The U.S. House Oversight Committee on Tuesday released the transcript of a closed-door interview in which Microsoft co-founder Bill Gates answered lawmakers’ questions about his ties to the late convicted sex offender Jeffrey Epstein.

Gates sat for the voluntary interview on June 10 in Washington, D.C., as part of the committee’s ongoing investigation into Epstein and his crimes.

In a statement Wednesday morning, a spokesperson for Gates said he appreciated the chance to appear before the House Oversight Committee and, as several committee members acknowledged, answered every question put to him over the nearly six-hour interview.

“With the full, unredacted transcript now publicly available, everyone can review the details for themselves,” the statement continued, reiterating that Gates “supports the full release of the files and hopes the Oversight Committee’s investigation will lead to justice for the victims.”

See the full transcript here and below, and continue reading for a summary of key points.

Bill Gates Transcript – U.S. House Oversight Committee by GeekWire

Meetings with Epstein

Gates described his association with Epstein as “one of the larger mistakes I’ve made,” saying he was foolish to spend time with him and that their interactions, from 2011 to 2014, were a “complete dead end.”

He said Epstein “certainly wasn’t a friend,” and that he declined Epstein’s social invitations — including to Epstein’s island — as Epstein tried to deepen the relationship.

Asked how often he saw Epstein, Gates gave this breakdown: three times in 2011, twice in 2012, and “five or six” times in each of 2013 and 2014, noting some of the 2013 contacts were Skype calls. He described the meetings as generally substantive rather than social.

Gates said that when he first met Epstein, at a January 2011 dinner in New York arranged by his former science adviser Boris Nikolic, he was aware Epstein had been convicted of a sex-related crime but had not looked into the specifics, acknowledging he “probably should have.”

He said it was not until 2018, when the Miami Herald detailed the extent of Epstein’s crimes, that he grasped their scope and learned Epstein had registered as a sex offender.

Gates said the primary reason he met with Epstein was Epstein’s claim that he could raise billions of dollars for global health from wealthy clients — money that never materialized. He acknowledged he also dealt with Epstein over a separate matter, the exit of his adviser Nikolic.

Gates said he was surprised to learn from the released files how extensive Nikolic’s own relationship with Epstein had been, and that reports Nikolic was named in Epstein’s will surprised him “a lot.”

He said he never witnessed Epstein engage in any sexual misconduct, was never offered any young women or girls, and never visited Epstein’s island, ranch, or Florida home.

He did acknowledge he “may have been in the presence of victims,” citing Epstein assistants he was photographed with and two who sat in the front cabin during a private New York-to-Palm Beach flight he took with Epstein — the one time, he said, that he flew with him. Gates said it was not Epstein’s 727, and he didn’t know who owned or chartered it.

Gates said neither he nor his representatives ever asked any victim to sign a nondisclosure agreement, secured any settlement, or held NDA discussions with victims or their lawyers regarding Epstein or Ghislaine Maxwell, the longtime Epstein associate who was convicted in 2021 of helping him sexually abuse underage girls.

Gates testified that Epstein flew to Seattle and visited his Gates Ventures office for a meeting focused on Nikolic’s departure — an encounter the committee dated to Aug. 8, 2013. Gates called it “kind of a worthless meeting.”

The next day, Gates emailed that Epstein had been “quite helpful,” but he told the committee he only “went along with the narrative” to close the deal, insisting Epstein’s involvement actually accomplished nothing.

Gates acknowledged making a $2 million donation to MIT during the period he knew Epstein, and said he told Epstein about it hoping to end Epstein’s requests that Gates give money in his name. He said MIT later investigated and found the gift was not Epstein-related.

He acknowledged three extramarital affairs — with a competitive bridge player, a nuclear scientist, and a doctor — and said Epstein had become aware of two of them, apparently through Nikolic.

However, Gates said, “I was not blackmailed,” characterizing Epstein’s notes as “emails to himself” that mixed true and false information and that he said he did not see until the Justice Department released the files. He allowed that the drafts looked like Epstein’s “brainstorming” heading toward blackmail.

Microsoft connections

Gates said the name Epstein “never came up” in his conversations with former Windows chief Steven Sinofsky, and that he learned of Epstein’s reported dealings with Sinofsky only through the press this year. (Sinofsky has declined to comment on the revelations and has not been accused of any wrongdoing.)

Regarding other Microsoft-connected figures, Gates said he never discussed Epstein with former CTO Nathan Myhrvold, though he had a “vague awareness of some connection” beforehand.

(Documents released in 2025 included an apparent letter and other materials from Myhrvold in Epstein’s 2003 “birthday book.” A spokesperson has said Myhrvold knew Epstein from TED conferences and as a donor to scientific research, doesn’t remember the letter, and regrets that he ever met him.)

As for LinkedIn co-founder and Microsoft board member Reid Hoffman, Gates said Epstein “may have come up” in conversation, that he’d had some prior awareness of a link through “some MIT connection,” and that both Hoffman and Epstein attended his final meeting with Epstein, a December 2014 breakfast. Hoffman has said he deeply regrets interacting with Epstein after his conviction and has called for full release of the files.

Other items

Rep. Lauren Boebert pressed Gates on Epstein’s interest in eugenics, transhumanism, and genetic engineering, asking whether Epstein ever discussed “genetic ambitions,” “population engineering,” or CRISPR-related DNA research with him, or tried to tie any of it to the Gates Foundation’s work. Gates said none of it ever came up and that Epstein had no influence on those initiatives.

At another point, pressed on whether he would support higher taxes on billionaires, Gates said he has paid “over $14 billion” in taxes and that the U.S. “has to find a way of taxing very rich people at a far higher level,” including himself.

Defending his foundation’s work, Gates said GAVI’s vaccine purchasing is “the primary reason childhood death has gone from 10 million a year down to below 5 million a year.” Separately, Gates said the foundation’s work “will be the focus the rest of my life.”

Microsoft announced Wednesday that over the past two decades, it has become dramatically more efficient in its use of water to cool data centers, slashing its consumption rate by 90% compared to levels when it opened its first facilities in the early 2000s. The company used 0.27 liters per kilowatt-hour last year, about three times better than the industry average.

Microsoft has for the first time replenished more fresh water globally than it withdrew, making important progress on its 2030 goal of being water positive across its operations.

And if this sounds familiar, you’re not wrong. Earlier this month, Amazon shared similar water usage stats (though it performed better) and Google came out with updated pledges around being water positive.

The tech giants are working to quench concerns about water use, which has become a key point of contention nationwide. Communities and local leaders are protesting and passing moratoriums on new data center construction. Other concerns include significant energy use that could drive up utility rates and noise complaints.

At the start of the year, Microsoft tried to get ahead of those fears by launching its Community-First AI Infrastructure initiative, in which it vowed to cover its electricity costs and forgo local tax breaks. Last week, it came out in support of the Ratepayer Protection Act, a congressional measure addressing data center utility bill impacts, though it earlier opposed Washington state legislation targeting some of the same concerns.

Microsoft remains “deeply committed” to water protections, said Judy Priest, CTO of Cloud Operations & Innovation, and Steve Solomon, vice president of Datacenter Engineering, in a blog post Wednesday.

“We continue to advance datacenter innovations that reduce water use intensity while supporting the growing performance demands of cloud and AI services,” Priest and Solomon said.

Data centers use a variety of strategies to keep electronics cool, including fans, evaporative cooling, air conditioning and direct liquid cooling. The approaches involve tradeoffs: air conditioning draws more electricity but saves water, while evaporative cooling is less energy-intensive but consumes more.

Microsoft’s approaches to curb its water use include:

• Cooling primarily with fans, supplemented by evaporative cooling when outside temperatures exceed 85 degrees.
• Using chip-level cooling that recirculates water through the system.
• Auditing data centers to ensure facilities are operating as designed and conserving water optimally.
• Expanding its use of recycled, reused or non-potable water.

Comparing companies on this front is tricky. Microsoft’s liters-per-kilowatt-hour figure applies only to data centers it owns, while Amazon’s includes both its own computing facilities and leased ones.

Though Microsoft has made strides on the goal set in 2020 of becoming water positive within a decade, that benchmark takes a global tally of water use and replenishment. In theory, that means water used in a desert climate could be offset by Microsoft’s actions in a wetter region as regards its overarching target. The Community-First AI Infrastructure initiative, however, pledges to replenish more water than it uses in each district where it operates AI data centers.

That aligns with the approach used by Amazon and Google, though Amazon’s replenishment goal covers only data centers, not all of its operations.

While concern about data center water use is growing, it remains relatively modest in the broader context: data centers account for about 0.5% of all industrial water use worldwide, as Amazon recently noted.

In terms of total volume, Microsoft withdrew 2.7 billion gallons of water in fiscal year 2024 across its data centers and its other operations. For context, Seattle Public Utilities delivers roughly 43 billion gallons each year to 1.6 million people in its service area.

Editor’s note: Story updated to clarify that progress on the 2030 water positive goal is ongoing.

Microsoft is promising relief to engineers who get woken up at 3 a.m. for outages and other cloud glitches: an agent informed by its years of experience running Azure, designed to diagnose whatever’s going wrong and recommend potential fixes. 

One big benefit over humans: the agent can operate without the stress, fatigue, or tunnel vision that often hampers people doing it on little sleep.

“Agents are a little bit less emotionally attached,” said Brendan Burns, a Microsoft technical fellow and corporate vice president who was one of the creators of Kubernetes. He pointed out that agents don’t feel the pressure when a manager asks for a rapid root-cause analysis.

The Azure Copilot Observability Agent, in preview since late last year, was made generally available Tuesday. It investigates incidents by connecting the logs, metrics, traces and other signals scattered across a company’s systems, then points engineers toward the likely cause. 

At this point, the agent does not fix problems on its own. Microsoft also introduced what it calls autonomous operations, in preview, letting the agent triage and investigate alerts without a person prompting it. But it still stops short of acting. It won’t restart a resource or change a configuration, for example, instead leaving it to humans to decide and execute. 

Microsoft is joining a crowded field. Datadog made its Bits AI SRE agent generally available in December, and Amazon’s AWS followed with a comparable DevOps Agent this spring. Microsoft said the agent is priced based on usage rather than a flat per-seat license, which is the same model AWS uses for its DevOps Agent. 

Established observability players including Dynatrace, Splunk, New Relic and Grafana are moving quickly in the same direction, alongside a wave of AI-focused startups. 

In an interview with GeekWire this week, Burns said he believes Microsoft’s breadth is one of its advantages, seeing more of a customer’s software than rivals do, from GitHub to Azure deployments to the signals systems generate. Knowing how those connect, he said, helps the agent trace a problem back to the line of code behind it. 

More than a decade ago, Burns and his then-Google colleagues Joe Beda and Craig McLuckie created Kubernetes, the open-source software that lets companies run applications across large, constantly changing infrastructure. It became foundational to cloud computing, and added to the complexity teams now have to manage.

Kubernetes brought a kind of self-repair to that world: when something breaks, it works automatically to restore the system to a healthy state. But it follows fixed rules, Burns said. It’s “very deterministic” — it “can’t make hypotheses, it can’t investigate solutions.” 

AI tools like the Azure observability agent are meant to add that missing layer: forming a theory about what went wrong, testing it against the data, and continuing to work to find a solution. 

Full autonomy — letting the agent act, not just investigate — is still down the road. In a blog post Tuesday, Burns framed the launch as part of a broader shift toward “agentic operations,” which reason across signals and will someday be able to act on them. 

For now, the agent can do a lot of the digging, even if a human still makes the call. 

Burns, who recalled once pulling a 36-hour on-call shift, said he can think of “a lot of late nights that would have been a lot nicer if I’d had this 10 years ago.”

Microsoft has spent years subsidizing Xbox rather than profiting from it, CEO Satya Nadella acknowledged this week, as he addressed the gaming division’s need for a new approach. 

His comments came during a Wednesday evening taping of The New York Times’ “Hard Fork” podcast, released Friday. Hosts Kevin Roose and Casey Newton pressed Nadella on the future of Xbox a few hours after the division’s leadership signaled an upcoming reset.

“No one can accuse Microsoft of not having invested for the last 25 years,” Nadella said of the Xbox and games business. “And now we have to turn this into a sustainable business.” 

For all the entertainment value Xbox provides, he said, Microsoft hasn’t been monetizing that entertainment, and has actually been subsidizing it. He added with a chuckle, “In fact, there’s more monetization of Xbox games happening on YouTube than at Microsoft.”

Earlier in the day, Xbox CEO Asha Sharma had told employees in a memo that the division’s heavy spending and declining revenue cannot continue. Sharma, about 100 days into the job, said Xbox will finish the fiscal year at roughly a 3% margin by an internal Microsoft measure, after the company spent more than $20 billion over five years even as annual revenue fell.

Bloomberg News reported that the division is planning major job cuts next month. 

On the podcast, Nadella described two pressures on the business. One is temporary: a run-up in prices driven by the shortage of semiconductors and memory, which is squeezing PCs, phones and other consumer electronics, and which he said Microsoft will get through. 

The other is lasting — the question of what the Xbox business model should be going forward. 

“I think we have to find ways to deliver the games in which it’s economically relevant for the customer and for us,” Nadella said when Newton asked whether he could offer any sort of “carrot” for gamers, or whether consoles and games would simply get more expensive. 

Nadella didn’t detail what the new model would look like. Sharma said in her memo that she’ll spend the next 100 days taking what she called a fresh look at the business.

The Information reported Friday that Microsoft hasn’t ruled out restructuring Xbox — potentially as a wholly owned subsidiary, a joint venture, or a spin-off — though it has no imminent plans to do so. The outlet, citing three people with direct knowledge, said Sharma plans to pair layoffs with heavier investment in big franchises like Halo and Fallout, a plan Nadella and CFO Amy Hood have signed off on.

See above for the full conversation, which otherwise focuses largely on artificial intelligence, including the AI backlash over data centers, AI’s impact on jobs, whether the U.S. government should take stakes in AI companies, and how much he buys the idea that AI is about to automate entire jobs.

Xbox CEO Asha Sharma, roughly 100 days into her tenure, delivered a blunt assessment of Microsoft’s gaming business in a memo to employees Wednesday, saying that heavy spending with thin profit margins and declining revenue “cannot continue.” 

The memo, posted publicly on the Xbox blog, came as Bloomberg News reported that the division is planning major job cuts next month, soon after the close of Microsoft’s fiscal year on June 30. Xbox is also planning significant cuts to marketing and other budgets, according to the report.

The exact scale of the layoffs is not yet clear. Microsoft declined to comment. The Verge also reported that Xbox “will be hit with significant layoffs next month,” citing people familiar with the plans.

Sharma’s memo did not mention layoffs but described a business that needs a sweeping reset. She and Xbox content chief Matt Booty, who co-signed the memo, cited rising hardware component costs, an overextended studio system, and aging platform infrastructure among the challenges facing the division.

Xbox will end the fiscal year at about a 3% “accountability margin,” an internal metric Microsoft uses to measure the profitability of the business, according to the memo.

“Excluding Activision Blizzard King, over the past five years, we have spent over $20 billion on ongoing investments in our content, platform, and hardware subsidy, but our annual revenue has declined nearly half a billion during that time,” Sharma and Booty wrote. “Going forward, this cannot continue.”

Microsoft’s most recent quarterly filing illustrates the challenge. Gaming revenue fell 7% to $5.3 billion in the quarter ended March 31, with Xbox hardware revenue down 33% on lower console sales, and Xbox content and services revenue down 5%.

The memo follows Sunday’s Xbox Games Showcase, where Sharma reversed course on the company’s multiplatform strategy, announcing that Gears of War: E-Day and Clockwork Revolution will be Xbox console exclusives. Bloomberg reported Wednesday that a PlayStation 5 version of the new Gears of War game had been in development, and was canceled, before the announcement.

Sharma took over in February from Phil Spencer, the longtime Xbox leader who announced his retirement after 38 years at Microsoft. A former Instacart COO and Meta product executive, she previously ran Microsoft’s CoreAI product organization.

Bill Gates appeared voluntarily behind closed doors before House Oversight Committee investigators on Wednesday to answer questions about his ties to Jeffrey Epstein, part of a larger turn in his image that has been rippling through the Seattle-area institutions he built and funds.

In a statement in advance of his testimony, the Microsoft co-founder gave his most detailed public account yet of the situation, saying he was introduced to Epstein in 2011 on the promise of raising billions for global health and cut off contact in late 2014. He said he should never have met with Epstein, calling the decision a “grave error in judgment” that put his work in global health and philanthropy at risk.

“If the time I spent with Epstein lent him any credibility, I am deeply sorry,” he said. “I have learned a significant lesson and am now far more careful about who I engage with even in a limited capacity.”

Gates also addressed the marital infidelity that surfaced in the files, saying Epstein learned of it and tried to use it as leverage to draw him back in, without success. Gates acknowledged two affairs with Russian women at a Gates Foundation town hall in February, the Wall Street Journal previously reported.

A Gates spokesperson said via email that he “welcomes the opportunity to appear before the Committee. While he never witnessed or participated in any of Epstein’s illegal conduct, he is looking forward to answering all the committee’s questions to support their important work.”

The Gates Foundation, which he chairs, said it commissioned an external review earlier this year to assess its past engagement with Epstein and its vetting of new partners. CEO Mark Suzman told employees at a February town hall that he felt “somewhat sullied” by the association and that it had made the foundation’s mission more challenging, according to the Journal.

The foundation has said it “regrets having any employees interact with him in any way.”

TerraPower, the Bellevue-based nuclear energy company Gates founded and chairs, held an all-hands meeting where employees raised questions about Epstein’s ties to both Gates and Nathan Myhrvold, the former Microsoft chief technology officer, who co-founded TerraPower and serves as its vice chairman.

The Journal also reported that Breakthrough Energy, his climate-investment firm, has struggled to raise a new fund, with some investors citing the Epstein connection.

Gates did not take part in Microsoft’s CEO Summit this year, and a dinner he typically hosts at his home in connection with the event did not take place. “While it didn’t work out this year, we’ve already extended an invitation for Bill to attend the CEO Summit next year,” a Microsoft spokesman told the Journal.

On Capitol Hill today, Gates appeared for a transcribed interview rather than public testimony, and a transcript is expected to be released in the coming days. He has not been charged with a crime or accused of wrongdoing by Epstein’s victims.

The interests of Microsoft and graduates rebelling against AI are actually aligned. 

That was one takeaway for Brad Smith, Microsoft president and vice chair, from a recent return to his alma mater, Princeton University, for its reunion weekend. Seniors wore class jackets labeled “100 percent cotton” and “100 percent human,” referencing allegations that an earlier design was created with AI — part of a broader backlash across campuses this spring.

In a blog post this morning, which he started drafting during that visit, Smith writes that graduates booing AI at commencements across the country are “telling us what we need to hear.” He points out that Microsoft’s own future depends on people staying employed. 

“Workers have been Microsoft’s lifeblood from the start,” he writes in the post. “If the world’s people don’t have jobs, then neither do we. And if we’re not doing our part to help people use technology to pursue better jobs, then we’re not doing the job we were born to do.”

Speaking with GeekWire this week, Smith acknowledged the tension between that message and job cuts across the tech sector, including at Microsoft. He addressed the issue in the post, as well, citing the industry’s desire to offset capital spending on AI, along with factors including geopolitical uncertainty, trade tensions, and a correction from earlier over-hiring.

“Our industry is going through one of the most extraordinary transformations in its history,” Smith said in the interview, while adding that the “expenses of capital expansion make it more difficult to afford the employment bubbles we’ve had, especially since 2020.”

Smith cited the automation of entry-level tasks among the challenges facing graduates, as well.

But he also took a larger view. Computer science jobs are changing, he said, not vanishing. Coding is becoming a smaller part of the work, while the roles around it — including designing software, managing product development, and reviewing code — are expanding. 

In the post, Smith places AI in a longer line of technologies that reshaped work without ending it, from the camera to the spreadsheet to email. He calls AI the next “general purpose technology,” akin to electricity, and argues its spread will take decades, not years, because the limit is how fast people and institutions change, not how fast the models improve. 

Some jobs go away, he writes, while new ones appear, and many are remade. 

Smith’s advice to workers is to treat a job as a bundle of tasks rather than a title, sorting them into what AI can do, what a person can do with AI, and what only a human can do. For this, he takes inspiration from a new book by LinkedIn’s Ryan Roslansky and Aneesh Raman, “Open to Work,” and its list of durable human attributes: curiosity, creativity, compassion, communications, and courage.

The post also offers a clear message for companies, aligning with Microsoft’s own business interests. Smith says organizations need to build their own AI systems on top of frontier models, using their own data and what Microsoft CEO Satya Nadella calls a “hill climbing machine” of evaluations and steady improvement, rather than simply renting intelligence from someone else.

Smith cites intellectual property and data sovereignty as a central concern, arguing that firms must adopt AI without handing their hard-won expertise to a rival’s model. 

In the interview, Smith said the blog reflects months of discussion among Microsoft’s senior leaders, including Nadella and Chief People Officer Amy Coleman, and that it’s intended to speak to the company’s own employees as much as to the outside world. 

Asked what he would have told new college graduates had he been the speaker at a commencement ceremony this spring, Smith said he would have focused on the resilience of humanity more than advances in technology — urging them to speak up for the values they care about, help contribute to a better world, and go forward with hope and optimism.

“That doesn’t mean these challenges may not be significant,” he said, “but I personally believe that the human spirit is far greater than any artificial intelligence the world is likely to create.”

Microsoft released a new pre-recorded Xbox Showcase on Sunday morning as part of this year’s Summer Game Fest event, which also marked new CEO Asha Sharma’s first big public event since taking over the company’s gaming division.

Back in February when Sharma took over Xbox, some analysts, including me, openly wondered if she was there to shut down the department. Instead, Sharma appears determined to give Xbox a shot in the arm, telling Bloomberg News earlier this week that she aims to make Xbox “the number one gaming and entertainment company” by 2030.

For a product that seems to be permanently stuck in third place behind Sony and Nintendo, and which is facing at least one significant consumer boycott, that’s a frankly awe-inspiring level of ambition. That set up high expectations for this year’s Showcase.

Instead, Sharma and Xbox chief content officer Matt Booty seemed content to let their games do the talking. The focus of this year’s hour-long Showcase was firmly on new and upcoming releases from the Xbox studio network and its partners, as part of a low-key celebration of the Xbox project’s 25th anniversary.

The Showcase began with a new look at gameplay for the forthcoming Gears of War: E-Day (Oct. 6). The Gears of War series has, since the beginning, been focused on the conflict between humanity and a subterranean species called the Locust Horde.

E-Day is a prequel set on the first day of that conflict (“Emergence Day”), 14 years before the original Gears of War. It once again puts the player in the role of the series’ traditional protagonists Marcus Fenix and Dom Santiago.

We’ve known that E-Day was coming for a couple of years, and Gears has traditionally been one of the bigger franchises in Xbox’s network. The big surprise here isn’t the game itself, but rather the quiet announcement that E-Day is an Xbox console exclusive.

This is a big reversal of policy from Microsoft, which made headlines over the last couple of years by deliberately publishing several of its first-party games on competitive platforms such as the PlayStation 5. While this appeared to be financially successful for the company, it’s also traditionally been the kind of move that video game companies (i.e. Sega) do right before they leave the hardware market.

Now Xbox is at least attempting to chart a new course. Both E-Day and the forthcoming steampunk action-RPG Clockwork Revolution (2027) were specifically identified as Xbox console exclusives. While several other first-party Xbox titles weren’t, including the Fable reboot and Halo: Campaign Evolved, any move towards console exclusivity is a big departure for modern Xbox.

The company later specifically confirmed via Xbox Wire that E-Day and Clockwork are not timed exclusives. For the foreseeable future, if you want to play either of these games on a console, you’ll have to own Xbox hardware to do so.

This suggests that Sharma’s Xbox may be moving back towards more proven market strategies for the platform, as opposed to the Spencer/Bond tactic of attempting to redefine the terms of success or the product itself.

Fable may be the next biggest news out of this year’s Showcase, as it’s been suspected of being vaporware for several years now. The original games were some of the biggest Xbox exclusives, as famously open-ended fantasy RPGs that allowed you to play as a hero, a villain, or something in between.

The series has been on hiatus since Fable III in 2010, so Microsoft got fans’ attention back in 2020 when it announced plans for a reboot. Then, nothing happened for quite some time. Every major press event at Xbox would feature some small piece of information about Fable, just as proof of life, before the project vanished again.

Now we actually have a release date for the new Fable: Feb. 23, 2027. In addition, the Showcase trailer marks the debut of Fable’s villain, Isabel, who’s played by British actress Hayley Atwell (Captain America).

Speaking of projects that seemed like they’d never come out: this year’s Showcase featured a new trailer for Seattle-based Undead Labs’ State of Decay 3.

This co-op zombie survival game, set in the post-apocalyptic Pacific Northwest, has been in development for years, but a 2022 scandal about its toxic work culture nearly sank Undead Labs before it could be released. It’s frankly shocking that Microsoft never pulled the plug. Instead, State of Decay 3 is coming in 2027.

This summer, we’ll finally see the next project from the rebranded Halo Studios, as the Unreal Engine remake of the original Halo is coming on July 28. In addition to the remake ofHalo’s story,Campaign Evolved will feature three new missions set one year before the game’s events, which team the Master Chief with fan-favorite character Sergeant Major Avery Johnson.

Other first-party news out of the Xbox Showcase includes:

• To celebrate the Xbox’s 25th anniversary, Microsoft plans to release a new edition of the Xbox Series X in November which features a translucent green shell and a matching controller. This is meant to recall the limited-edition Halo Xbox that released in 2004.
• This year’s Call of Duty is a fourth entry in its Modern Warfare subseries. It features a new mode, DMZ, set in the fictional Hajin Exclusion Zone in Korea, which is the setting for a high-stakes extraction shooter.
• Mojang Studios’ Minecraft Dungeons II is coming out on Sept. 29.
• Activision plans to revive the Spyro series, featuring one of the most recognizable mascots of the 2000s, with A Realm Beyond, coming in the spring of 2027.
• Ninja Theory’s award-winning Hellblade series is getting a third entry next year, simply titled Senua.
• Last year’s Doom: The Dark Ages will receive a DLC expansion, Revelations, on July 7.
• Microsoft Flight Simulator 2024 has a new “world update” planned for July 4, which is meant to “celebrate the beauty of the United States” by adding multiple national parks to the game’s environments. Now you can fly through an ultra-realistic Grand Canyon or over the Dry Tortugas.

Announcements from Xbox partners included:

• Atlus debuted a remake of its popular Japanese urban-fantasy RPG Persona 4 and, after a long period of silence on the topic, has confirmed it’s making Persona 6. While nothing is known about P6 aside from its existence, that’s still new information.
• As a celebration of the Castlevania series’ 40th anniversary, Konami plans to publish Belmont’s Curse, the first new entry in the core series since 2008’s Order of Ecclesia. It follows Rose Belmont as she and her father Trevor investigate a monster attack on the streets of Paris in 1499. It’s due out on Oct. 15.
• The French developer Asobo Studio revealed another trailer for its forthcoming prequel Resonance, which is a new entry in the Plague Tale series. These action-horror games, set in 14th-century France, deal with a supernatural take on the Black Plague and its protagonists’ attempts to survive it.
• Sega plans to revive its Crazy Taxi racing series with a new entry, World Tour, coming next year. Within minutes of the reveal, fans noticed that World Tour‘s Steam page features an AI-usage disclaimer, which has already begun a mild furor on Western social media.
• Koei Tecmo and Team Ninja revealed Wo Long 2, a sequel to its fantasy action-game, set during China’s Three Kingdoms period.
• Serenity Forge, an indie studio that specializes in both weird horror and “cozy” games, seems to be splitting the difference with its newest project. Vivarium is an anime-inspired, hand-animated game about life inside a small town… which is contained in a glass jar in what appears to be an abandoned house. It’s due out next year.

Reid Hoffman, the LinkedIn co-founder who has served on Microsoft’s board since 2017, will not stand for re-election at the company’s 2026 annual meeting, ending a board tenure that coincided with some of the most consequential years in the tech giant’s history.

Hoffman informed Microsoft on June 2 that he will not stand for re-election at its 2026 annual shareholder meeting, according to an SEC filing. He will remain a director until that meeting, later this year.

The filing states that Hoffman’s decision was not the result of any disagreement with management over the company’s operations, policies or practices. Microsoft thanked him for his contributions.

On an episode of his “Possible” podcast, released Friday morning, Hoffman told Microsoft CEO Satya Nadella that he wanted to return to “founder mode” to focus on Manas AI, the cancer drug-discovery startup he co-founded, citing early progress there.

Nadella thanked him for his board service and said he was eager to see what Hoffman builds next. “I am so grateful for all of your contributions to Microsoft and the board over the years, and excited to see you get back to founder mode with Manas,” Nadella wrote in a LinkedIn comment.

Hoffman joined the board after Microsoft’s $26.2 billion acquisition of LinkedIn, the professional network he co-founded in 2003. A longtime Greylock Partners investor and prominent Democratic donor, he served on the board’s Environmental, Social and Public Policy Committee.

His exit follows three consecutive years of opposition from the National Legal and Policy Center, a conservative shareholder group that urged investors to vote against his re-election, citing his political activities and other concerns. Shareholders re-elected him each time, most recently in December 2025.

Hoffman’s name has also surfaced in the long-running fallout over the late financier Jeffrey Epstein. Hoffman has said his dealings with Epstein were limited to fundraising for MIT’s Media Lab, including a 2014 visit to Epstein’s island that he has said he regrets, and he has not been accused of wrongdoing.

Documents and emails released in the federal Epstein files indicated more contact than he had previously acknowledged. In November 2025, the Justice Department said it would investigate Epstein’s ties to Hoffman and other Democrats after President Trump called for a probe. Hoffman has called the scrutiny baseless and politically motivated.

Hoffman has been one of Silicon Valley’s most active figures in AI, both as an investor and a hands-on experimenter. He was an early investor in OpenAI and sat on its board until resigning in March 2023, citing potential conflicts with his AI investments at Greylock and his role as a founder of Inflection AI, the startup he co-founded in 2022 with DeepMind co-founder Mustafa Suleyman.

Microsoft hired Suleyman and most of Inflection’s staff in March 2024, installing Suleyman as CEO of a new Microsoft AI division, and struck a licensing deal with Inflection reportedly worth $650 million.

Editor’s note: This story was updated to add background on Hoffman’s ties to Jeffrey Epstein.

The token-hungry developers were there. Nvidia CEO Jensen Huang was there (virtually). The Chainsmokers were there. But Microsoft’s rumored Copilot “Super App” was not.

According to various reports and screenshots posted on social media over the past couple of days, Microsoft’s Copilot Super App was ready for its close-up. Reports indicated that the Copilot Super App is meant to provide a single Copilot experience, or shell, with various modes, possibly including a Copilot Chat mode; GitHub Copilot coding mode; Cowork mode for knowledge workers and prosumers; the “Scout” OpenClaw-based work mode; and some kind of Autopilot always-on agent mode.

Some had expected Microsoft could make the Copilot Super App its “one more thing” announcement during the kick-off keynote at the Microsoft Build 2026 conference on June 2. But the app, in whatever form it currently may exist, was a no-show.

It wasn’t a total wash, however. CEO Satya Nadella did mention the Super App in passing.

“Come summer, we will be bringing coding to all knowledge work within one Copilot Super App. That’s really exciting. So you’re going to have Chat, Cowork, and Code all in Copilot,” Nadella told the Build audience in San Francisco.

The premise of a Copilot super app makes sense on several fronts. Microsoft is looking for a way reclaim its early-mover position in AI coding that it carved out with GitHub Copilot. The company needs an answer to the growing popularity of Anthropic’s Claude Code and OpenAI’s Codex. And given Microsoft is working to unify its consumer and commercial Copilot experiences, an all-in-one Copilot workspace could provide a neat solution.

Jacob Andreou, recently appointed executive vice president of Copilot, is charged with this unification and reports directly to Nadella as part of a small team replacing long-time head of Microsoft’s Experiences and Devices unit, Rajesh Jha.

Andreou has what I’d consider a daunting task. And not just because he is based in Los Angeles and came to Microsoft just a year and a half ago via an unconventional path (Snap and then Greylock Partners).

Microsoft originally tried to position its various Copilots as a single product, even though they used different data sources, had different interfaces and provided different types of access.

More recently, officials acknowledged this and made distinctions between consumer Copilot, GitHub Copilot and Microsoft 365 Copilot for businesses as separate, but related offerings. But now the company seems to be veering back toward trying to make Copilot seem like a single entity in terms of brand and across consumer and enterprise lines.

Last week, Microsoft took a step toward improving the Microsoft 365 Copilot user experience with a redesign, which made the prompt box bigger and results appear more quickly. But it didn’t go so far as to show off how the new UI will dovetail with the coming Super App.

A couple of the supposed elements of the Super App did get airtime at Build. Scout, which Microsoft describes as a “personal agent for work” is built on the open-source OpenClaw framework. Scout can access data in Microsoft apps like Teams, Outlook and SharePoint thanks to Microsoft’s WorkIQ context layer, so it can proactively handle tasks such as prepping for meetings and fixing scheduling conflicts without having to ask users for approval.

Microsoft is making Scout available to its customers in its “Frontier” testing program starting today, June 2. Up until now, it’s been in testing inside Microsoft.

If Scout sounds familiar, it should. Scout is the official name of the OpenClaw skunkworks project that’s been the focus of Microsoft Corporate Vice President Omar Shahine and team (profiled here on GeekWire last month). Microsoft has been working to add guardrails around OpenClaw and Scout to try to allay security fears that many enterprise companies, including Microsoft itself, have expressed about OpenClaw’s always-on way of operating.

Scout is considered the first public example of this new category of always-on agents that Microsoft is calling “Autopilots.”

After sitting through the three-hour (!) Build opening keynote, I was left wondering why Microsoft didn’t show off, even fleetingly, the coming Copilot Super App.

Was it because execs felt they had so many other announcements that they didn’t want it to get lost in the mix? They’re waiting for the “Ask Copilot” taskbar feature to go live on Windows 11? Or maybe the Super App is just not yet stable enough to demo? (Given how quickly Microsoft is moving from idea to private testing with Scout, making sure a product is baked before showing it publicly doesn’t seem to be much of a concern at Microsoft.)

Sure, the pressure is on with AI to announce or be eclipsed, like never before. And Microsoft is no stranger to “creatively architected” demos of not-yet-finished products. (I see you, Longhorn.)

But can Microsoft really move from pilot to shipping products at this pace and not alienate enterprises that have substantial security, compliance, data-residency and other hefty requirements? I guess we’ll see….

Microsoft has based much of its AI business on models from OpenAI, before expanding more recently to Anthropic. On Tuesday, the company showed how it plans to rely less on both.

At the Build developer conference, the Microsoft AI Superintelligence Team unveiled a family of seven models built from scratch. It’s part of an ongoing effort by the company to build credible in-house alternatives to models from partners and rivals with competing allegiances.

“This is all about long term self-sufficiency for Microsoft and our partners. It’s about models you can trust,” wrote Mustafa Suleyman, CEO of Microsoft AI, in a post announcing the models.

Microsoft is OpenAI’s largest backer, having invested a cumulative total of $13 billion in the ChatGPT maker over multiple funding rounds. The company last year announced an investment of up to $5 billion in Anthropic, and later integrated its technology into a Copilot Cowork AI assistant.

However, Anthropic is also backed by Microsoft rivals Google and Amazon, and OpenAI is increasingly cozy with Amazon — showing the need for Microsoft to control its own AI destiny.

The flagship of the seven newly announced MAI models is MAI-Thinking-1, a reasoning model that Microsoft says draws even with Anthropic’s Claude Sonnet 4.6 in blind human testing, and matches the more capable Claude Opus 4.6 on a widely used coding benchmark.

Suleyman stressed that MAI-Thinking-1 was trained from the ground up with no distillation from other companies’ models, looking to appeal to enterprises that care about clean data lineage.

It’s available in private preview on Microsoft Foundry, where the company also hosts the latest models from OpenAI and Anthropic, including the recently released Claude Opus 4.8.

Microsoft AI also released MAI-Code-1-Flash, a 5-billion-parameter coding model now rolling out in Visual Studio Code and GitHub Copilot, and MAI-Image-2.5, which Microsoft says ranks second on a leading image-editing leaderboard, ahead of Google’s Nano Banana Pro.

The full set of models spans image, voice, transcription, coding and reasoning.

[Editor’s Note: Agents of Transformation is an independent GeekWire series, underwritten by Accenture, exploring the adoption and impact of AI and agents. See coverage of our related event.]

A team inside Microsoft has been quietly building a platform for devices that run AI agents instead of apps, based on Android instead of Windows, with two working hardware designs so far, and an initial set of big-name companies lined up to run pilots.

The platform, dubbed “Project Solara,” is Microsoft’s bet that AI will open up entirely new scenarios for computing — using agents to avoid the constraints of traditional software, and off‑the‑shelf components to develop new devices quickly and inexpensively.

Microsoft is racing against Google, Amazon, OpenAI and others to bring AI to devices and provide the technical backbone for a new generation of computing. In effect, the company is attempting to repeat with AI what it did for personal computers five decades ago, with much stiffer competition this time but also far greater technical freedom.

“Boundaries are collapsing,” said Stevie Bathiche, the Microsoft corporate vice president and technical fellow who leads its Applied Sciences Group. “You don’t necessarily need the traditional app model. You don’t need the traditional way of developing experiences.”

The company unveiled Solara on Tuesday at its Build conference in San Francisco, describing it as a new platform that spans from chip to cloud. GeekWire got a behind-the-scenes look at the project during a briefing last week in Redmond, including demos of the first two concept devices based on the platform:

• A desktop hub that sits beside a PC and responds to voice commands, signs users in using facial recognition, and surfaces the day’s most pressing items. With a monitor attached, it becomes a full Windows machine running in the cloud.
• A wearable badge that reimagines the standard employee ID card. A fingerprint button wakes an agent in one press; a single tap records and transcribes a conversation; and a built-in camera lets the agent act on what the user sees.

Microsoft says it won’t ship these devices itself. Instead, it envisions hardware makers and other industry partners turning the reference designs into implementations of their own, each intended for a specific industry, company, or scenario.

For example, in one demo shown by the company, the high-tech badge ran on agents designed for use by a health-care worker, including the ability to scan a patient’s QR code, record and transcribe the visit, log vitals, and start a prescription.

In another application of the same badge, the built-in camera scanned a brainstorm board with ideas for an office revamp, and made a suggestion: add some plants.

The two devices are a starting point. The bigger opportunity, the company says, is all the tasks and workflows where a PC or phone gets in the way or isn’t practical to use. 

A display inside the Microsoft Applied Sciences lab gave a hint of where things could be headed, including smart glasses, rings, earbuds, scanners, and other form factors.

“This is a way to put computing in those spaces easily and cheaply, but more importantly, it’s a way to put your agent into those spaces,” Bathiche said.  

In the coming months, companies including AccuWeather, Best Buy, CVS Health, Levi’s, and Target are expected to begin pilots of devices based on the reference designs.

The operating system is the Microsoft Device Ecosystem Platform, or MDEP, an enterprise version of Android that Microsoft developed for devices including Teams meeting-room hardware.

The company says it chose MDEP over Windows deliberately, to run on smaller, lower-power devices while keeping the management and security features IT departments expect: patch and over-the-air updates, device integrity, Microsoft Defender, Intune, and Entra ID sign-in.

What’s different

At first glance, the concept devices raise a couple of natural questions: 

1) Why not just use a phone? Bathiche said companies have tried, particularly in healthcare, and it didn’t go well. Asking a nurse to pull up patient data on a personal device felt wrong to patients and created security problems. 

A purpose-built device, he said, has a far smaller attack surface, can last a week on a single charge, and can orient its camera for face-to-face interaction rather than forcing the user to hold up a screen. 

“Computers are continuing to specialize,” he said, describing a trend he has been calling out for years now. “Computers are continuing to come closer to you.” 

2) Isn’t the desk device basically an Amazon Echo? Here, Bathiche drew a distinction: Alexa is one agent trying to do everything, while Solara is designed for each organization’s own agents, secured and managed by its IT department. 

The practical difference was visible in the demo. The desk hub pairs with a PC over Bluetooth, hands off tasks between the two, and keeps them locked in sync. An Echo Show sitting next to the same PC wouldn’t know it was there.

Pushing the timeline

Still, the project is very early, by Microsoft’s own admission. Bathiche said CEO Satya Nadella liked what the team was doing and suggested showing it at Build this week, much sooner than the company would normally show its behind-the-scenes work in public.

That underscores just how competitive and fast-moving the AI world is right now, but it also illustrates the pace that the new technologies are enabling. For example, Bathiche said the team got the badge running on the platform in about three days, using the same software as the desk device on a different chipset from a different company. 

Yet some fundamental details still need to be figured out. Asked by GeekWire about the business model for the platform, Bathiche pointed to one clear piece: the devices run on Microsoft’s Azure cloud. Beyond that, he said, the economics are still taking shape. 

Even the potential scenarios are in the preliminary stages. For example, Bathiche said the healthcare demo was designed to illustrate the concept, not to serve as an actual clinical tool.

The devices can run multiple agents at once, with a coordination layer that taps whichever agent a task requires. Microsoft offers its own agents, including Microsoft 365 Copilot, but the platform is designed for organizations to use other agents, as well.

Qualcomm and MediaTek are the first chip partners. The badge runs on a new Qualcomm wearable chip; the desk hub runs on MediaTek IoT silicon. Both are off-the-shelf, not custom, which is central to how Microsoft plans to keep devices cheap and fast to build.

Notably, OpenAI’s reported AI-agent phone is also being developed on MediaTek and Qualcomm silicon, underscoring the competition emerging in this category.

For Bathiche, Solara is a bet on what the next computer looks like. “What is the next thing that comes closer to you?” he asked. That, he contends, is where computing is ultimately going.

More about Project Solara on this week’s GeekWire Podcast:

Yusuf Mehdi, one of Microsoft’s best-known and longest-serving business leaders, whose tenure has spanned 35 years from Windows 3.1 to Copilot, plans to leave the company after one more year — his “final season,” as he called it in an interview.

Mehdi, 59, is Microsoft’s EVP and consumer chief marketing officer, overseeing product marketing for Windows, Surface, Copilot, Microsoft 365 consumer, Edge, and Bing search engine. He announced his plan Thursday, saying he intends to work at full intensity through the next fiscal year, ensuring succession plans are in place, before stepping away from the company that has been, as he put it, “the canvas of my life’s work.”

He compared it to picking a ship date for a product, something he’s had a lot of experience with during his time at the company. You put it on the calendar, and you work toward it.

“There will be time later to reflect and celebrate, but for now, it’s full speed ahead on our mission,” Mehdi wrote in an internal email to his team on Thursday afternoon. 

After that? He’s not sure, but he’s not calling it retirement. He said he still feels young and energetic, and he has some ideas in mind, but he emphasized that he hasn’t made any plans. 

Mehdi said he has been working closely with Microsoft CEO Satya Nadella and Chief Marketing Officer Takeshi Numoto on the transition plan. Microsoft has not named a successor, and he said it’s too early to determine what the leadership structure will look like after his departure. 

His priorities over the next year will include positioning Windows for the agentic era, unifying Microsoft Copilot as a seamless experience across work and personal lives, and scaling the Microsoft 365 consumer business as it approaches 100 million subscriptions.

His work on Windows, in particular, is “a little poetic,” he said, since that’s where he started.

Mehdi’s career at Microsoft has touched nearly every major consumer product the company has shipped. He started as an intern in 1991, in his mid-20s, after a two-year stint at Reuters, where he worked on computer-based products for the foreign exchange trading business. 

He joined Microsoft full-time in 1992 after earning his MBA from the University of Washington, adding to a bachelor’s degree in economics from Princeton. 

After his early work on the launch of Windows 3.1 and Windows 95, Mehdi went on to help launch and market Internet Explorer during the browser wars. He spent more than a decade running Microsoft’s online services and search businesses, helping build Bing and playing a central role in the company’s search and advertising partnership with Yahoo.

In 2011, he moved to the Xbox division, where he helped launch the Xbox One and forged the NFL partnership that put Surface tablets on the sidelines. He later oversaw the rollout of Windows 10 to more than 1 billion devices, and took on responsibility for the Windows and Surface businesses after Panos Panay’s departure in 2023.

That same year, he was promoted to his current role on Microsoft’s senior leadership team, becoming the public face of the company’s consumer AI push, from the launch of the AI-powered Bing search engine to the marketing of Microsoft Copilot.

Mehdi plans to stay through the end of Microsoft’s next fiscal year, June 30, 2027.

Seattle’s business and tech community just delivered a World Cup-sized assist for a great cause.

A premium 2026 World Cup suite experience, donated by Microsoft, raised $100,000 for Seattle Children’s Hospital after an auction organized by GeekWire. 

UiPath, the business orchestration and AI automation giant which is expanding its Bellevue, Wash. operations, submitted the winning bid last week. 

The auction kicked off at the GeekWire Awards earlier this month. Sounders FC captain and U.S. Men’s National Team midfielder Cristian Roldan and Microsoft deputy general counsel Brian DeFoe promoted the effort on stage. 

“We’re proud to join together to support the incredible, life-changing work Seattle Children’s does every day for kids and families in our own backyard,” said Michael Atalla, chief marketing officer at UiPath. “When the opportunity came up to use the magic and reach of the 2026 World Cup to support our community, it was an easy decision — and a powerful reminder of what’s possible when we can rally together around something bigger than any one of us.”

UiPath employs 195 people in the Seattle area, recently expanding its product and engineering hub at Bellevue’s Lincoln Square. The New York-based company employs 4,800 worldwide.

Microsoft — a Seattle FIFA World Cup 2026 Host City Supporter — donated the suite as part of an effort to connect the global event with local community impact, turning one of the hottest tickets into a fundraiser just weeks before the tournament begins.

“Moments like this are exactly what we hoped to inspire as a host city sponsor,” said Jane Broom, senior director of community leadership at Microsoft. “It’s a great reminder that our business community shows up not just for big events, but for the people who live here long after the final whistle. Thanks to GeekWire for bringing Microsoft, UiPath, and Seattle Children’s together so we can do more.”

As Seattle prepares for one of the biggest global events in the city’s history, the auction offered a glimpse of how the World Cup is bringing together the region’s tech, business and philanthropic communities.

The World Cup starts in North America on June 11, with the first match in Seattle taking place on June 15 when Egypt faces Belgium. 

UiPath’s winning auction bid was for the last match in Seattle, a round of 16 knockout stage fixture on July 6. The teams for that match have not yet been determined, though it could be the U.S. Men’s National Team if they advance far enough in the tournament. 

S. “Soma” Somasegar, a fixture in the Seattle tech community who led Microsoft’s Developer Division as part of his 27-year tenure at the company before supporting a generation of cloud and AI startups as an investor, board member and advisor, has passed away.

The news was confirmed Tuesday afternoon by Microsoft and Madrona, the Seattle-based venture capital firm where Somasegar had been a key figure for the past 11 years.

Microsoft CEO Satya Nadella, who first met Somasegar at Microsoft in the early 1990s, remembered him in a statement as “a remarkable leader who helped grow and shape Microsoft’s developer ecosystem, and a dear friend and colleague that I valued greatly.” 

“He brought depth, humility, and a real commitment to empowering developers everywhere and his impact on Microsoft and the broader technology community will live on!” Nadella said.

Somasegar was 59. No cause of death was given. He is survived by his wife, Akila, and two daughters.

“Soma was beloved by so many people in all aspects of his life, and he had such a generous spirit for helping others,” said Matt McIlwain, Madrona managing director. “We are deeply saddened by this loss, most importantly for his wife and his two beloved daughters.”

McIlwain added, “We are focusing on supporting his family, the Madrona team and all those who knew and loved Soma, including the broader Microsoft community.”

Tuesday evening on its website, Madrona posted an initial tribute to Somasegar, saying, in part: “We all loved Soma, as everyone who knew him did.”

On a personal level, Nadella and his wife, Anu, formed a close friendship with Soma and Akila over the decades. Nadella and Somasegar were among a group of tech leaders who co-own the Seattle Orcas, a professional cricket team based in the region. 

“For Anu and me, this loss is very personal,” Nadella said. “Soma was there for us during some of the toughest moments in our lives, always with quiet strength, kindness, and a sense of steadiness we depended on. We will miss him very much.”

From Puducherry to Seattle

Born Aug. 13, 1966, in the southern Indian coastal town of Puducherry, Sivaramakrishnan Somasegar — known throughout his life by the nickname “Soma” — grew up in a household where education came before everything else, according to a 2008 profile in Mint, the Indian business newspaper. His father worked as a technician at a hospital, his mother stayed home, and neither had attended college. 

“Food was a secondary priority in our house because education was a first priority,” Somasegar recalled in a 2024 oral history conducted for the Microsoft Alumni Network. “Whatever little I’ve done so far, it’s a direct result of that.” 

He arrived in the U.S. in 1987 to pursue a master’s in computer engineering at Louisiana State University, having mistaken the “LA” in his admission letter for Los Angeles. He realized his mistake only as the plane was landing in New Orleans, he recounted in the oral history. 

After 18 months at LSU, Somasegar enrolled in a PhD program at the State University of New York at Buffalo. He left after a single harsh winter semester to join Microsoft, arriving in Redmond on Jan. 23, 1989 — a date he remembered precisely decades later.

He joined the OS/2 team as a software design engineer in test, working on memory management and file systems. Within six months, Microsoft’s relationship with IBM on the joint OS/2 project was fraying, and Somasegar was drafted in March 1990 onto what would become one of the most consequential projects in the company’s history: Windows NT.

Somasegar spent his first decade at Microsoft on the NT team, ultimately contributing to eight releases of the Windows operating system, as recounted in a 2015 GeekWire “Geek of the Week” profile. He rose from software design engineer to test lead to test manager.

During the NT years, Somasegar designed the team’s overnight stress test program and ran the daily reliability check himself, arriving around 5:30 a.m. to walk the halls, leave yellow sticky notes on crashed machines, and report findings at the 9 a.m. bug meeting. 

He also founded Microsoft’s India Development Center in Hyderabad in 1998, which has grown into one of the company’s largest engineering operations outside the United States. Especially given his roots, he often called the India effort one of his proudest contributions to Microsoft.

By the time Windows Server 2003 shipped, Somasegar had risen to vice president. In December 2003, Microsoft’s then-server and tools chief Eric Rudder asked him to take over the Developer Division — the group responsible for Visual Studio, .NET, and the tools used by millions of software developers.

Somasegar held the role for the next 12 years, eventually as senior vice president. Under his leadership, the division extended its reach from Windows into mobile and the cloud. 

In 2014, Somasegar was an internal advocate and leader for Microsoft’s decision to open-source the .NET core server runtime and framework, a surprise move that marked a significant shift in the company’s approach toward the broader developer world.

Shifting to startups at Madrona

Somasegar announced his departure from Microsoft in October 2015, and within weeks had joined Seattle-based Madrona Venture Group as a venture partner. He was promoted to managing director in January 2017.

At Madrona, Somasegar focused on early-stage investments in cloud infrastructure, developer tools, AI, and what the firm calls intelligent applications. 

He led or played a key role in Madrona’s investments in Snowflake, UiPath, Pulumi, Statsig, Common Room, Rhythms, and RelationalAI, among others. Several became multibillion-dollar companies. Statsig was acquired by OpenAI for $1.1 billion in 2025.

Somasegar also served on the boards of UiPath and other portfolio companies. 

He remained an active writer and commentator on the industry, including a February 2024 GeekWire guest post reflecting on Satya Nadella’s decade as Microsoft CEO and their friendship that began at Microsoft in 1992. He conducted interviews and served as the primary on-stage host of Madrona’s IA Summit in downtown Seattle last fall. 

Just this week, Somasegar was named to Business Insider’s Seed 100 list of the best early-stage investors of 2026. 

Beyond venture capital, startups and technology, Somasegar was deeply involved in Seattle’s sports and cultural community. In addition to co-owning the Seattle Orcas, he was part of the ownership group of the Seattle Sounders FC.

Ed Lazowska, the longtime University of Washington computer science professor and a fixture in Seattle’s tech community, said Somasegar’s spirit fit a tradition going back to Madrona’s earliest days.

“Soma was a wonderful human being, in the tradition of the four Madrona co-founders,” Lazowska said, referencing Tom Alberg, Jerry Grinstein, Bill Ruckelshaus, and Paul Goodrich.

Manuela Papadopol, executive director of the Microsoft Alumni Network, said she was heartbroken by the loss. Somasegar was a mentor and advisor to Papadopol, and a steadfast supporter of the alumni network who “embodied the very best of Microsoft,” she said.

“He was a world-class technologist and investor, but what set him apart was his generosity with his time, wisdom, and encouragement,” Papadopol said. “His impact will live on through the countless founders, developers, leaders, and alumni whose lives he touched.”

In one of his last extended conversations with GeekWire, recorded earlier this year at Madrona’s 30th anniversary celebration, Somasegar reflected on the venture firm’s role in the region and its philosophy of being a “trusted partner” to founders from day one. 

He spoke about being part of an early Madrona-led effort during the pandemic that helped raise more than $25 million for the All In Seattle campaign to support those impacted by homelessness and other community needs.

“We have a day job. We want to be the best venture capital guys,” he said. “But we are also all about the community. We are about embracing the community.”

Nadella said he will remember Somasegar’s “warmth, his thoughtful advice, and the integrity he brought to everything he did.”

“Our thoughts are with Akila and his daughters, and with everyone who had the privilege of knowing him,” he said. “He will be deeply missed and remembered for all he did and contributed to our industry and our community.”

Friends and colleagues with memories of Soma to share can reach Todd Bishop at todd@geekwire.com.

The wave of departures from the Allen Institute for AI to Microsoft is bigger than previously known: A total of at least 10 former Ai2 staffers and researchers have joined the tech giant, including the core of the Seattle-based institute’s flagship OLMo open-source model effort.

In addition to the previously reported Microsoft hires — former Ai2 CEO Ali Farhadi, former COO Sophie Lebrecht, and research leaders Hanna Hajishirzi and Ranjay Krishna — former Ai2 researchers now at Microsoft include Luca Soldaini, Kyle Lo, Dirk Groeneveld, Pete Walsh, Matt Jordan, and Jake Poznanski.

They have joined the Superintelligence team led by Microsoft AI CEO Mustafa Suleyman, working on its core mission and AI model post-training, a Microsoft spokesperson confirmed in response to GeekWire’s inquiry. Formed in November, the team is developing what Suleyman has called “humanist superintelligence,” advanced AI systems in areas such as health care, energy, and AI companions. 

More broadly, Microsoft is working to reduce its dependence on OpenAI for its own AI models. The hiring of the Ai2 group brings Microsoft a team with deep expertise in efficient, fully open model development, an area where Ai2 has punched well above its weight. 

Ai2 confirmed that the researchers are no longer with the institute.

“While we saw a small number of departures earlier this year, Ai2’s mission remains unchanged,” a spokesperson said. “We remain focused on developing a fully open AI ecosystem and advancing AI for good across health, science, and environmental research.”

As evidence of Ai2’s continued momentum, the spokesperson cited a new computing cluster brought online last week as part of the $152 million initiative backed by the NSF and Nvidia, known as Open Multimodal AI Infrastructure to Accelerate Science, or OMAI.

When Farhadi’s departure as CEO was announced in March, Ai2 board chair Bill Hilf said the cost of competing at the frontier of AI as a nonprofit had become a fundamental challenge. 

“The cost to do extreme-scale open model research is extraordinary,” Hilf said at the time, adding that it’s “really hard to do extreme-scale model work inside of a nonprofit.” He said the board had to weigh whether philanthropic dollars were best spent trying to keep pace with tech giants spending billions on infrastructure to train the most advanced models. 

Behind the scenes, the changing nature of Ai2’s funding environment has also been playing a role in the exits, according to people with knowledge of the situation. Ai2’s primary backer is now the Fund for Science and Technology, a $3.1 billion foundation created under Allen’s instructions. The funding process has shifted from providing an overall annual budget to a proposal-based process.

A spokesperson for FFST said previously that Ai2’s “work and mission remain the same.” 

Ai2’s approach to open-source AI has set it apart in the industry. Unlike most leading AI labs, the institute releases the full weights, training data, code, and evaluation tools behind its models, allowing outside researchers to inspect, reproduce, and build on the work. 

Among the leaders remaining at Ai2 is Noah Smith, the Ai2 senior research director and University of Washington professor who leads the $152 million OMAI project.

The late Microsoft co-founder Paul Allen started Ai2 in 2014, with the mission of advancing AI research for the common good. Farhadi had led the institute since July 2023, succeeding founding CEO Oren Etzioni. The Ai2 board is searching for a new permanent CEO.

In a Q&A posted May 4 on the Ai2 site, interim CEO Peter Clark outlined the path forward, emphasizing longer-term research, open models, and applied AI in areas such as scientific discovery, embodied AI, and environmental science. 

“Ai2 was created to take that longer-horizon view,” Clark said in the Q&A. “From the beginning, Paul Allen’s vision was to advance AI in ways that push science forward while also delivering meaningful benefit to the world, and, critically, doing it in the open.”

He said that commitment has “become even more important in the current landscape.”

Mythos has been MDASH’d.

A new AI-powered system from Microsoft surpassed a headline-grabbing rival from Anthropic on a leading cybersecurity benchmark, using more than 100 specialized AI agents working together across multiple AI models to find real-world software vulnerabilities.

Microsoft’s system, codenamed MDASH, was introduced this week alongside the disclosure of 16 new vulnerabilities it found in different versions of Windows, including four “critical” remote code execution flaws fixed in this month’s Patch Tuesday release. 

The company, which has faced persistent criticism over security lapses, is betting that multiple models can discover vulnerabilities at a pace that individual models can’t match. 

MDASH, derived from the term “multi-model agentic scanning harness,” works by running specialized AI agents through a staged pipeline. Different agents scan code for potential vulnerabilities, then a separate set of agents debate whether each finding is real and exploitable, and a final stage constructs proof-of-concept attacks to confirm the bugs exist.

By comparison, Anthropic’s Mythos, which raised concerns over its ability to find and exploit software vulnerabilities when it was previewed earlier this year, is a single AI model running inside an agent framework. Anthropic restricted its release to a handful of companies through a consortium called Project Glasswing, which includes Microsoft.

OpenAI’s GPT-5.5 and others on the leaderboard are also single-model systems.

MDASH scored 88.45% on the CyberGym benchmark, a test developed by UC Berkeley researchers that measures how well AI systems can reproduce real-world vulnerabilities across 1,507 tasks drawn from 188 open-source software projects.

Mythos Preview was second at 83.1%, followed by GPT-5.5 at 81.8%. 

The benchmark gives each system a description of a known vulnerability and an unpatched codebase, and measures whether it can produce a working attack that triggers the bug. 

The scores on the CyberGym leaderboard are self-reported by the companies, including Anthropic’s Mythos result. The benchmark code is public, but no independent party has verified any of the scores. Also, benchmark results don’t necessarily reflect real-world performance.

The results also highlight growing concerns about AI’s use as an offensive hacking tool. The same capabilities that allow AI to find vulnerabilities in friendly hands can be used to discover them for exploitation by attackers. Microsoft said MDASH is being used internally by its security engineering teams and will be entering a limited private preview with customers. 

Microsoft is telling customers to expect bigger Patch Tuesdays going forward as AI accelerates the discovery of vulnerabilities.

Microsoft CTO Kevin Scott took the stand Wednesday and, for the first time, publicly addressed the internal email that Elon Musk’s lawyers have cited to support allegations that Microsoft knew OpenAI was abandoning its nonprofit mission before investing billions in the company.

That email, sent by Scott on March 7, 2018, read in part, “I wonder if the big OpenAI donors are aware of these plans? Ideologically, I can’t imagine that they funded an open effort to concentrate ML [machine learning] talent so that they could then go build a closed, for-profit thing on its back.”

Musk alleges in the suit that Sam Altman and OpenAI secured his donations to found a nonprofit AI lab and then, with Microsoft’s help, converted it into a for-profit venture that enriched its leaders.

On the stand Wednesday, Scott said he was asking whether OpenAI even had standing to pursue the commercial plans it was pitching to Microsoft, not raising bigger questions about its mission. He explained that both companies were behind Google in AI, that OpenAI had recently left Azure for Google, and that he was worried the conversations would be “a big distraction.” 

Scott said the OpenAI donor he had in mind was not Musk but rather his friend Reid Hoffman, the LinkedIn co-founder, who sits on the Microsoft board.

But later that year, Scott testified, over dinner with Altman and retired Microsoft exec Craig Mundie at Flea Street Cafe in Menlo Park, he learned a key detail: Hoffman, the donor he had wondered about, was actually investing in OpenAI’s new for-profit entity and joining the non-profit board.

Also at the dinner, Scott said he learned that OpenAI was raising a $500 million round, that Altman was leaving Y Combinator to lead the company full time, and that OpenAI had created a new “capped profit” corporate structure as part of the new funding round. Scott called that structure “surprising and interesting” — something he said he had never seen before.

The path to a deal: But Microsoft was still far from committing. Scott testified that the company had “a substantial amount of diligence we needed to do,” including technical, financial, legal, and governance. 

By June 2019, the stakes were becoming more clear. In a confidential memo at the time, filed as an exhibit in the case, Scott and Microsoft CFO Amy Hood formally asked Microsoft’s board to approve a $1 billion investment in OpenAI. Scott warned that Google had used its proprietary AI training infrastructure to pull ahead, and that Microsoft was “scrambling to replicate” the results. 

Without OpenAI, Scott wrote in an appendix to the memo, Microsoft faced “gaps in experience and talent” that would make building its own program “time-consuming and risky.” 

A key part of the strategic case was that Microsoft needed what Scott called a “frontier AI workload” on Azure — a customer pushing the platform at a scale that would reveal what infrastructure needed to be built. Google had that advantage; Microsoft did not.

The board approved the investment. Microsoft announced the deal in July 2019, the first investment in a multi-year partnership that would see the company commit a total of $13 billion to OpenAI.

Within six months of that first deal, the companies had built their first AI supercomputer together, and OpenAI used the computing horsepower to train what would become known as GPT-3.

On the stand Wednesday, Scott called the partnership a success. “I’m very proud of our infrastructure capabilities,” he said, adding that he was proud overall of what Microsoft enabled OpenAI to do.

Pushback from Musk’s team: One of Musk’s lawyers challenged elements of Scott’s account in a brief but pointed cross-examination.

For example, Scott had testified that he did not have any understanding when writing the March 2018 email of whether OpenAI was releasing its technology as open source. Musk’s lawyer showed Scott an email he had received earlier, in which Microsoft chief scientist Eric Horvitz wrote OpenAI had “been sharing their work openly, per their basic tenet.” Scott confirmed he received it. 

Musk’s lawyer also pressed the Scott on whether Microsoft had conducted legal due diligence specifically for compliance with nonprofit law. Scott said he didn’t know, adding that the legal work was handled by others on Microsoft’s team.

New financial details: Also on the stand Wednesday, Microsoft corporate development leader Michael Wetter addressed the scale of Microsoft’s commitment to OpenAI. He testified that Microsoft’s total spending related to OpenAI — including its $13 billion in investment commitments, Azure infrastructure, and hosting costs — is “upwards of $100 billion” as of this fiscal year end in June. 

Wetter testified that Microsoft had generated approximately $9.5 billion in direct revenue from the partnership through March 2025. Separately, The Information reported this week that Microsoft’s total OpenAI-related revenue (including Azure server rentals, Copilot sales, and revenue-sharing payments) exceeded $30 billion between 2023 and 2025.

Under their deal announced last fall, Microsoft received a stake of roughly 27% in OpenAI, with a commitment by OpenAI to spend $250 billion on Microsoft’s Azure cloud services. 

On cross-examination by a lawyer for Musk, Wetter acknowledged that Microsoft, having contributed 98% of the capital in OpenAI’s for-profit entity at one point in time, held effective approval rights over major corporate transactions. This is a level of influence Musk’s lawyers have argued amounted to control.

Wetter said Microsoft has never rejected an approval request. 

Under the latest renegotiation of their deal, announced as the trial began, OpenAI gained the ability to serve its products on any cloud platform, ending its exclusive commitment to Azure. Amazon Web Services quickly moved to offer OpenAI’s models on its own platform. 

Microsoft’s license to OpenAI’s technology was extended through 2032 but became non-exclusive, and the companies removed a clause that could have cut Microsoft off from future models if OpenAI declared it had achieved artificial general intelligence. 

Musk’s legal case: Lawyers for the SpaceX and Tesla founder have argued that Microsoft’s approval rights gave it effective control over OpenAI’s transformation from nonprofit to for-profit, and that the company proceeded despite its own CTO flagging the potential problem in 2018.

Microsoft has maintained that it relied on OpenAI’s contractual assurances that the partnership would not violate any third-party rights. Wetter testified that Microsoft found “no conditions related to Elon Musk” in its normal process of due diligence.

Microsoft is named as a defendant in the case on allegations of aiding and abetting what Musk asserts was a breach of charitable trust by Altman and OpenAI in the for-profit conversion. 

What’s next in the suit: Testimony in the case ended around 1 p.m. today in federal court in Oakland. Closing arguments are set for Thursday, with jury deliberations expected to begin on Monday.

The jury will determine whether OpenAI breached its charitable trust and whether Altman and others were unjustly enriched. If the jury finds for Musk, the judge will determine the amount of financial damages.

Musk is seeking up to $134 billion across all defendants, though U.S. District Judge Yvonne Gonzalez Rogers has questioned the methodology behind those financial calculations. Musk, the world’s richest person, has said he would donate the proceeds to charity.

GeekWire reported on today’s proceedings via the court’s audio livestream.

LinkedIn is laying off workers across engineering, product and marketing, Bloomberg reported Wednesday, as the tech sector continues shedding roles.

CEO Daniel Shapero disclosed the cuts in an internal memo cited by Bloomberg. The professional networking platform, which is owned by Microsoft, did not say how many of its 17,500 employees would be affected or where they are based. A report from Reuters put the cuts at 5% of the workforce.

“As part of our regular business planning, we’ve implemented organizational changes to best position ourselves for future success,” a company spokesperson told GeekWire via email.

Shapero took the helm at LinkedIn last month after serving as chief operating officer since 2021. He succeeded Ryan Roslansky, who was elevated to executive vice president overseeing both LinkedIn and Microsoft Office.

The cuts come despite strong financial performance. In January, LinkedIn reported crossing $5 billion in quarterly revenue for the first time, and last month said its annual revenue grew 12% year-over-year. Microsoft acquired the company a decade ago for $26.2 billion.

The layoffs are the latest in a string of workforce reductions at Microsoft. The tech giant cut 6,000 employees, roughly 3% of its global workforce, about a year ago, then trimmed an additional 9,000 jobs last July. It recently offered voluntary retirement to thousands of employees for the first time in its 51-year history, targeting workers whose age plus years of service total 70 or more, and has flattened management layers while overhauling its compensation structure.

Microsoft has repeatedly denied a direct link between the cuts and its growing use of artificial intelligence to automate coding tasks. But as AI efficiencies expand and the company invests billions in data centers, it continues trimming its payrolls.

Former Google Cloud CEO Diane Greene was the name that drew the “strong, strong no” from Kevin Scott when the Microsoft CTO was weighing potential candidates for a revamped OpenAI board in a November 2023 text thread with Satya Nadella, Sam Altman, and Brad Smith.

That’s one of the revelations from a newly unredacted version of the thread, which was previously released with the subjects of the brainstorm blacked out. The exhibit was introduced as part of Nadella’s testimony Monday at the Musk v. Altman trial — revealing many of the names for the first time.

William “Bing” Gordon was suggested by Scott, until Nadella noted his Amazon connections. The veteran gaming exec and Kleiner Perkins partner had been on Amazon’s board for 14 years and retained ties to the company even after stepping down.

On the stand Monday in federal court in Oakland, Nadella acknowledged that he objected to both Gordon and Greene because of their ties to companies that compete with Microsoft in AI.

Belinda Johnson, the former COO of Airbnb, got the opposite reaction, according to the unredacted text thread. Scott called her “great,” and Nadella liked the message.

Sue Desmond-Hellmann, the former CEO of the Gates Foundation, was Nadella’s own suggestion. She was later appointed to the OpenAI board.

Nadella also floated Ursula Burns, the former CEO of Xerox.

Smith, Microsoft’s president, pitched Anne Sweeney, a former president of Disney-ABC Television Group and Netflix board member, as “solid, thoughtful, calm.”

He also suggested Leslie Kilgore, the former CMO of Netflix, calling her “incredibly smart, firm, practical, while also a good listener.”

“Yep,” Scott agreed, noting that Kilgore was on the LinkedIn board (where Scott worked prior to its acquisition by Microsoft). “Very reasonable,” was his assessment.

Scott ran through a longer list: Amy Rao, co-chair of Human Rights Watch; Emilie Choi, president and COO of Coinbase; Julia Hartz, co-founder and former CEO of Eventbrite; Ciporra Herman, the former Facebook finance leader and former CFO of the San Francisco 49ers; Maynard Webb, the former CEO of LiveOps; and Jeff Weiner, the former CEO of LinkedIn.

Microsoft’s CTO even threw a wild card into the mix: himself. “I can quit for six months and do it,” he texted, “Ready to be downvoted by Satya on this one, and not really serious.”

Nadella disliked that suggestion, according to the thread.

By evening, a framework had emerged. Altman proposed a three-person board of Bret Taylor, Larry Summers, and Adam D’Angelo, with himself restored as CEO but not on the board.

Smith raised concerns about Summers. The former U.S. Treasury secretary was smart, Smith wrote, “but so mercurial. I think it’s too risky a proposition.”

Altman acknowledged as much. “id accept it given my conversations with him and where we right now,” he wrote. “it’s bullshit but i want to save this … can you guys live with it?”

Nadella’s response: “Can I call Larry tonight?” Altman gave him Summers’ cell number.

Summers served on the OpenAI board until November 2025, when he resigned following revelations about his correspondence with Jeffrey Epstein.

The OpenAI Foundation board today is Sierra co-founder Bret Taylor as chair; Quora CEO Adam D’Angelo; Desmond-Hellmann; Carnegie Mellon professor Zico Kolter; retired U.S. Army Gen. Paul Nakasone; Global Infrastructure Partners chairman Adebayo Ogunlesi; former Sony Corp. of America President Nicole Seligman; and Altman, who was reinstated to the board in 2024.

Following last year’s restructuring, the foundation holds a minority stake in OpenAI’s for-profit public benefit corporation, making it one of the wealthiest nonprofits in the world. Musk’s lawsuit accuses OpenAI and Microsoft of betraying the organization’s original nonprofit mission, seeking up to $134 billion in damages.

At the time of the text thread, on Nov. 21, 2023, Altman had been ousted as OpenAI CEO and was soliciting input from Microsoft’s executives on potential board members as he negotiated his return, given the Redmond company’s massive financial stake in OpenAI and the partnership between the companies.

Musk’s lawyers argue that the thread shows the degree of influence Microsoft exercised over OpenAI’s nonprofit board despite having no formal governance role. Microsoft is a named defendant in the lawsuit, which accuses the company of aiding and abetting OpenAI’s alleged breach of charitable trust.

Nadella has said publicly that Microsoft didn’t want a board seat. He testified Monday that the discussions were initiated by Altman and that the board could have ignored his suggestions.

In one of the most widely cited lines from his testimony, Nadella said the old board’s handling of Altman’s firing was “amateur city.” He said he never got a specific explanation beyond the public statement that Altman had not been “consistently candid.”

Asked by Musk’s attorney why none of his board suggestions came from a safety background, Nadella said he didn’t think safety was the problem. The challenge, he said, was ensuring that the OpenAI board “had the right sort of maturity to make sure that OpenAI doesn’t blow up.”

Satya Nadella drew a historical parallel to Microsoft’s early PC partnership with IBM as the tech giant prepared to invest $10 billion more in OpenAI in April 2022 — writing in an internal email that he didn’t want Microsoft to become IBM while OpenAI became the next Microsoft. 

That email, presented as evidence by Elon Musk’s lead trial attorney Steven Molo, was one of the new details to emerge from the Microsoft CEO’s turn on the stand Monday morning in Musk’s lawsuit against Sam Altman, OpenAI and Microsoft in federal court in Oakland. 

Nadella described the decision to invest in OpenAI as a “one-way door,” saying Microsoft couldn’t build two supercomputers — one for itself and one for OpenAI — and had to accept the opportunity cost of diverting scarce computing resources away from its own AI teams. 

“We were outsourcing essentially a lot of the core IP development and taking a massive dependency on OpenAI,” Nadella testified, explaining that he wanted to ensure Microsoft had access to the intellectual property generated by the partnership, and continued to build its own knowledge and capabilities at the same time.

Board considerations unredacted: The testimony also provided new information from messages among Microsoft execs and Altman in the days following his brief ouster as OpenAI CEO in 2023. The names of potential candidates from that thread were previously redacted in public court records.

From Nadella’s testimony Monday, it emerged that two potential OpenAI board candidates for whom he voiced his disapproval were Diane Greene, the former Google Cloud CEO, and Bing Gordon, the veteran gaming exec and Kleiner Perkins partner previously on Amazon’s board. Nadella said he objected to both as potential candidates because of their ties to companies that compete directly with Microsoft in AI.

[Follow-up: Filing reveals who Microsoft favored (and opposed) for OpenAI’s board]

He said the discussions were initiated by Altman and other OpenAI insiders seeking his input, and that the board could have ignored his suggestions. One candidate he suggested, former Gates Foundation CEO Sue Desmond-Hellman, was later appointed to the board.

Musk argues that Microsoft’s efforts to protect its interests in the OpenAI partnership came at the expense of the OpenAI nonprofit’s original mission to develop AI for the benefit of humanity. His lawsuit alleges that Microsoft aided and abetted a breach of the charitable trust that governed OpenAI’s founding, misusing his original investment, estimated at $38 million to $44 million.

Enabling a massive nonprofit: Nadella offered a different view on the stand, describing a collaboration built on mutual benefit in which Microsoft took on enormous risk to support a fledgling AI lab that no one else was willing to fund. He said the partnership had created “one of the largest nonprofits in the world,” enabling products like ChatGPT and Copilot that put AI tools in the hands of millions of people.

Under cross-examination, however, Nadella acknowledged that he was not aware of any full-time employees at the OpenAI nonprofit before March 2026, or of any grants, research, or open-sourced technology it had produced. 

One of Microsoft’s attorneys in the case, Jay Jurata of Dechert, also sought to undermine Musk’s standing in the case. He walked Nadella through three major milestones in the Microsoft-OpenAI partnership — the 2019 announcement, a 2020 exclusive license to GPT-3, and the 2023 $10 billion investment — and asked each time whether Musk had reached out to object. 

Each time, Nadella said no. He and Musk have each other’s phone numbers, he added. 

Microsoft estimates the OpenAI return: Musk’s attorney, on cross-examination, sought to show the benefits Microsoft has received from the partnership.  He walked Nadella through a January 2023 memo from Microsoft President Brad Smith to the company’s board, projecting a $92 billion return on Microsoft’s cumulative $13 billion investment in OpenAI. 

According to the testimony, a footnote in the memo showed a 20% annual increase kicking in starting in 2025, which could roughly double the return within four years.

Under the restructured deal announced last year, the caps on Microsoft’s returns were removed entirely. Microsoft and OpenAI also recently amended the partnership to make Microsoft’s IP license non-exclusive and open all OpenAI products to any cloud provider.

[Update: The Information reported Monday that revenue-sharing payments from OpenAI to Microsoft under the new deal are capped at $38 billion.]

Asked about the memo on the witness stand, Nadella confirmed the figures but noted that the investment carried real risk, saying the return could just as easily have been zero. 

The trial, before U.S. District Judge Yvonne Gonzalez Rogers, is expected to continue through May 21, with OpenAI CEO Sam Altman also expected to take the stand this week. 

GeekWire reported on today’s proceedings via the court’s audio livestream. Correction: The name of Microsoft’s outside counsel for Nadella’s testimony has been corrected since publication.

Microsoft employees eligible for the company’s first-ever voluntary retirement program are learning the details of the package Thursday morning, including the size of cash payments, length of healthcare coverage, and vesting of stock awards if they take the company’s offer.

As described in an internal summary viewed by GeekWire, lump-sum cash payments will range from eight weeks to 39 weeks (about nine months) of base pay, depending on level and tenure. 

Participants would also receive up to five years of continued access to Microsoft’s medical, dental, and vision coverage for themselves and their dependents. Microsoft would fully subsidize the cost in the first year, with participants paying standard COBRA rates after that. The coverage could end sooner for those who reach Medicare eligibility at age 65.

Unvested stock awards would continue to vest for six months after an employee’s departure, extending to 12 months for those with 24 or more years at Microsoft. 

Some longer-tenured employees who meet additional age and service thresholds could qualify for continued vesting of all eligible unvested awards on their original schedule.

Eligible employees have 30 days to decide whether to accept the offer. There are no apparent restrictions on what employees could do after accepting the offer, such as finding other employment.

Announced by the company on April 23, the program is rare in the tech industry, where companies have relied on layoffs, stricter performance reviews, and return-to-office policies to manage headcount. Microsoft itself laid off more than 15,000 employees last year and began requiring Seattle-area workers to return to the office three days a week in February. 

An estimated 7% of Microsoft’s 125,000-person U.S. workforce, or roughly 8,750 employees, is eligible for the program, which is open to those at Level 67 and below whose age plus years of service totals 70 or more. Microsoft said it is a one-time offer.

On Microsoft’s earnings call last week, CFO Amy Hood disclosed that the company expects to take a $900 million charge related to the voluntary retirement program in the current quarter. She also said headcount declined year over year and will continue to decline in fiscal 2027.

[Editor’s Note: Agents of Transformation is an independent GeekWire series, underwritten by Accenture, exploring the adoption and impact of AI and agents. See coverage of our related event.]

A new Microsoft study of 20,000 artificial intelligence users in workplaces around the world concludes that the biggest barrier to getting real value from AI isn’t the technology or the workers themselves — it’s the ingrained culture of the organizations where they work.

That “Transformation Paradox” is one of the central findings from Microsoft’s annual Work Trend Index, released Tuesday morning, which paints a picture of employees eager to reshape their jobs and organizations that aren’t really in a position to make it happen.

Sixty-five percent of the AI users surveyed said they fear falling behind if they don’t adopt AI quickly. But only 13% said they’re rewarded for using and experimenting with AI in their jobs.

“Employees are ready to reinvent how they work, but the system around them—metrics, incentives, and norms—continues to reinforce the old way,” Microsoft says in the report. 

The takeaway: For companies to truly capitalize on the AI revolution, leaders need to fundamentally overhaul how work is structured, managed, and rewarded, rather than simply handing workers new tools and expecting them to figure it out.

Matt Firestone, general manager of Microsoft’s Frontier Firm initiative, said the message to leaders has changed. Two years ago, executives were under pressure from their boards to unlock value from AI. Now, he said, the message is that their people are already there.

It’s the job of leaders to “re-architect work,” Firestone said in an interview ahead of the report’s release. “Your job is to convert the individual agency and capacity and abilities of your people to unlock that and apply it to increase business value for the enterprise.”

Leaders who encourage employees to experiment with AI and share their experiences create “these incredible systems of learning that drive us forward into the agentic era,” he said.

Of course, this also serves Microsoft’s interests: the company is betting heavily on agents for the next phase of its outsized AI product ambitions, and a report that says organizations need to change how they work is also a pitch for more tools, training, and licenses.

Alongside the report, Microsoft is announcing new capabilities for Copilot Cowork, including a mobile app and a plugin ecosystem for connecting to third-party business systems.

New data on how workers use AI 

The report is the latest installment of a survey that has tracked the transformation of work from the early days of the COVID-19 pandemic through the rise of AI in the workplace and the “infinite workday.” Last year’s edition introduced the concept of the “Frontier Firm” and foresaw a world in which workers served as “agent bosses” managing AI teammates.

This year’s Work Trend Index was narrower, covering 20,000 workers across 10 countries, down from 31,000 across 31 countries in recent years. The survey was conducted by Edelman Data x Intelligence. In a new twist, it also excluded anyone who doesn’t already use AI at work.

As it has in the past, Microsoft also analyzed trillions of anonymized productivity signals from Microsoft 365. The company partnered with Harvard Business School and in-house organizational psychologists to interpret the findings. 

New this year was an analysis of more than 100,000 Copilot chats, classified by the type of work involved. That analysis found that 49% of all Copilot interactions involved cognitive work — analyzing information, solving problems, and thinking creatively — rather than simpler tasks like summarizing documents or finding information. 

Microsoft is using that data point to assert that AI is not just making workers faster but expanding the types of work people can accomplish.

The rise of ‘Frontier Professionals’

Fifty-eight percent of AI users surveyed said they are producing work they couldn’t have a year ago, rising to 80% among a group the report calls “Frontier Professionals.”

These are the 16% of AI users who routinely use agents for multi-step workflows, redesign how their work gets done, and share what they learn with their teams. Frontier Professionals are also more deliberate about when not to use AI: 43% said they intentionally do some work without it to keep their skills sharp.

The largest group of AI users in the study (42%) sat in what Microsoft called the “emergent” middle, where both individual skills and organizational support are still taking shape.

On the organizational side, the report found that culture, manager support, and talent practices account for more than twice the AI impact of individual factors like mindset and behavior. 

When managers actively modeled AI use, employees reported a 17-point increase in the value they got from AI and a 30-point boost in trust in agents, according to a separate Microsoft study of 1,800 workers. But only one in four AI users said their leaders are clearly aligned on AI.

Emerging AI adoption patterns 

The report also includes Microsoft’s first Work Trend Index data on AI agents, showing a 15x year-over-year increase in active agents on Microsoft 365, rising to 18x in large enterprises. Microsoft did not disclose the baseline, making it difficult to assess the actual scale of adoption.

The new report says adoption patterns vary by industry. As would be expected, software and technology companies showed the broadest use of agents across job functions.

But Microsoft said it was surprised by the depth of adoption in manufacturing, where fewer companies were using agents but those that did were deploying them heavily in specific tasks. Banking and capital markets, retail, and education also showed significant agent adoption. 

In a blog post accompanying the report, Jared Spataro, Microsoft’s chief marketing officer for AI at Work, described four emerging patterns for how humans and AI agents work together:

• Author: The worker produces the work, calling on AI for help as needed.
• Reviewer: The worker sets the intent and AI creates a first draft to edit and approve.
• Director: The worker hands off entire tasks for AI to execute and signs off on the outcome.
• Orchestrator: The worker designs a system where multiple agents run in parallel, flagging exceptions back to the human.

Firestone compared the current moment in AI to the early days of mobile apps, when people were building apps before app stores and permission models existed.

“People are building agents. They’re hobbyists,” he said. “Their personal knowledge is extending the professional workplace. This is a new wave of technology, but all of the fundamental instincts of how to transform the workplace haven’t changed.”

Bob. Clippy. Cortana. Copilot. Microsoft has been trying to unlock the personal-assistant puzzle for decades. Now a fledgling team inside the company that’s been experimenting with OpenClaw — an open-source framework that acts both a virtual assistant and platform for building and managing proactive agents — is taking a stab at the problem.

That team, headed by Corporate Vice President Omar Shahine, already has a working agent prototype and, as of May 1, more than 3,000 daily users inside Microsoft testing “ClawPilot,” the team’s OpenClaw-based desktop environment that is part of “Project Lobster,” up from 100 the previous week.

Not bad for a technology that CEO Satya Nadella dismissed as a security risk akin to “a virus” just a few months ago. A number of other companies, including OpenAI and NVIDIA, are also rushing to integrate the technology with their own.

The vision of Shahine’s team is to create “an always-on agent team (a Chief of Staff agent, an Executive Assistant agent, and a roster of specialist agents) that works 24/7 on your behalf within the Microsoft 365 ecosystem,” as he described it in a blog post.

It’s a “persistent runtime that monitors your signals continuously, prepares your day before you wake up, triages your inbox while you’re in meetings, and follows up on action items without being asked,” he explained.

OpenClaw, developed by Peter Steinberger (who, as of Feb. 2026, works for OpenAI) has only been publicly available since Nov. 2025, originally under the name Clawdbot.

Shahine had been dabbling with OpenClaw since earlier this year to automate tasks at home, such as drafting an email or investigating concert-ticket prices. He demonstrated how Lobster works during a presentation to Microsoft’s AI Accelerator group on Feb. 26. And by March 31, he had a new role at Microsoft: To bring OpenClaw and personal agents to Microsoft 365.

Microsoft recently has made forays into the autonomous-agent space with Copilot Tasks, an agent in preview for consumers that is designed to help with chores like triaging email and booking travel. On the business side, Microsoft is integrating Anthropic’s Cowork technology with Microsoft 365 Copilot in the form of Copilot Cowork, which takes action inside the various Microsoft Office apps.

But neither of these approaches provides a virtual assistant working on users’ behalf 24/7 with access to people’s full, real lives, Shahine maintains. They can’t do things like order from DoorDash if a user is in back-to-back meetings or reschedule a call if it interferes with a family dinner. That gap is why he decided to target knowledge workers, he says.

Shahine’s team, known as Ocean 11, includes a handful of people, each running his/her own Lobster agent. The team is building out the runtime and supporting infrastructure needed to make Lobster work in an enterprise environment.

As Lobster is currently envisioned, it will work across all kinds of apps and Microsoft 365 and other data sources. It won’t need constant prompting, but instead, will suggest courses of action it can take, pending user approval.

And this is why Nadella and other security-minded professionals have qualms about OpenClaw: It works autonomously, can ingest untested inputs, maintains persistent credentials, and could turn things like prompt-injection attacks into action-injected ones.

Microsoft’s own Defender security team’s current guidance states: “OpenClaw should be treated as untrusted code execution with persistent credentials. It is not appropriate to run on a standard personal or enterprise workstation.”

In an interview, Shahine acknowledged that enterprise-hardening Microsoft’s OpenClaw-based offerings needs to be job No. 1. His team is designing prototype agents to have their own Microsoft 365 identities, meaning their own Entra IDs for governance, their own Exchange mailbox, their own Teams presence, and integration with the Microsoft Graph.

“My goal is to contribute to make OpenClaw better but also consume it and run it so that it’s also a reference design, reference pattern that people can look to and say, ‘Well, you know, it’s great. Microsoft figured out how to make this thing enterprise great,” he said.

Shahine wasn’t ready to talk timetables or deliverables, though there is an OpenClaw-developed Teams plug-in already available. Shahine’s team already has developed the ClawPilot Mac and Windows desktop environment (no relation to clawpilot.ai) that it’s using internally to work with “claw-like agentic workflows.” Shahine said ClawPilot is acting as his personal assistant and goes by “Sebastien” (a nod to “The Little Mermaid”).

Microsoft Vice President Scott Hanselman has built a Windows node for OpenClaw which could get some airtime at Microsoft’s upcoming Build developer conference in San Francisco in June. Shahine said “there will be some concrete information about how we’re working to make Windows a fantastic environment for OpenClaw and other agentic systems to operate.”

Microsoft and Amazon joined other leading artificial intelligence companies in signing deals to deploy their technology in classified Pentagon networks, the Defense Department announced Friday, accelerating a push to build what the military is calling an “AI-first fighting force.”

The agreements — which also include OpenAI, Google, Nvidia, SpaceX and the startup Reflection — will give those firms’ AI systems access to the military’s most classified network environments, known as Impact Level 6 and Impact Level 7. The Pentagon said the technology will be used to analyze data and improve battlefield decision-making.

“Together, the War Department and these strategic partners share the conviction that American leadership in AI is indispensable to national security,” the Pentagon said in a statement, using the Trump administration’s preferred name for the Defense Department.

The Pentagon says the effort is already well underway. More than 1.3 million Defense Department personnel have used GenAI.mil, the military’s official AI platform, generating tens of millions of prompts and deploying hundreds of thousands of agents in just five months, according to the department. Officials say the technology has cut some tasks from months to days.

The deals come as the Pentagon is locked in a legal battle with Anthropic, one of the nation’s leading AI labs, which had sought guarantees its technology would not be used for mass domestic surveillance or fully autonomous weapons. The Defense Department moved to blacklist Anthropic earlier this year, calling the company a national security risk — a designation Anthropic is contesting in court.

On Thursday, Defense Secretary Pete Hegseth called Anthropic CEO Dario Amodei an “ideological lunatic” and slammed the company during a Senate Armed Services Committee hearing.

Bloomberg reported that the Pentagon negotiated its deal with Amazon Web Services late into Thursday night, according to two officials briefed on the talks.

“We look forward to continuing to support the Department of War’s modernization efforts, building AI solutions that help them accomplish their critical missions,” AWS spokesman Tim Barrett said in a statement.

Hundreds of Google employees sent a letter to company leadership this week urging them to refuse to let the Pentagon use its AI on classified data.

“We want to see AI benefit humanity; not to see it being used in inhumane or extremely harmful ways,” they wrote, according to The Washington Post.

Microsoft’s Azure cloud business accelerated in the March quarter, growing 40% and topping the company’s own forecast, giving the tech giant a new answer to questions about its ability to translate record capital spending on AI infrastructure into stronger financial results.

The company’s revenue rose 18% to $82.9 billion, beating the $81.4 billion analyst consensus, and earnings per share jumped 23% to $4.27, above the $4.06 expected by Wall Street. 

AI run rate: In its earnings news release, Microsoft also disclosed that its AI business has reached an annual revenue run rate of $37 billion, up 123% from a year ago. It’s the first time the company has updated the figure since it reported a $13 billion run rate in January 2025.

Capex trends: Capital spending came down to $31.9 billion from $37.5 billion the previous quarter. Microsoft had said the decline would come and that it reflected the timing of data center construction and hardware deliveries, not a slowdown in demand for cloud and AI services.

Copilot: For the quarter, Microsoft 365 Copilot now exceeds 20 million paid seats, up from 15 million in January. That means about 4.4% of the company’s commercial base is on its paid enterprise AI plan.

Cloud overall: Microsoft Cloud revenue, which includes Azure, commercial Microsoft 365, LinkedIn, and Dynamics 365, rose 29% to $54.5 billion. The company’s remaining performance obligations, a measure of contracted future revenue, was $627 billion, with a significant part of that backlog tied to OpenAI.

Elsewhere in Microsoft’s business:

• Revenue in the More Personal Computing segment fell 1% to $13.2 billion, with Xbox content and services revenue down 5% and Windows OEM and devices revenue down 2%. Search advertising revenue grew 12%.
• The Productivity and Business Processes segment, which includes Microsoft 365, LinkedIn, and Dynamics 365, grew 17% to $35 billion. LinkedIn revenue rose 12%, and Dynamics 365 revenue increased 22%.
• The Intelligent Cloud segment, home to Azure, grew 30% to $34.7 billion, making it nearly equal in size to the productivity segment for the first time.

The results come three months after Microsoft’s stock dropped 10%, wiping out $357 billion in market value, despite the company beating expectations on revenue and earnings.

Investors focused on the record capital spending, a Copilot product that had reached just 3.3% of Microsoft 365’s commercial base at that time, and a revenue backlog heavily dependent on OpenAI.

The OpenAI relationship has shifted significantly since then.

This week, the two companies restructured their partnership, with OpenAI ending its exclusive commitment to Microsoft’s Azure cloud and gaining the ability to run its products on other platforms, notably Amazon Web Services. Microsoft, in turn, locked in its revenue-sharing arrangement and removed a clause that could have ended it if OpenAI had declared artificial general intelligence.

Update, April 30: On the earnings conference call Wednesday evening, CFO Amy Hood said Microsoft expects capital expenditures to exceed $40 billion in the current quarter, which would be a new record, and roughly $190 billion for calendar year 2026. She said approximately $25 billion of that reflects higher component pricing, driven in part by a global memory crunch.

Hood said the company expects to remain supply-constrained through at least the end of the year but anticipates “modest acceleration” in Azure growth in the second half of the calendar year.

Microsoft shares fell about 5% in trading Thursday morning.

OAKLAND — Did Microsoft knowingly help OpenAI abandon its nonprofit mission?

That question sits at the center of a trial starting here this week, pitting the world’s richest man against the AI nonprofit he helped found and the tech giant that bankrolled its transformation.

It’s being called the “AI Trial of the Century,” with Elon Musk and Sam Altman in starring roles, and a supporting cast that includes Microsoft CEO Satya Nadella, CTO Kevin Scott and CFO Amy Hood. Current and former OpenAI execs and board members are also on the witness list.

Monday morning in Oakland, Altman and OpenAI President Greg Brockman were on hand for jury selection, with the OpenAI CEO sitting in the front row behind the lawyers’ tables in a dark suit and light blue tie, quietly scrolling on his phone as he waited for the process to begin.

Musk was not present for jury selection. He is expected to take the stand later in the trial. 

A protest was scheduled for midday outside the courthouse, organized by the Tesla Takedown activist group under the banner “Whoever Wins, We Lose” — arguing that a billionaire power struggle over AI’s future has little to do with ordinary people.

Jury selection: Inside, U.S. District Judge Yvonne Gonzalez Rogers described the case to the jury pool: Musk alleges breach of charitable trust and unjust enrichment against Altman, Brockman, and OpenAI, and aiding and abetting breach of charitable trust against Microsoft — centered on OpenAI’s operation as a nonprofit and its creation of a for-profit affiliate.

Prospective jurors were asked about topics including their views on AI and the parties involved. 

One man said he was an avid news reader who continues to subscribe to a newspaper — drawing applause from journalists listening in the overflow room. He was more pointed about the plaintiff: “I do have some strong feelings with regard to Elon and just how he does things. Elon doesn’t care about people, much like our president. He cares about money.”

A nurse said AI creates more work in her job, requiring frequent checking and correction. 

One prospective juror, when asked by the judge if she has worked in teams, asked if the judge was referring to the conferencing app. “Microsoft is happy that you asked that question,” the judge said.

When another prospective juror expressed concern about being able to follow the technical nuances of the case, the judge replied, “This is just a case about promises and breaches of promises.”

What’s at stake for Microsoft: Amid the feud between two of tech’s most polarizing personalities, Microsoft might seem like a subplot, but its actions are at the heart of the case.

The company has invested more than $13 billion in OpenAI since 2019, building its products around the partnership and betting its competitive future on the deal, before hedging its bets more recently with rival AI firms and its own in-house models.

A victory for Elon Musk would mean a federal judge ordering Microsoft to hand over a slice of what its OpenAI partnership has been worth — not to Musk, but to the OpenAI nonprofit. 

Musk’s damages expert puts the combined demand as high as $134 billion across both defendants, with Microsoft’s share between $13.3 billion and $25 billion. However, the judge has already called these figures into question, saying Musk’s expert was “pulling these numbers out of the air.” Microsoft called the methodology “unverifiable” and “unprecedented.”

A loss could also hand regulators in the United States and Europe new ammunition just as the company tries to defend its OpenAI relationship from antitrust scrutiny. In that way, it could force every major tech company to rethink how it invests in mission-driven AI labs.

The story took a new twist Monday morning when Microsoft and OpenAI announced a major amendment to their partnership — loosening the terms of their alliance and, perhaps not coincidentally, demonstrating that their fortunes aren’t as aligned as they once were.

Microsoft’s defense: In short, the company says it was kept in the dark, that it invested as a commercial partner, never informed by OpenAI of any charitable restrictions attached to Musk’s contributions or any duties the company owed to the Tesla and SpaceX founder. 

Former OpenAI CTO Mira Murati appeared to back that up in her deposition, testifying that she never told anyone at Microsoft about those restrictions. In a filing over the weekend, Microsoft’s lawyers flagged a discrepancy: Murati’s answer to that question was missing from the official deposition transcript. It was audible on the video recording, but absent from the written record. 

Microsoft has also pointed to its work with Musk’s own AI company, xAI, as evidence of its neutrality — arguing in pretrial motions that hosting xAI’s Grok model on Azure proves it is simply a platform for competing AI models, not a partisan actor in OpenAI’s transformation. 

Microsoft’s cleanest path to victory, however, may be procedural. The company contends Musk’s claims are barred by the statute of limitations, and its primary evidence is his own words. 

In a September 2020 tweet, Musk publicly declared that “OpenAI is essentially captured by Microsoft.” If Microsoft can convince the jury that Musk knew about its involvement more than three years before he filed suit, the multibillion-dollar exposure disappears entirely.

Smoking gun? Musk’s lawyers will point to an internal Microsoft email from March 2018 in which Microsoft’s own CTO raised the very question that will come before the jury. 

Writing to Nadella ahead of a call with Altman, Scott made an observation about OpenAI’s commercial transformation: “I wonder if the big OpenAI donors are aware of these plans? Ideologically, I can’t imagine that they funded an open effort to concentrate ML [machine learning] talent so that they could then go build a closed, for profit thing on its back.”

Microsoft went on to invest billions anyway.

It’s one of many behind-the-scenes emails revealed so far in the case, including internal Microsoft exchanges showing Nadella and other executives weighing in on the composition of OpenAI’s board during the crisis that briefly ousted Altman as CEO in November 2023.

When Musk’s lawyers confronted Nadella with Scott’s email in his deposition and asked whether he shared those concerns, the Microsoft CEO deflected: “I think that the nonprofit board of OpenAI gets to make the decision on what’s the best way for them to realize their mission.” 

Nadella also said he did not recall ever raising Scott’s concerns directly with Altman. 

Microsoft says Scott’s email shows due diligence, not guilt: Scott asked the right questions, OpenAI’s board provided contractual assurances that its agreements “would not impinge any third party’s rights,” and Microsoft was legally entitled to rely on those representations. 

Backstory: Musk co-founded OpenAI in 2015 as a nonprofit dedicated to the safe development of AI, contributing tens of millions of dollars before leaving the board in 2018. He filed suit in late 2024, claiming Altman and others had transformed OpenAI into a for-profit venture, betraying the mission he helped fund and enriching themselves and their investors.

What’s next: Addressing prospective jurors this morning, Judge Gonzalez Rogers said she expects the trial to wrap by May 21 — including roughly three weeks of evidence followed by deliberations, with nine jurors deciding the case. If the jury finds for Musk, the judge will then determine in a separate proceeding how much Microsoft and OpenAI must pay out. 

RELATED: The Microsoft-OpenAI Files: Documents reveal the realities of AI’s defining alliance

Follow-up: Microsoft tops Wall Street expectations, reports accelerating Azure growth and $37B AI run rate

The last time Microsoft reported earnings, it seemed to do everything right, at least by the traditional metrics. Revenue was up 17%, profits soared 24%, and the company’s closely watched Azure cloud business beat internal forecasts. 

And then it got absolutely punished.

Microsoft’s stock dropped 10% the next day, wiping out $357 billion in market value. Investors looked past the traditional numbers, focusing on the company’s record $37.5 billion in quarterly capital spending, an AI revenue backlog heavily dependent on OpenAI, and a Copilot product that had reached just 3.3% of Microsoft 365’s commercial base.

The stock still hasn’t recovered, finishing last week down 22% from its 52-week high.

On Wednesday, Microsoft gets another chance, reporting its fiscal Q3 results after the market closes. Here’s a preview of the key numbers and storylines to watch.

Core earnings estimates: Analysts expect Microsoft to report revenue of about $81.4 billion, up 16% from a year ago, and earnings of $4.06 per share, up 17%, according to Yahoo Finance. Microsoft has beaten Wall Street’s estimates four quarters in a row.

Cloud expectations: Microsoft has said it expects Azure to grow 37% to 38% in constant currency (adjusted for fluctuations in exchange rates) in Q3. That would be a slight slowdown from the 38% it posted in Q2. Last time, Azure beat Microsoft’s own forecast but fell short of what analysts were privately expecting, a major factor in the historic stock plunge.

But the Azure number doesn’t tell the full story. CFO Amy Hood said on the last earnings call that if Microsoft had allocated all the GPUs it brought online in Q1 and Q2 solely to Azure (i.e., the company’s cloud customers), the growth rate would have been over 40%. 

Instead, the company split that capacity across Azure and its own products and operations, including Copilot, GitHub Copilot, and internal R&D. That means Azure growth is as much a reflection of how Microsoft chooses to allocate its resources as it is a measure of demand. 

A leaner Microsoft: Even in just the past few months, Microsoft has moved to cut costs and streamline its operations even as it continues to spend aggressively on AI infrastructure — attempting to demonstrate to Wall Street that it’s staying disciplined on operating expenses. 

• The company offered voluntary retirement to thousands of employees for the first time in its 51-year history, targeting workers whose age plus years of service total 70 or more. Hood is expected to discuss the financial details of the program on the earnings call.
• It flattened its management layers and overhauled its compensation structure, reducing the number of pay points from nine to five and decoupling stock awards from bonuses.
• Cloud and sales teams were put under spending and hiring freezes.
• Several senior execs announced their retirement, including Experiences and Devices chief Rajesh Jha, Developer Division leader Julia Liuson, and Xbox chief Phil Spencer. 

Capital spending: Microsoft is on pace to spend more than $100 billion on infrastructure in fiscal 2026, up from $88.7 billion the year before, mirroring spending surges across Big Tech. About two-thirds goes to GPUs and other hardware for AI and cloud workloads. 

Hood said capex spending would come down from the Q2 figure of $37.5 billion in the last quarter, but it will still be far above the company’s historical levels. Investors will be watching for any signal about whether the pace of spending is set to continue, level off, or accelerate. 

Copilot and AI monetization: Microsoft disclosed in January that its Copilot product had reached 15 million paid seats, roughly 3.3% of the Microsoft 365 commercial base of about 450 million, which has since been cited repeatedly as an example of the company falling short.

At $30 per user per month, Copilot represents a large revenue opportunity if adoption accelerates, and any new disclosures about overall usage will make big headlines. If the company doesn’t disclose this number in the new report, it could be telling, as well.

Microsoft’s contracted future revenue more than doubled to $625 billion last quarter, but about 45% of that was tied to OpenAI, thanks to the company’s renegotiated partnership with the ChatGPT maker, raising questions about risk of so much revenue connected to one company.

William Blair analyst Jason Ader noted after last quarter that Microsoft’s contracted future revenue still grew 28% after stripping out OpenAI, and that new contract signings surged 228%.

Microsoft CEO Satya Nadella also introduced a new metric last quarter: “tokens per watt per dollar,” a measure of how much AI output the company gets for each unit of energy and capital it invests. He didn’t give an overarching number, but as an example, Nadella said Microsoft was able to process 50% more OpenAI workload on the same amount of infrastructure as before. 

The bigger picture: Not everyone is pessimistic. Wedbush analyst Dan Ives, in two notes to clients last week, argued that the market is underestimating cloud growth and that fears about OpenAI and Anthropic displacing the big cloud providers are overblown. 

Ives pointed to more than $650 billion in combined AI infrastructure spending from Microsoft, Google, Amazon, and Meta in 2026, and estimated $3 trillion in enterprise and government AI spending over the next three years. He called the recent sell-off a buying opportunity. 

ServiceNow, a major enterprise software company, saw its stock drop 17% on its own quarterly results last week, a sign that business technology spending may be softer than expected. 

But Intel surged more than 20% after strong earnings, driven by a 22% jump in data center and AI revenue, a sign that demand for the computing infrastructure behind AI is broad-based. 

Earnings avalanche: Amazon, Google, and Meta all report the same afternoon as Microsoft, which means investors will be comparing Azure, AWS, and Google Cloud growth in real time. 

Check back Wednesday afternoon for coverage. 

The dreamy part of Mikaela Kiner‘s life is easy to picture. She has spent her recent winters working from a small Costa Rica beach town, taking surfing lessons before dawn, sunset walks in the sand, and Zoom calls with real palm trees swaying in the background.

But “The Reverb Way,” her new book about building and running the Seattle-based HR consulting firm of the same name, is not the postcard version of the story. 

Kiner describes what happened when new business dropped to half its usual volume, as tech layoffs, a rocky economy, and the rapid rise of AI hit Reverb’s client base. She battled insomnia so severe she couldn’t get through a workday without napping. Her daughter, watching her scramble through a client crisis, told her she’d never seen her this stressed. 

The book is a candid account of the ups and downs, detailing what Kiner has learned in a decade of reorienting her work to support the life and the company she wanted to create.

“I didn’t want to give the impression that owning a business is easy,” Kiner said in a recent conversation about the book on the porch of a Seattle coffee shop. “You can still be tired, you can still be overworked, you can still be drained, and you can still struggle.”

At the same time, she wanted to convey the fun and joy that comes from the freedom of doing your own thing. Kiner spent 15 years in HR leadership at companies including Microsoft, Amazon, and Starbucks, often working 60 to 80 hours a week, before starting Reverb in 2015.

“I made a choice to try and do something different,” she said. “And I’m so happy I did. Really, really happy. The key words there being made a choice.”

“The Reverb Way” is her second book, following “Female Firebrands” in 2020.

The new book is part memoir and part leadership guide. It draws on Kiner’s corporate career and her decade running Reverb to offer insights on everything from hiring and delegation to performance management and company values, and the daily mechanics of productivity and protecting your time.

Practical takeaways

Here are some of the insights from the book that resonated with me: 

Park your ideas. Instead of chasing every good idea the moment it came up, Kiner started logging them in a “Future Goals spreadsheet” and reviewing the list during quarterly business reviews. Some items got done as part of other initiatives. Others became irrelevant. But the team stopped getting pulled in a dozen directions at once.

Use your freedom. Reverb takes Fridays off from Memorial Day through Labor Day, with one person on call to check email a few times in case clients need help.

When Kiner offered to go further and adopt a formal four-day work week, the team turned her down. They already had the flexibility they needed. One employee had been going to a rock climbing gym at 3 p.m. every day, and Kiner never knew, because the work was getting done.

Don’t apologize for your schedule. Kiner writes about watching male executives cancel meetings for their kids’ soccer games without explanation or apology, and realizing she’d been justifying every time she was unavailable. Her rule now: no meetings before 9 or after 5, and no explanation necessary.

Build your own community. After being rejected from a business accelerator — possibly, she suspects, because she’d listed family time as a personal value — Kiner created her own informal group of women CEOs called WISE. They meet quarterly, share business insights, and support each other. Some are direct competitors. Friendship comes first.

Celebrate more than you think you need to. Kiner describes herself as a recovering perfectionist who used to hesitate to praise someone doing one thing well if they were struggling in another area.

For leaders who struggle with this, she suggests a simple tracking method: write down your team members’ names and add a checkmark each time you recognize them. Her point: everyone needs to hear they’re on the right track, probably more often than you think.

In that spirit, while the book is about Kiner’s experience, it also puts a big focus on the team that makes Reverb work, including co-owner and COO Sarah Wilkins, whom Kiner describes as the person who kept the company running during the worst stretches.

What’s happening now

As candid as the book is about the downturn, things have shifted since Kiner finished writing. In the weeks before our recent conversation, she said, new deal volume had jumped 50%, across tech, nonprofits, and small businesses. Reverb is hiring consultants again.

“I literally can’t explain it,” she said, noting that the turnaround has been happening despite inflation, gas prices, and geopolitical turmoil such as the war in Iran. 

AI is a frequent backdrop and topic of conversation in their work. Kiner writes in the book, for example, about teams at some companies being told to double productivity with AI but getting little support. 

In our conversation, she described a split: companies using AI as a way to demand more, and those actually bringing people along, showing them how to save time.

She’s not worried about AI replacing the human side of her work. One of her advisors uses a term she likes: “connective labor,” referring to empathy, conflict resolution, and the work of helping people and teams get unstuck. That part, she said, isn’t going away.

“I think there’s room for all of us,” she said. “Us and the agents, too.”

“The Reverb Way” is available in paperback and e-book versions.

Editor’s Note: GeekWire is a Reverb client.

Microsoft is offering a one-time voluntary retirement program for the first time in its 51-year history, giving thousands of long-serving U.S. employees a chance to leave with a financial payout and extended healthcare as it works to control costs amid a massive buildup in AI infrastructure.

An estimated 7% of Microsoft’s 125,000-person U.S. workforce, or about 8,750 employees, would be eligible based on a formula that takes into account their years at the company and their age.

It’s a highly unusual move in the tech world. Voluntary retirement programs are common in older industries, such as telecom and manufacturing, but the largest tech companies have instead turned to layoffs, stricter performance reviews, and return-to-office policies to thin their ranks. 

Microsoft itself laid off more than 15,000 employees last year and began requiring workers in the Seattle region to return to the office three days a week in February.

The retirement program was outlined Thursday in a memo to employees from Chief People Officer Amy Coleman, who described it as a one-time offering for long-serving workers.

The program is open to U.S. employees at Level 67 — the equivalent of senior director — and below, excluding those on sales incentive plans, whose years of service plus age total 70 or more. Eligible employees will be notified May 7 and will have 30 days to decide.

“Many of these employees have spent years, and in some cases, decades, shaping Microsoft into what it is today,” Coleman wrote. “Our hope is that this program gives those eligible the choice to take that next step on their own terms, with generous company support.”

In the same memo, Coleman outlined changes to Microsoft’s compensation system, reducing the number of pay levels from nine to five. It’s also decoupling stock awards from bonuses, giving managers flexibility to use stock to reward long-term contributors regardless of their latest performance rating.

Microsoft isn’t providing specific details of the retirement package yet, saying eligible employees and their managers will receive more information on May 7. Details of the healthcare component will be significant for employees who are not yet old enough to qualify for Medicare at age 65.

There are not expected to be any restrictions on future employment for those who take the deal. 

The program would take effect in Microsoft’s fiscal fourth quarter, and CFO Amy Hood is expected to discuss it on the company’s earnings call next week.

Listen to GeekWire’s Todd Bishop discuss the news on KIRO Newsradio:

LinkedIn has a new CEO for the first time in six years.

Daniel Shapero, the company’s chief operating officer since 2021, is stepping into the top job, reporting to Ryan Roslansky, who was elevated last year to executive vice president overseeing both LinkedIn and Microsoft Office.

The changes come as LinkedIn crosses $5 billion in quarterly revenue for the first time, putting it on an annual run rate of more than $20 billion. The business social network has been owned by Microsoft since its acquisition by the Seattle-area tech giant for $26.2 billion in 2016.

Roslansky announced the changes Wednesday (in a post on LinkedIn, of course) saying he also asked Mohak Shroff, LinkedIn’s longtime engineering leader, to take on the new role of president of platforms and digital work. Both Shapero and Shroff report to Roslansky.

Roslansky put the moves in the context of the accelerating impact of AI on the labor market.

“Last year when Satya Nadella asked me to lead LinkedIn and Microsoft Office, I knew what he was betting on: AI is going to transform how people work and grow in their careers faster than most people expect,” he wrote. “And LinkedIn and Office would be at the center of that.

As LinkedIn CEO since 2020, Roslansky nearly tripled the company’s revenue and grew the platform to more than 1.3 billion members, 70 million companies, and 42,000 skills, according to the company.

He took on the additional role of EVP of Office last June, adding oversight of Outlook, Word, Excel, PowerPoint, and Microsoft 365 Copilot as Microsoft pursued its “agentic web” AI strategy.

The new leadership structure is designed to free Roslansky to focus on that broader portfolio. Shapero will run LinkedIn day to day, while Shroff will work across LinkedIn and Microsoft on longer-term technology strategy and innovation.

Shapero joined LinkedIn in 2008 as roughly its 300th employee. He rose through sales, product, and operations before becoming COO in 2021. In his own LinkedIn post, he called his time at the company “one of the most meaningful experiences of my life” and said he would start by “learning and listening.”

Feroze Motafram is an operations consultant based in Sammamish, Wash., and founder of Avestan LLC. This piece is adapted from a LinkedIn post.

Someone asked me recently what made me think about writing this. The trigger, I told them, was simpler than you might expect.

I live in Sammamish, in the shadow of Microsoft’s looming presence. Microsoft employees are my neighbors, my social circle, the people I run into at weekend gatherings. Over time I noticed that conversations with them had a distinctive gravitational pull — always inward, toward reorgs, internal politics, who reports to whom now, who’s ascendant, who’s out. Customers were rarely part of the conversation. This usually means navigating the organization has become more consuming than building anything within it.

Microsoft’s stock decline and the softening of real estate in this corridor (both affecting me personally) were the prompts to write it down. The material was already sitting in front of me.

I should be clear about what I am and am not. My formal training is in electrical engineering. The primary instruments of my early career were set squares and slide rules, which will tell you something about both my vintage and my domain. I have spent the intervening decades as a senior executive at Fortune 100 companies and, more recently, as an operations and supply chain consultant. I build and fix things: supply chains, organizations that have lost their way. What I can offer is not insider knowledge. It is 30 years of pattern recognition, applied to what is visible from where I stand.

This is the lens I am bringing. Take it for what it is worth.

The market is asking a question

Microsoft stock declined roughly 25% in Q1 2026, representing its worst quarterly performance since the 2008 financial crisis despite blockbuster results. The market may overreact, but it is not stupid. When the stock of a company of this scale underperforms that of its peer group by double digits, the question worth asking is not “is this a buying opportunity.” The question is: what does the market understand about this organization that the headlines don’t capture?

Part of the answer is visible in the financials. A striking portion of Microsoft’s forward revenue backlog is tied to a single counterparty, OpenAI, an unprofitable startup that has since signed a landmark cloud agreement with Amazon, directly challenging the Azure exclusivity Microsoft had treated as a cornerstone of its AI strategy. Meanwhile, Microsoft is building its own internal AI model as a hedge, an expensive bet layered on top of an already expensive bet.

But the part that does not show up in an earnings report may be the more consequential story. That is what I want to offer here.

The monopoly dividend, and its hidden cost

For the better part of three decades, Microsoft enjoyed something very few companies in history have had: a captive market. Enterprise customers did not use Office because they loved it. They used it because leaving was more painful than staying. That distinction between loyalty and lock-in matters enormously, and it is one that organizations rarely make honestly about themselves.

When your customers cannot leave, the feedback loops that drive genuine innovation go silent. The tendency is to stop asking “what does the customer need?” and start asking “what can we get away with?” Processes multiply. Committees proliferate. Bureaucracy thrives. The organization optimizes for defending territory rather than creating it.

This is not a character failing. It occurs insidiously and unconsciously. It is an entirely rational organizational response to a monopolistic competitive environment. But it leaves a mark. And that mark does not disappear simply because the competitive environment changes.

Satya Nadella earned his laurels, but the work isn’t finished

The Azure pivot was a genuine strategic achievement, and Microsoft CEO Satya Nadella’s cultural reset from “know-it-all” to “learn-it-all,” as he framed it, was real and necessary. The stack-ranking era that preceded him did generational damage to Microsoft’s ability to collaborate, retain talent, and take meaningful risks. He arrested that decline and deserves full credit for it.

But here one must tread carefully. Stack ranking was formally abolished in the final months of Steve Ballmer’s tenure. The announcement was celebrated, the headlines were laudatory. What is rather more interesting is what one hears in conversations since. Ask Microsoft employees about the performance review system that replaced it, and the response is rarely enthusiastic. Whether the underlying mechanics genuinely changed, or whether the organization simply learned to dress the same instincts in more palatable language, is a question I cannot answer from the outside. What I can observe is that the people doing the work don’t appear to believe the answer is reassuring.

Cultural transformation in a 220,000-person organization moves at a glacial pace. You can change the language in a decade. Changing the instincts takes considerably longer. One has to wonder how many of the engineers and managers who learned to survive the Ballmer years by navigating politics rather than building products have since moved on, and how many remain, in leadership positions, still oriented by instinct toward self-protection over bold action.

What I can observe is the output. Copilot (inarguably Microsoft’s most strategically critical product) has converted just 15 million paid subscribers from a captive base of 450 million Microsoft 365 users. That is 3.3%. When your own customers will not buy what you are selling at scale, it is worth asking whether the product is genuinely solving a problem or simply a feature in search of a use case.

Microsoft’s internal preoccupations do not stay inside the building. I have observed versions of this dynamic before, most vividly when I lived in Brookfield, Wis., in the orbit of GE Healthcare’s then-headquarters. But what I observe in this corridor is of a different magnitude. It is not just politics that dominates the conversation. It is the organization itself — its structure, its hierarchies, its shifting priorities — that has become the primary subject of intellectual energy.

The campus, in a very real sense, has become the product. When navigating the organization becomes more consuming than building anything within it, that is not a criticism of the individuals. It is a diagnosis of the system they are operating inside.

The human capital story no one is writing

There is a dimension to this that the financial press has largely missed, and I raise it because I see it in my community every day… including, in ways I did not anticipate, in my own backyard.

A significant proportion of Microsoft’s engineering talent (and the engineering talent of the broader Seattle tech corridor) consists of H-1B visa holders. These are exceptional professionals: highly educated, deeply skilled, often carrying decade-long career investments in the United States. They have built lives here. Many have children born here. They have been, in many cases, the intellectual engine of the products Microsoft is depending on to compete in the AI era.

That population is operating under a level of personal anxiety that is, in my observation, without modern precedent. Travel advisories from their own employers. A $100,000 petition fee for new visa applications. Proposed rule changes touching birthright citizenship. A policy environment that sends a clear and unambiguous message: your presence here is conditional, negotiable, and subject to revision without notice.

The behavioral consequence of that anxiety is not visible in a quarterly earnings report. But it is real and consequential. People operating under existential personal uncertainty do not take professional risks. They do not champion the bold new initiative. They do not volunteer for the high-visibility project that could fail. They execute reliably on what already exists and protect their position. In an organization that already has a cultural predisposition toward risk aversion, this compounds the pathology in ways that will show up — perhaps not this quarter, but in the product decisions made over the next eighteen months.

The effects are visible beyond the campus walls. Conversations with real estate professionals in this corridor tell a consistent story: demand from this community, which has historically been among the most financially capable buyers in the region, has softened measurably. Not because the finances have changed, but because the horizon has. When you are uncertain whether your visa will be renewed, or whether your children’s citizenship status may be revisited, you do not buy a house.

The softening of demand is not merely an abstraction for those of us who live here. But the more significant consequence is not measured in property values. It is measured in the quality of risk-taking inside those campuses. And risk-taking is precisely what Microsoft needs most right now.

The case for optimism, and why it requires more than patience

None of this is to suggest Microsoft is broken beyond repair. Betting against Microsoft has historically been an enterprise for the foolhardy. The balance sheet remains stellar. The enterprise relationships are genuinely extraordinary. Ripping out Azure, Teams, and the M365 stack is not a decision any CIO makes lightly. The installed-base moat is real, and should not be underestimated by anyone, least of all an operations consultant from the suburbs.

What I would offer, more modestly, is this: the bull case requires more than a great balance sheet and sticky products. It requires an organization capable of genuine innovation at speed. Which in turn requires a culture that rewards risk, retains its most creative talent, and executes with urgency. Whether Microsoft can summon those qualities at this particular moment is a question I cannot answer with conviction.

What I can say is that the market, which is considerably more qualified than I am, appears to be asking the same question. The valuation has compressed to levels not seen in a decade, briefly falling below the S&P 500 for the first time in a generation. That is not the posture of a market betting with conviction that the answer is yes.

Perhaps it should be. I honestly don’t know. What I do know is that the signals visible from outside the building — from the neighborhood, from weekend gatherings, from the casual conversations — are worth paying attention to. They usually are.

This week on the GeekWire Podcast, a week of Seattle-area startup news shows how the AI era is reshaping the regional tech scene. Q1 venture numbers reveal bigger checks going to fewer companies, with Seattle slipping behind the likes of Austin and Miami on deal volume.

And yet the distributed nature of modern startups is complicating what it even means to be a regional tech hub. (Does a mailbox in Pioneer Square really count as a Seattle headquarters?)

Founders and CEOs are navigating this in different ways. Those with enough cash are eyeing strategic acquisitions, including opportunities to absorb startups caught up in the AI shakeout.

Many are also rethinking how they hire and expand. More than a third of the GeekWire 200, our ranking of top Pacific Northwest startups, saw year-over-year employment declines, as agents boost individual productivity and reshape the workforce.

Plus: Andy Jassy’s shareholder letter signals Amazon is making bets again, in areas including chips and robotics. Driving home the point, the tech giant’s Amazon’s ambitious Globalstar acquisition effectively means it’s inheriting Apple’s satellite roadmap.

Of course, we have to talk about Allbirds. The sustainable shoe brand, which once challenged Amazon over knock-off sneakers, pivoted to AI infrastructure and saw its stock soar.

And in our final segment, a trivia challenge on the No. 1 companies in GeekWire 200 history.

With GeekWire co-founders John Cook and Todd Bishop. Edited by Curt Milton. 

Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you listen.

When Bill Belichick famously spiked a Microsoft Surface tablet as the New England Patriots head coach in 2016, the moment was a symbol of frustration with technology on the sidelines.

But inside the company, it actually became a point of pride. Thomas Labuzienski, a senior partnerships manager at Microsoft, shared the rest of the story with us this week: the Belichick device made it back to the company — where testing showed, amazingly, that it still worked.

That rugged (and apparently indestructible) hardware remains the tech giant’s most visible presence on NFL game days. But these days, the action is inside: Copilot, AI agents, and live data tools that are changing how coaches, players, and scouts prepare and compete.

That was one of the takeaways from an event Wednesday at Seattle’s Lumen Field, where Microsoft walked content creators and this reporter through some of its latest NFL technology. High above the field, social media and YouTube influencers posed for photos and selfies with a Microsoft Surface Copilot+ PC much like NFL fans would with their favorite players.

It was an off-season glimpse into the evolution of Microsoft’s partnership with the league, which started in 2013 and was extended last summer to go deeper on AI and cloud tools.

Just as with business workflows, these tools are changing how teams operate on and off the field. One big benefit of AI, Microsoft says, is speed: getting the right information to the right person fast enough to shape the next play call or adjustment. But the coaches, players, and staff are ultimately still calling the plays, both literally and figuratively.

“We’re not using AI and technology to make decisions for them,” Labuzienski said. “They’re the experts. They know what they’re trying to do.”

Even with baseball and soccer well under way, Microsoft seems as excited as ever about football, especially with its hometown Seahawks coming off a Super Bowl championship.

Here’s what we saw and learned during the event:

Copilot filtering: The NFL’s Sideline Viewing System now features Copilot-powered filters that let coaches and players instantly sort plays by down, distance, quarter, and gain type. 

Labuzienski demonstrated how a play that took a minute to find manually could be located in seconds using the filters. “That five seconds can mean giving my coach the right information so they can make the right adjustment or the right play call,” he said.

Excel on the sidelines: One analyst per team has access to a real-time Excel dashboard in the coaches booth that pulls live play-by-play and player usage data from the NFL. 

Coaches can load custom templates before the game and use Copilot to run analyses on the fly — tracking formation tendencies, snap counts, and player load without having to manually write formulas mid-game.

NFL Combine: Microsoft built a custom AI agent trained on 10 years of NFL Combine data, allowing scouts to query prospects using natural language. This helps teams assess players faster than ever during one of the most time-sensitive periods of the NFL calendar.

For example, Labuzienski said, “If I really love this defensive lineman, I can pull his data compared to other prospects over the last 10 years.”

Tablets on game days: Hardware is still a key part of the equation, of course. Labuzienski said there are 2,500 Copilot Plus PCs throughout the league on game day, with 20 on each sideline, and 10 in each coach’s booth.

Tech as a coaching advantage: Back when Belichick got his start, NFL sidelines relied on thermal printers hardwired to stadium cameras. It was a process that took minutes, with runners ferrying photos into binders for coaches and players on the sidelines. 

Technology has transformed all of that, and a new generation of coaches has grown up never knowing anything different.

Seattle Seahawks head coach Mike Macdonald, who has openly embraced his reputation as a “football nerd,” is a prime example. Labuzienski said he represents exactly the kind of coach Microsoft has built its NFL partnership around. Macdonald views technology as a differentiator across the league and has said it’s an advantage for the Seahawks to embrace it.

Even coaches less enthusiastic about AI are coming around, Labuzienski said, if only to avoid falling behind. “They do, sometimes begrudgingly, study up and make sure they get the most out of the technology,” he said.

Digital-first players: Labuzienski said he’s seen players become some of the most engaged users of sideline technology — sometimes more so than the staff around them.

“I was at the preseason game with the Bills against the Bears, and after every single drive, Josh Allen picked up this device,” Labuzienski said of the Bills QB. “He was talking with the other quarterbacks, and they were using it to go through the plays from that previous series.”

The system is designed to make that kind of real-time collaboration easy. The 20 devices on each sideline are all connected, with coaches in the booth able to write notes directly to individual player profiles — so a quarterback picking up a tablet can instantly see what his position coach upstairs is flagging, without having to radio up or wait for a break in the action.

That dynamic is likely to only deepen as younger players enter the league. 

Microsoft was at last year’s Rookie Premier — an event where roughly 40 top draft picks gather for brand activations ahead of their first NFL season — and showed the incoming class the Sideline Viewing System for the first time. The reaction was telling.

“A lot of them really quickly picked it up and were able to click around,” Labuzienski said. “They’re very digital-first — very savvy.”

Microsoft says its Recall app — which captures and stores screen shots every few seconds — is safe. Security researchers keep saying otherwise.

Recall was originally billed as a “photographic memory” to store everything Windows users do on their computers. People could then see some of those screen shots at a later time by searching AI with plain-text queries such as red barn. (See illustration above.)

Select members of Microsoft’s exclusive Windows Insider program have had access to Recall for more than a year. Users of AI-enabled Copilot+ PCs started receiving Recall as an opt-in feature in April 2025, one year ago this month.

But since its debut, experts have repeatedly demonstrated that hackers can access the data Recall stores. This raises questions about whether a tool that records your entire digital life can ever be adequately secured. The situation is creating uncertainty about Microsoft’s plans to make Recall more widely available on all PCs.

Alexander Hagenah, executive director of SIX — a Zürich-based technology company that operates infrastructure for stock exchanges in Switzerland and Spain — described Recall’s security weaknesses in a LinkedIn post in April 2025. He also released an app he called TotalRecall that could “extract all captured windows and images taken by Recall … nothing encrypted, no rocket science needed.”

Joining other researchers, the University of Pennsylvania’s Office of Information Security released a warning on Apr. 14, 2025, about the version of Recall that was then available. The university’s announcement stated that Recall “introduces substantial and unacceptable security, legality, and privacy challenges” [emphasis in the original]. The statement added that administrators of “Windows environments at Penn are strongly urged” to disable Recall.

In response to criticisms such as these, Microsoft — to its credit — pulled back on its plans to roll out Recall to all Windows 11 PCs that met fairly high system requirements (including a neural processing unit and eight logical processors, according to an MS Learn document). Instead, the company announced in a blog update on June 13, 2024, that Recall would become available only to participants in the company’s much smaller Windows Insider program.

In the time since that decision, the fate of Recall has become even murkier. Journalist Zac Bowden wrote in a Windows Central blog post on Jan. 30, 2026, that Microsoft is “pulling back its Windows 11 AI push with a major Copilot and Recall rethink.”

The problem is that it’s tough for software engineers to make data ultra-convenient for end users to access while simultaneously securing it so it’s impervious to hackers.

It’s hard to remember that the company’s original goal was ease of use, now that Microsoft’s focus has changed to making the security of its screen-cap app impenetrable.

Microsoft says Recall blurs images of credit-card numbers, bank passwords, and other personal data — or doesn’t store them at all. But security experts are still not convinced.

After testing the latest version of Recall, Swiss technologist Hagenah recently issued a new proof-of-concept called “Total Recall Reloaded” on a GitHub page. In his comments, Hagenah said any malware running on a user’s PC can copy every Recall screen shot as it passes through in-process memory: “No admin required. Standard user. No kernel exploit.”

Hagenah has not publicly disclosed some security holes, saying he’s reported them to Microsoft and won’t release the technical details until the Redmond company has fixed the problems.

Already, malicious hackers have written code to take advantage of Recall’s screen shots. The malware can access Recall’s own memory to copy screen caps and send them to a faraway server. Hackers no longer need to write such code from scratch. (The procedure is described in a technical overview by cybersecurity writer Kevin Beaumont.)

At this writing, fewer than 10% of Windows 11 PCs can enable and run the current version of Recall. Microsoft representatives responded to my inquiries about plans for the app’s future availability by pointing to a Sept. 27, 2024, security update and an Apr. 25, 2025, blog post.