I have an ace of spades on the outside of my left wrist and Sic Parvis Magna across the left side of my chest. The first is a card I keep ready. The second is a line that has defined how I have lived and how I will live. For those who cannot read Latin, I have been a gamer since SimCity 2000, Worms, Jazz Jackrabbit, Star Wars: Dark Forces, three decades of hardware changing underneath me while the thing on the screen kept being the thing I came back to. I still play. I still follow gaming news the way other people follow politics or sport.

So when I tell you that the most important thing Microsoft is doing in 2026 is happening inside Xbox, I want you to know where the observation is coming from. And from what I can see, no one is reading it correctly.

On the twentieth of February, Microsoft announced that Phil Spencer was retiring after thirty-eight years at the company and twelve leading the gaming division. His successor was Asha Sharma, until that day the President of Microsoft’s CoreAI Product organisation. CoreAI is the group that runs GitHub Copilot, Visual Studio Code, Azure AI Foundry, and the developer-facing AI tooling that Microsoft has spent the last two years putting in front of every enterprise customer it has. Sharma had been at Microsoft for two years. Before that, COO at Instacart and VP of Product at Meta. She was not a gamer. Her Xbox profile was created in January, one month before her appointment. Sarah Bond, the Xbox President widely expected to inherit, departed alongside Spencer.

Sharma’s opening statement to the company contained a sentence that has been quoted everywhere since. “We will not chase short-term efficiency or flood our ecosystem with soulless AI slop.” The line was striking enough that the gaming press ran it as the headline. A new CEO arriving from the AI division of Microsoft, on day one, adopting the audience’s own vocabulary against AI. Slop was not a Microsoft word. Slop was the word the gaming community had coined as Microslop, a term of derision aimed at Microsoft’s own AI push that had become culturally specific enough to get banned briefly from the official Copilot Discord earlier in the year.

Two months later, the moves arrived in sequence. April twenty-first, the Game Pass Ultimate price cut from $29.99 to $22.99, and Call of Duty pulled from day-one inclusion in the subscription. April twenty-third, the “We Are Xbox” memo, with daily active players named as the new north star, replacing revenue and engagement metrics that had governed the division under Spencer. May fifth, the platform reorganisation. Four senior executives parachuted in from CoreAI, the division Sharma had just left. Jared Palmer, formerly at Vercel and GitHub, taking product, engineering, developer tools, and infrastructure. Tim Allen, design. Jonathan McKay, formerly head of growth at OpenAI, taking growth, data platform, and analytics. Evan Chaki, forward-deployed engineering. David Schloss from Instacart, subscription and cloud. Two twenty-four-year Microsoft veterans, Kevin Gammill and Roanne Sones, departed the same day. Same announcement, same memo, Sharma confirmed that Copilot for Gaming would be wound down on mobile and stopped on console.

Read separately, each move was covered by the gaming press as a piece of consumer-friendly leadership. Game Pass cheaper. AI assistant killed. Microsoft Gaming renamed back to Xbox. Console recommitted to. The “This is an Xbox” multi-device campaign axed. The community responded with cautious optimism. Hideo Kojima posted a photograph with Sharma. Mountain Dew replied to an Xbox post with a green heart. Threads lit up with the green-is-back rallying call that had been dormant for five years.

That is not the shape of consumer-friendly leadership. That is the shape of something else.

What Sharma did on day one was not principled refusal. It was positioning. Skilled product leaders read the room and adapt language accordingly. Meeting an audience in the words they already use is a competence, not a manipulation. But the adaptation is real, and naming it matters, because the public-facing register a leader chooses for a hostile audience is not necessarily continuous with the strategic register the same leader operates in privately.

Sharma had large shoes to fill taking over from Phil “one of us” Spencer. She needed the same buy-in. But how do you earn the “one of us” label without being a gamer? Quite simply, you talk shit about your boss’s dumb new direction at the watercooler. It worked in Office Space, Horrible Bosses, The Office, Grandma’s Boy, and Friends. The cultural air bought by the disavowal is what protects the work happening underneath.

What I am calling the substrate is older than Sharma’s appointment, and it is not hidden. Microsoft has said most of it out loud.

In March 2025, fourteen months before Sharma took the Xbox role, Microsoft quietly retired the distinction between personal and work Microsoft accounts. A single email-based identity now signs a person into Windows, Xbox, and Microsoft 365 consumer services. The change reached roughly a billion users globally and was framed at the time as a friction reduction. It is also the architectural precondition for treating those surfaces as one behavioural environment rather than three.

The advertising business has been operating against that environment for over a year, and the corporate vehicle that operates it has been quietly absorbed. Activision Blizzard Media, the in-house advertising sales business that came with the Activision purchase, has been renamed Xbox Media Solutions and folded into Microsoft Advertising’s product hierarchy. Visiting the original activisionblizzardmedia.com domain redirects to a Microsoft Advertising page where gaming sits under the Technology and Services category, sold as ad tech rather than as media inventory. The April 2026 cross-platform pitch describes King, Xbox, and Microsoft Casual Games as a single buyable surface for brand campaigns, with the materials noting that one in three players engaged with Microsoft gaming properties engages with more than one of them. The first menu item under Technology and Services is Agentic Commerce. Gaming Solutions is the third. The same word the gaming community had spent months using as criticism, the word Sharma’s day-one statement positioned Xbox against, sits as the lead category in Microsoft Advertising’s product hierarchy in the same quarter Sharma’s disavowal was issued. Two audiences. Two registers. One company.

In April 2026, the United Kingdom’s Competition and Markets Authority opened a formal investigation into Microsoft’s integrated AI offering across Copilot, Windows, and Office. Among the questions the CMA is examining is whether Microsoft should be required to provide more open access to that data ecosystem for competing AI services. A regulator has now formally treated the cross-product Microsoft data layer as something competitors might need protection from. That is not the regulator getting ahead of an emerging concern. It is the regulator catching up to one.

At the executive level, the framing has been more direct, and the response from the audience has already taught Microsoft something. In November 2025, the President of Windows and Devices, Pavan Davuluri, posted a short statement ahead of Microsoft Ignite describing Windows as “evolving into an agentic OS, connecting devices, cloud, and AI to unlock intelligent productivity.” The reaction was immediate and negative enough that Davuluri locked replies within hours. Within weeks, reporting from Windows-focused outlets confirmed that Microsoft was quietly reevaluating its AI strategy on Windows. By early 2026, visible Copilot integrations in Notepad and Paint were being trimmed. Recall, the screenshot-and-search feature that had drawn earlier privacy backlash, was reportedly viewed internally as a failure. The visible layer was being scaled back. At the same time, the substrate-layer AI work, Semantic Search, the Agentic Workspace, Windows ML, the Windows AI APIs, continued as planned.

That sequence matters because it is the rehearsal for the playbook Sharma now runs at Xbox. The visible AI surface that drew community backlash was trimmed. The substrate underneath it was not. Three months later, Sharma arrives at the only consumer surface in the Microsoft estate more hostile to AI than the Windows power-user community. She walks in with the lesson already learned. Disavow the visible layer on day one. Continue the substrate work without naming it. By the time the visible Copilot for Gaming gets wound down in May, the move reads as responsiveness to the community. It is also the same move Microsoft was forced into on Windows three months earlier, executed this time without needing to be forced.

I will be honest with you. I always had a soft spot for Phil Spencer. I believe he operated with the gaming community in mind. We were the audience, flawed misfits in an easy-to-categorise online user segment, harder to manage, and he understood us in a way only someone from inside that layer could. The decision to deplatform games and put us on what felt like a too-good-to-be-true subscription deal in Game Pass was something I genuinely respected. When there was a storm, he took the uncomfortable interviews himself, often in fan channels, talking to us rather than around us. It was refreshing and it felt authentic. If the latter was just a façade, he deserves an Oscar.

But the most explicit public statement Spencer made on AI, four months before his retirement was announced, was a careful articulation of how Xbox studios would relate to AI tools. AI use at Xbox, Spencer said, was much more operational than creative. Studios were free to use AI development tools or not. There were no production mandates. Sarah Bond’s public position throughout her tenure was AI-positive in the same gaming-internal register, framed around creator democratisation and graphics innovation. Both positions treated AI as a tool inside the studio process. Useful, optional, applied to the work of making games. That is a real and coherent view of AI. It is also not the view of AI the rest of Microsoft has been operating under for the same period. The unified sign-in, the advertising graph across King and Xbox and Microsoft Casual Games, the Xbox Media Solutions rebrand, the agentic commerce navigation, the agentic-OS framing that triggered the Davuluri backlash, the CMA investigation. All of that was being assembled at the company layer while Xbox ran the gaming surface with the AI-as-tool framing intact.

Sharma’s CV on the other hand was spearheaded with platforms not products. Instacart was the company that turned grocery shopping into a behavioural-data platform underneath the surface of an app. Sharma was its COO. CoreAI was the division building the developer-facing AI tooling Microsoft sells across its enterprise estate. Sharma ran it. In an interview six months before her Xbox appointment, she described products as organisms that adapt rather than artefacts that ship, predicted that agents would outnumber humans inside companies within the year, and noted that Microsoft’s planning framework had moved from annual roadmaps to seasons. The vocabulary is product-organism, agent-fluent, dataflow-native. None of it is the vocabulary of running a console business. All of it is the vocabulary of running platform infrastructure that surfaces feed into.

The surface produces inputs. The platform underneath is the asset. That mental model is the inheritance Sharma actually arrives with.

To understand why those moves matter, the financial picture has to be on the table.

In the quarter ending March 2026, Microsoft reported $82.9 billion in revenue. Productivity and Business Processes brought in $35.0 billion, up 17 percent year-over-year. Intelligent Cloud brought in $34.7 billion, up 30 percent, with Azure growing 40 percent. More Personal Computing, the segment that contains Windows, Xbox, Surface, and Bing, brought in $13.2 billion and was the only declining segment. Inside it, gaming revenue fell 7 percent. Roughly 84 percent of Microsoft’s revenue is now enterprise commercial cloud, productivity, and developer infrastructure. The 16 percent that touches consumers is shrinking. The AI business surpassed an annual revenue run rate of $37 billion, up 123 percent year-over-year. Microsoft’s commercial enterprise pillars are growing at a pace that has few comparators in technology. Its consumer pillar is the slowest-growing piece of the company.

That gap is where the structural problem sits. Microsoft has no consumer mobile operating system. Windows Phone died in 2017. iOS and Android divide mobile, where most consumers spend the largest share of their attention. Microsoft has no consumer social graph. LinkedIn is professional. There is no Microsoft equivalent of Facebook, Instagram, WhatsApp, or TikTok. Microsoft has no consumer media platform at scale. No Netflix, no Spotify, no YouTube. Bing is a distant second to Google in consumer search. There is no Microsoft consumer payments platform with mindshare. Consumer hardware is a niche line. Maps, photos, voice assistants in the home, wearables, all minimal or absent. Microsoft built the strongest professional behavioural graph in technology and has almost nothing on the consumer-leisure side.

Read against that gap, the Activision Blizzard King acquisition stops looking like a gaming expansion and starts looking like the only available answer to a structural problem. Sixty-nine billion dollars, the largest tech acquisition in history, paid for the only Western-owned target that delivered top-tier coverage across all three consumer-leisure platforms in a single transaction. Mobile through King at 238 million monthly active users. PC through Battle.net and the Blizzard catalogue. Console through Activision and Call of Duty. Electronic Arts would have given Microsoft console and PC, with thin mobile. Take-Two pre-Zynga was console and PC only. Ubisoft was console and PC. Embracer was a roll-up of smaller franchises with no mega-asset. ABK was the only one that closed the gap. The eight days between Take-Two’s announcement of the Zynga acquisition on the tenth of January 2022 and Microsoft’s announcement of the ABK deal on the eighteenth tell their own story. Two of the largest Western publishers moved within a week of each other to consolidate platform breadth across mobile, PC, and console. Microsoft simply paid for the larger answer.

Sharma’s job, read structurally, is not to grow Xbox as a gaming business. It is to operate the only consumer-leisure pillar Microsoft has, at the moment its enterprise pillars are growing 17 to 30 percent annually and its consumer pillar is the only declining segment in the company. The mismatch between commercial momentum and strategic need is severe. Enterprise pays the bills. Consumer-leisure closes the gap that the agentic platform thesis cannot work without.

The clearest evidence of how the pillar is being operated sits in the studios that have been closed.

Since the ABK acquisition closed in October 2023, Microsoft has run four waves of cuts inside its gaming division. January 2024, 1,900 employees. May 2024, three studios shuttered outright: Tango Gameworks, Arkane Austin, and Alpha Dog Games. September 2024, another 650. July 2025, the largest wave: 9,000 across Microsoft, with The Initiative closed, Perfect Dark and Everwild cancelled, Turn 10 cut by roughly half, and significant reductions at Raven, Sledgehammer, ZeniMax, Blizzard, and King.

The cost-reduction explanation does not survive contact with Microsoft’s financials. Quarterly net income last quarter was $31.8 billion. The annual cost of operating every studio Microsoft has closed since the ABK deal does not measurably register against numbers like that. Microsoft could have absorbed every closure as a rounding error and kept the studios running for prestige, recruiting, or creative-portfolio reasons. The choice not to do so is the data point.

Look at which studios closed and which survived, and the optimisation function becomes visible.

The closed studios produced critical acclaim without continuous transaction-behaviour data. Tango Gameworks shipped premium one-time-purchase titles. Hi-Fi Rush at thirty dollars, three million players, almost all of whom were already inside the Microsoft gaming ecosystem through Game Pass, Xbox console, or PC. The behavioural signal Tango generated was largely redundant with signal those players were already producing through their other Microsoft gaming activity. Arkane Austin, the same, especially when their latest triple-A title “Redfall” was met with such an underwhelming reception Phil Spencer had to personally apologise for it. Whilst Microsoft never released any sales numbers, they admitted in court filings that Redfall had generated “minimum sales.” The Initiative had not shipped a game, pitching a premium Perfect Dark reboot whose audience overlapped with Halo and Call of Duty. Alpha Dog made a free-to-play mobile Doom variant whose user base overlapped with King’s much larger mobile audience.

The surviving studios run continuous transaction engines at scale. Call of Duty generates among the largest microtransaction revenue in Western gaming, with battle passes, operator skins, and weapon blueprints producing high-frequency purchasing decisions across tens of millions of players. Blizzard runs four parallel monetisation engines. World of Warcraft on DLCs, subscriptions and in-game items, Diablo IV with DLCs, cosmetic shop and battle pass, Diablo Immortal with an aggressive in-game economy, Hearthstone with card-pack purchasing patterns going back over a decade. King is the canonical mobile microtransaction machine, with Candy Crush operating on impulse-purchase signal so dense that King’s analytics were called out as a strategic asset when Activision originally bought the studio in 2016.

The studios that produce premium one-time titles and survived nonetheless are the ones with anticipated future transaction-rich franchises. Bethesda Game Studios sits on Elder Scrolls VI, with a fourteen-year-old waiting fanbase and an almost certain expansion-pack tail. Fallout 76 already runs a cash shop. They also have Starfield, which launched to strong initial reviews but met an underwhelming user base, and has seen both DLC and ongoing updates since. What Bethesda extended onto Starfield in June 2024, eight months after the ABK acquisition closed, was an existing community-focused mod storefront. Creation Club had launched on Skyrim and Fallout 4 in 2017, before Microsoft acquired Bethesda. Microsoft inherited the model and chose to extend it. Bethesda takes a cut of every transaction. Microsoft captures purchase-behaviour data from an audience commercially difficult to reach through other channels. id Software’s Doom franchise carries growing cosmetic layers. MachineGames runs Wolfenstein and the Disney-IP Indiana Jones partnership, which delivered Indiana Jones and the Great Circle in late 2024 with further titles anticipated. The pattern is not “no microtransactions equals closure.” It is “no current transaction signal and no anticipated future transaction signal at franchise-event magnitude.”

The optimisation function being maximised across these decisions is not revenue or cost. It is unique-user-data-and-transaction-signal per engineering dollar. The studios whose audiences duplicate audiences served by surviving studios get closed. The studios whose audiences are unique, or whose next title is anticipated as a transaction-event of franchise magnitude, survive. The function does not require anyone at Microsoft to consciously devalue creative work. It just produces the closures the data shows. Critical acclaim is not weighted in the function. The franchises that survive every wave are the ones whose user bases the substrate cannot replicate elsewhere in the portfolio.

With all this in mind, let’s return to February 2026. Read the moves through the substrate.

The order of operations matters. On April 6, three weeks before the consumer-facing moves arrived, Microsoft Advertising’s gaming-vertical content function broke a six-month editorial silence to publish a manifesto under the new Xbox Media Solutions byline. The piece, titled “The Future Is In Play,” positioned gaming-derived advertising as outperforming online video and social media on viewability, attention, and brand impact. It named the products: Rewarded Video, Playables, Xbox Landing Experiences, Click-to-Engage on Xbox. It quantified the brand-fit multipliers across Forza, Xbox, and Candy Crush. The substrate’s commercial expression went live first. Everything that followed protects it.

The Copilot for Gaming wind-down removes the most visible target for the Microslop critique. It buys cultural permission for everything else. Visible AI dies. The audience cheers. The substrate work proceeds at the depth where the audience has no vocabulary to object.

The “AI slop” disavowal performs the same function in language. It imports the audience’s word into Microsoft corporate speech on day one. The leader who came from CoreAI six months earlier and predicted that agents would outnumber humans inside companies is now the leader who will not flood the gaming ecosystem with soulless AI slop. Same person. Two registers. The cultural air bought by the second register is what protects the work continuous with the first.

The Game Pass price cut maximises subscriber breadth at the cost of per-user margin. The new north star is daily active players. Each engaged subscriber compounds in value because the substrate makes the count produce more than ARPU multiplied by MAU. The same subscriber who cancels under thirty dollars but stays at twenty-three is now generating continuous engagement, transaction, and cross-platform identity signal that feeds the layer underneath. The price cut reads as consumer-friendly leadership. It is also a calibrated instrument tuned against a function in which the marginal subscriber is worth more than they were under Spencer.

The “Everything is an Xbox” reversal is the move that looks most like it contradicts the substrate strategy and actually expresses it most cleanly. Spencer’s strategy of putting Xbox identity onto Samsung TVs and iPhones and Steam Decks made sense as a service-everywhere bet on a content business that had lost the console war. It was substrate-blind in a specific way. Cloud gaming on a Samsung TV gave Microsoft the gaming session metadata and gave Samsung everything else. The cross-context behavioural signal, the device-level activity, the agentic AI execution layer all sat on hardware Microsoft did not own. Native cross-platform games on iOS and Android continue and expand, because Microsoft owns the application code and captures the behavioural data regardless of which app store distributed the binary. The cross-platform pitch in the April 6 manifesto is specifically about that layer. Killing the brand-only cross-device service distribution closes the leak. The cross-platform that produces substrate value stays. The cross-platform that did not produce it goes.

The CoreAI executive imports bring the operational expertise to extract the substrate value the manifesto promises advertisers. Palmer for developer tools and infrastructure. McKay, formerly head of growth at OpenAI, for growth, data platform, and analytics. Chaki for engineering simplification. Allen for design. Schloss for subscription and cloud. The departures, Gammill and Sones, held the platform-and-ecosystem jobs the new hires now do. The hiring profile is not gaming. It is the operational profile of a consumer-leisure platform being rebuilt for substrate value extraction.

Each move read alone is defensible as good leadership of a struggling gaming business. Read together, they are coordinated around protecting the cultural air for a substrate strategy Microsoft has openly stated, that a regulator is openly investigating, that Microsoft Advertising is selling to brand budgets right now, and that the gaming press has covered as five separate stories without seeing the choreography. The advertiser-facing pitch goes live on April 6. The Game Pass cut on April 21. The “We Are Xbox” memo on April 23. The CoreAI hires and the Copilot wind-down on May 5. Four weeks. The order is the evidence.

What changed under Sharma is not the leader’s competence relative to her predecessor. It is the company’s definition of the job. Spencer ran Xbox as a gaming business with creative integrity, and the substrate was built around him by other parts of Microsoft. Sharma runs Xbox as the consumer-leisure pillar of a platform whose substrate is now operationally live. Same surface. Very different objectives.

Let me grab a whisky diluted with some honest qualifications first.

The substrate is announced. Microsoft has not hidden it. The article connects what is on the corporate website to what Sharma has done in eight weeks. That connection is the contribution. The rest is public record.

The crisis-response reading is partially true. Hardware revenue is in free fall. Sharma is moving fast on every available lever because the financial pressure is real. The crisis sets the pace. It does not set the direction. The moves could equally be read as four independent responses to a single inherited crisis, and that reading is defensible. The reason I do not adopt it is that the substrate-protecting interpretation explains the order of operations more completely, including the April 6 advertiser-facing manifesto landing three weeks before the consumer-facing moves.

Players are not being experimented on. The substrate is a normal commercial system at the scale of a trillion-dollar company, operating with users’ continuous behavioural participation as one input. Nobody is being deceived. The thing worth noticing is just that the surface most players think they are on, the gaming surface, the Game Pass surface, the console surface, is one piece of a much larger system whose shape is now visible if you stand far enough back.

Phil Spencer was the leader who told the gaming community the truth about losing the console war and tried to build something honest around the loss. Game Pass at its peak felt like the deal nobody at a trillion-dollar company should have signed off on, and that was the point. He understood us as the audience that had been dismissed and overcategorised everywhere else, and he made room for us. I will genuinely miss the version of Xbox he ran. The version that comes next is a different job for a different reason, and the substrate that operates underneath it does not require anyone to grieve what came before.

I have been carrying this observation since February, deliberately avoiding theories like Copilot-integrated games with Copilot-driven Kinect and controllers, AI-generated remasters of classic games, and the cargo-cult IQ verdict of “diversity hire.” Fifteen years of building digital systems and three decades as a gamer gave me the kind of patience required, and that was the missing factor other publications didn’t have. Granted, that’s an expensive commodity for outlets relying on engagement. Watching Sharma’s first eight weeks. Reading the Microsoft Advertising pages. Looking at the closures, the timing, the pivot in language, the people. It has not been clean. The “This is an Xbox” campaign died loudly. Copilot for Gaming was demoed, beta-tested, then quietly killed. Game Pass prices were hiked and rolled back. From the surface, retreat after retreat. From the substrate, one consistent direction. Cloud, productivity, identity, advertising, commerce, and now consumer leisure, integrated into a single platform. Strategy at a trillion-dollar scale.

The hand has been played. We are looking at $37 billion USD worth of chips. The pitchforks and torches have for now been put down. Friends of tech and the misfits of gaming are calling it a retreat from AI. From where I am standing, a $37 billion all-in is not what I would consider a fold.