Microsoft

Microsoft is a US-based public technology company and one of three global hyperscalers (with AWS and Google Cloud) underwriting frontier-scale AI compute. Its AI strategy rests on two pillars: a cloud and commercial partnership with OpenAI, begun in 2019 with a cumulative investment of more than $13 billion, that gives it access to GPT-series models across Azure and its enterprise stack; and an internal Microsoft AI division, led since 2024 by Mustafa Suleyman, that builds in-house models and consumer products. Following a late-April 2026 restructuring of the OpenAI partnership, Microsoft has moved to reduce single-vendor dependence on OpenAI, releasing its own MAI model family and positioning itself, per Suleyman, to join Google DeepMind, OpenAI, and Anthropic as a fourth frontier lab.

Snapshot

Revenue and earnings

Capital expenditure and margins

Users

OpenAI partnership

The Microsoft–OpenAI relationship is the central commercial arrangement underpinning Microsoft's AI strategy. Microsoft has invested more than $13 billion across multiple rounds beginning in 2019 and holds rights to a large share of OpenAI profits up to a cap. Before April 27, 2026, OpenAI trained and served models on Azure exclusively, with Azure OpenAI Service reselling those models to enterprise customers; GPT models power Copilot across Windows, Microsoft 365, GitHub, and Bing.

A structural constraint in the 2019 deal gave Microsoft rights to anything OpenAI develops or acquires. This surfaced in the July 2025 Windsurf acquisition saga: Windsurf CEO Mohan refused Microsoft IP access, Microsoft blocked the carve-out, and OpenAI's $3B Windsurf deal collapsed, with GitHub Copilot's competitive position the underlying driver (Source: Raw Sources/Windsurf Acquisition Saga 2025.md).

On April 27, 2026, OpenAI and Microsoft announced an amended partnership. OpenAI can now sell its products across any cloud provider rather than exclusively through Azure. Microsoft no longer collects a revenue share on its own sales of OpenAI services but retains a 20% share of OpenAI's revenue through 2030, an equity stake reported at approximately 27% (valued near $135 billion per April 28 reporting), and access to OpenAI's models through 2032. The long-disputed clause that would have transferred OpenAI's IP rights to Microsoft once OpenAI declared "AGI" was scrapped (Sources: openai.com; theinformation.com).

A further amendment on April 30, 2026 extended the commercial agreement through 2032, converting Microsoft's license to non-exclusive, allowing OpenAI to distribute across any cloud while still prioritizing Azure, and capping Microsoft's revenue share through 2030. The amendment extended Microsoft's IP access through 2032; its net effect was to preserve Microsoft's long-tail access while removing the legal friction blocking OpenAI's bilateral hyperscaler deals (AWS, Oracle, and others) (Source: openai.com). The residual-stake mechanics are documented further at OpenAI–Microsoft residual stake.

Regulators have scrutinized whether the arrangement is effectively a merger. The FTC's January 2025 6(b) staff report examined the Microsoft–OpenAI partnership, alongside Amazon–Anthropic and Alphabet–Anthropic, as a potential vector for competitive foreclosure in frontier AI. Microsoft is also named as a $250B Azure counterparty in circular-financing concerns raised in Sen. Warren Letter to OpenAI (2026-01-28) (see Circular Financing in AI).

Microsoft AI division and in-house models

In March 2024 Microsoft hired Mustafa Suleyman, co-founder of DeepMind and Inflection AI, to run a new Microsoft AI division, effectively absorbing most of Inflection's team in a transaction structured to avoid traditional merger review. The division's mandate is to build Microsoft's own frontier models and consumer AI products, reducing single-vendor dependence on OpenAI while keeping the partnership intact.

Cybersecurity models and agents

Microsoft launched MAI-Cyber-1-Flash, its first in-house cybersecurity model, at a San Francisco event on July 27, 2026, alongside Perception, an agentic security platform running red, blue and green agent teams. Mustafa Suleyman, CEO of Microsoft AI, said the model paired with GPT-5.4 inside the MDASH harness "beats out Gemini, GPT 5.5 Cyber, GPT 5.6 Sol, and Mythos 5 on Cyber Gym." Microsoft said the model carries roughly 90% of MDASH's workload, and prices Perception in Security Compute Units. Microsoft reported the MAI-Cyber-1-Flash and GPT-5.4 pairing scoring 95.95% on the benchmark it cited, at roughly half the cost of the prior configuration (Source: thehackernews.com; qz.com). Accounts of availability differ: SiliconANGLE reported public preview inside Microsoft Defender on August 3, TechCrunch on November 3 (Source: techcrunch.com; siliconangle.com). See AI and Cybersecurity, Autonomous cyber-agents.

MAI model family

Microsoft's in-house MAI model family, led by Suleyman, is the substance of its strategy to become an AI provider less dependent on OpenAI. As of late May 2026 the lineup spanned MAI-1-preview (a text/foundation model trained on approximately 15,000 Nvidia H100 GPUs), MAI-Voice-1, MAI-Image-2, and MAI-Transcribe-1, framed as a direct competitive move against OpenAI and Google (Sources: geekwire.com, venturebeat.com). The ~15,000-H100 training run for MAI-1-preview is modest relative to frontier-scale clusters. Microsoft's first-party strategy has drawn scrutiny over whether it can sustain frontier development alongside its OpenAI distribution role.

To support the in-house effort, Microsoft hired at least 10 former Allen Institute for AI (Ai2) researchers and staffers, including the core of Ai2's OLMo open-source model effort, for its Superintelligence team (Source: geekwire.com).

Build 2026 launch (seven new MAI models)

Microsoft first confirmed on May 28, 2026 that it would release a new in-house coding model the week of June 2, 2026, timed to its two-day Build 2026 developer conference in San Francisco on June 2–3, without initially disclosing the model's name, parameter count, architecture, or benchmarks. The release was positioned as a response to Cursor and Windsurf's encroachment on GitHub Copilot's developer base and followed reports that Microsoft was winding down internal Claude Code use (Source: cryptobriefing.com; see AI Coding Agents and AI Bubble Debate). Microsoft also confirmed Build would unveil new transcription, reasoning, speech, and image models alongside the coding model; its shares rose roughly 3% on the May 28 disclosure, which came as Anthropic's Claude Code overtook GitHub Copilot in AI-assisted-coding share. The reveal was timed against Anthropic's same-day Opus 4.8 release and Dynamic Workflows research preview, raising the question of whether GitHub Copilot would counter at the orchestration layer or only at the single-agent capability layer (Source: reuters.com).

At Build 2026 (June 2, 2026) Microsoft announced seven in-house AI models, including its first reasoning model, MAI-Thinking-1 (35B active parameters, 128K context), trained "from scratch" with no distillation, per Suleyman, who said the goal was to join Google DeepMind, OpenAI, and Anthropic as a fourth frontier lab following Microsoft's late-April separation from OpenAI. Microsoft committed to first-class OpenClaw support on Windows. The Verge characterized the event as a post-breakup competitive turn, headlining "Microsoft and OpenAI broke up — now they're ready to fight" (Source: theverge.com).

Per June 6 reporting, the seven models, beyond MAI-Thinking-1, include MAI-Code-1-Flash (a 5-billion-active-parameter coding model integrated into GitHub Copilot and VS Code), MAI-Image-2.5, MAI-Transcribe-1.5, and MAI-Voice-2. Suleyman said MAI-Thinking-1 was preferred to Anthropic's Claude Sonnet 4.6 in blind human side-by-side evaluations and was "trained from the ground up on clean data, without distillation from third-party models." Microsoft described the launch as a "hill-climbing machine" aimed at "long-term self-sufficiency" from OpenAI, co-designed with its own Maia 200 silicon, and paired it with a new Mayo Clinic collaboration to build a Mayo-owned frontier healthcare model (Source: microsoft.ai; nlp.elvissaravia.com; see Healthcare — AI Deployment). Follow-on June 5 reporting characterized MAI-Thinking-1 as reportedly trained entirely on Microsoft's own IP, consistent with Suleyman's "from scratch, no distillation" framing (Source: theverge.com).

Products and agents

Scout and the agent platform

At Build 2026 Microsoft introduced Scout, an OpenClaw-inspired, always-on personal AI agent inside the Microsoft 365 ecosystem (Outlook, OneDrive, Teams), billed as the first of a planned set of "Autopilot" agents. Users name it, train it with custom skills, and run it across desktop and browser. It ships through the early-adopter Frontier program (which requires a GitHub Copilot subscription) with a built-in "policy conformance system" that audits whether the agent is operating within set guidelines, presented as a response to the agent "inbox-rampage" incidents OpenClaw surfaced earlier in 2026. VP Omar Shahine said the agent "becomes more capable, better understanding you and gaining more agency" (Source: techcrunch.com; theverge.com; see Agentic AI).

Microsoft also showed Project Solara, an Android-based OS for agent-first devices, demonstrated as a desk hub and a wearable badge and built with Qualcomm and MediaTek; AccuWeather, Best Buy, CVS, and Target are slated to run pilots (Source: theverge.com).

Ahead of Build, two sources told Fortune on May 29, 2026 that Microsoft was building a "super app" to unify its fragmented Copilot tools — GitHub Copilot, Copilot chat, Copilot Cowork, and a new agentic capability code-named Autopilot — into a single interface. The effort, addressing the proliferation of separately-branded Copilot surfaces, is spearheaded by newly appointed Copilot chief Jacob Andreou under the internal slogan "Delivering one Copilot," with a launch targeted by end of summer 2026 and some elements possibly previewed at Build (Source: fortune.com). "Autopilot" as an agentic layer is positioned as Microsoft's answer to the orchestration question posed by Anthropic's Dynamic Workflows.

Andreou is the 33-year-old former product lead at Snap whom Satya Nadella promoted to executive vice president of Copilot in March 2026; he oversees more than 11,000 people. Fortune reported on June 27, 2026 that, as part of retooling the assistant, Andreou merged Microsoft's consumer and enterprise Copilot teams, eliminated extraneous versions of the assistant, and introduced consumption-based pricing such as Copilot Cowork billed by usage, while pursuing the phased Copilot "super app." About 4.5% of Microsoft 365's roughly 450 million customers pay for Copilot features, and Microsoft shares were down by double digits over the prior year (Source: fortune.com). An internal memo that became public on July 2, 2026 confirmed the consolidation: Microsoft is merging its consumer and enterprise Copilot apps into a single application with AI coding tools and agents, with the memo stating the product must "earn the right to exist" (Source: theinformation.com; seekingalpha.com).

Xbox Copilot wind-down

On May 5, 2026, Xbox CEO Asha Sharma announced that Microsoft is winding down Copilot on mobile and halting Copilot development on console, as part of a broader platform reorganization that brought CoreAI executives into Xbox (Source: theverge.com).

Security business overhaul

Microsoft's overhaul of its cybersecurity business — the world's largest by software sales — to capture AI-security demand became public on July 15, 2026. New security chief Hayete Gallot has replaced at least eight executives who reported to her predecessor and is investing in AI tools that find software flaws and respond to threats automatically (Source: theinformation.com). See AI and Cybersecurity.

A first product from the overhaul surfaced on July 17, 2026: Project Perception, an AI bug-finding security product that routes work across Anthropic, OpenAI, and Microsoft models, reported to debut as soon as July 2026 (Source: theinformation.com).

Pricing model

Microsoft has shifted its AI products from seat-based pricing toward consumption-based metering. On April 29–30, 2026, CEO Satya Nadella confirmed a structural shift to consumption-based pricing for GitHub Copilot, adding usage-based fees on top of seat fees announced the prior week, and signaled the same playbook would extend to 365 Copilot, security, and other "per-user" Microsoft AI products. CFO Amy Hood said the new consumption pricing already lifted Office 365 AI margins, and Microsoft's E7 enterprise bundle ($99/user/month) launched with a usage cap on agent features that triggers overage billing once exceeded (Sources: theinformation.com, newsletter.pragmaticengineer.com). The Pragmatic Engineer reported the same day on token-spend budget breaks across 15 tech companies (Source: newsletter.pragmaticengineer.com).

In the May 5 earnings call, Nadella told analysts that Microsoft 365 will shift "from a traditional seat model to seats plus consumption," and Hood reframed it as "a license business plus a consumption business" with Azure-style metering (Source: stratechery.com).

Effective June 1, 2026, GitHub (Microsoft) shifted GitHub Copilot to token-based billing, replacing the flat "Premium Request Units" model with AI Credits priced at roughly $0.01 each, so that charges scale with token consumption rather than a predictable per-request allotment. The change drew developer backlash on May 30–31, 2026 over unpredictable and, in many cases, higher bills, with users warning that heavy or agentic workflows could spike costs without warning (Sources: techcrunch.com, indiatoday.in). The move operationalizes the broader consumption-based pricing pivot at the developer-tool layer, passing per-token agentic-coding costs through to users, and follows the Q1 disclosure that GitHub Copilot usage was dragging Azure margins.

Developer tooling and competitive positioning

Microsoft began canceling most of its Claude Code licenses on May 14, 2026, winding down usage in its Experiences + Devices group by the end of June and pushing developers toward GitHub Copilot CLI. Sources cited a mix of cost and competitive dynamics, with one framing being "Claude Code undermining Microsoft's own tool." The move came the same week OpenAI launched Codex mobile (4M+ weekly users) (Source: theverge.com). The Verge reported on May 15, 2026 that Microsoft was requiring employees to stop using Claude Code and switch to Copilot CLI to drive in-house tooling improvements, framing the directive as dogfooding (forcing Microsoft engineers to identify Copilot CLI gaps by removing the alternative) rather than pure cost cutting (Source: spyglass.org).

In April 2026 Microsoft was reportedly evaluating a Cursor acquisition, in parallel with separate xAI and Mistral interest (Source: theinformation.com). Stratechery framed an alternative outcome as a Musk-led "SpaceXAI" via SpaceX's reported $60B Cursor option (see Cursor (Anysphere)). By April 25, 2026, Microsoft had walked away from Cursor, and SpaceX-xAI acquired the $60B option (Source: Fortune, April 25, 2026).

Azure AI and compute

Azure is one of three global hyperscalers (with AWS and Google Cloud) underwriting frontier-scale compute, and Microsoft has made multi-year, multi-billion-dollar commitments on GPUs, data centers, and power. The Three Mile Island nuclear restart (Constellation Energy, 2024) contracted to Microsoft illustrates the energy demand of Microsoft's AI buildout, a thread picked up in AI Environmental Impact. Microsoft disclosed in its annual sustainability report, released July 9, 2026, that its carbon emissions climbed 25% in 2025 amid the data-center buildout, a setback for its goal of becoming carbon negative by 2030 (Source: geekwire.com; bloomberg.com).

On April 24–25, 2026, The Information reported that Microsoft and other cloud providers have tightened access to Nvidia GPU capacity, diverting supply to internal teams and large customers such as OpenAI while smaller AI startups struggled — a sign of constrained supply rather than slowing demand (Source: theinformation.com).

Azure passed $100 billion in annual revenue for the first time in fiscal 2026, rising 41% from $75 billion, with Chief Financial Officer Amy Hood guiding first-quarter fiscal 2027 Azure growth to about 45%. On the July 29, 2026 earnings call Satya Nadella cited the OpenAI agent incident as an argument for model diversity on the platform — "you can't depend on any one model" — noting that Azure now offers 11,000 models (Source: fortune.com). Reported after-hours share moves on the results differ by outlet: Fortune put the rise at more than 8%, Reuters at 4.4%.

In-house accelerators (Maia)

Microsoft introduced the Maia line of AI accelerators in November 2023 and unveiled the second-generation Maia 200 in January 2026, built by TSMC on 3-nanometre technology with a large SRAM allocation. Microsoft has lagged Alphabet and Amazon in scaling in-house silicon; Google began recognising revenue from direct Tensor Processing Unit sales in the quarter ended June 2026.

The Information reported on August 10, 2026, citing people with direct knowledge, that Microsoft plans to unveil Maia 300 in autumn 2026, potentially as soon as September. The company has been in talks with TSMC to secure manufacturing capacity for more than 300,000 units for delivery in 2027, and ultimately aims to secure capacity for more than 1 million Maia 300 chips, though component supply and continuing capacity negotiations could constrain those plans. Microsoft is also seeking to persuade major cloud customers including Anthropic to adopt the chip (Source: reuters.com). The unit figures and the September timing are unconfirmed plans attributed to unnamed sources, not company announcements.

On July 21, 2026 Microsoft agreed to fund Mistral's European AI expansion in a multibillion-dollar deal that deepens its ties with the French model developer and targets European enterprise customers (Source: reuters.com). As part of the arrangement, Microsoft said it will spend billions of dollars on Nvidia GPUs it plans to share with Mistral (Source: wsj.com).

Hardware partnerships

On May 31, 2026, Microsoft and Nvidia co-announced the first Windows PCs purpose-built for personal AI agents, running on Nvidia's N1X Arm chip under the "RTX Spark" platform delivering roughly 1 petaflop of AI performance, with Dell, Lenovo, and HP as launch OEM partners (Source: nvidianews.nvidia.com). The announcement extends Microsoft's agent-era hardware push beyond the existing Copilot+ PCs category toward on-device agent execution; Nvidia's role is tracked on Nvidia & TSMC — AI Compute Infrastructure.

Ahead of Computex, Microsoft and Nvidia teased a consumer-PC announcement under the shared tagline "A new era of PC." Analyst M.G. Siegler argued on May 29, 2026 that the tease points to Nvidia's long-rumored general-purpose "N1"/"N1X" CPUs and a possible new Microsoft Surface device, framing it as a second attempt at the Arm-based Windows PC that failed with the original Surface RT (Source: spyglass.org). Siegler's reading is analyst speculation, not a confirmed product.

Government and enterprise deployment

On May 1, 2026, Microsoft was one of eight AI vendors signed by the U.S. Department of War to deploy frontier AI on classified networks "for lawful operational use," alongside SpaceX, OpenAI, Google, Nvidia, Reflection, AWS, and Oracle; Anthropic was excluded (Source: bloomberg.com; see DOD — Department of Defense (AI Deployer)). Microsoft Azure AI Foundry is also a vehicle for government deployment via GSA OneGov Program and USAi Platform (August 2025). On July 22, 2026, Microsoft committed $60 million to the Energy Department's Genesis Mission — $40 million in Azure compute credits and $20 million in engineering — with initial projects at Pacific Northwest National Laboratory, Lawrence Livermore National Laboratory, Johns Hopkins APL, and Idaho National Laboratory (Source: blogs.microsoft.com). See Department of Energy (DOE).

On July 2, 2026, Microsoft announced Microsoft Frontier Company, a new operating business backed by $2.5 billion and 6,000 industry and engineering experts — forward-deployed engineers, consultants, and salespeople embedded with clients — to run enterprise AI deployments, led by Rodrigo Kede Lima, with early partners including the London Stock Exchange Group, Unilever, Land O'Lakes, and Accenture (Source: techcrunch.com; cnbc.com). The move followed Amazon Web Services' June 30, 2026 launch of a $1 billion forward-deployed-engineering organization and earlier OpenAI and Anthropic enterprise-services joint ventures valued at $4 billion and $1.5 billion respectively (Source: techcrunch.com).

Microsoft and Accenture confirmed on April 27, 2026 that Copilot 365 would be rolled out to approximately 743,000 Accenture employees, the largest enterprise Copilot deal to date, giving Microsoft a reference customer for global rollouts in IT services (Source: reuters.com). England's National Health Service rolled out Microsoft 365 Copilot to more than 500,000 staff (Source: fortune.com).

Data and platform competition

On May 22, 2026, Microsoft moved to block partners from connecting their data-management tools to one of its popular products, opening a new front in competition over AI-agent access to data. The Information framed the move as Microsoft defending its business-software position in the agent era, treating control over the data-integration layer as a strategic chokepoint as AI agents increasingly act on enterprise data. The action came alongside Microsoft's parallel moves to consolidate around its own developer tooling and consumption-based AI pricing (Source: theinformation.com; see Agentic AI).

Policy positions

Brad Smith has articulated Microsoft's public AI policy stance across speeches, congressional testimony, and the book Tools and Weapons:

• "Time for AI Government": a Smith framing arguing that governments must build AI capacity, not just regulate it, also deployed in favor of government cloud procurement.
• Federal legislation over state patchwork: Microsoft supports risk-tiered federal AI legislation rather than a patchwork of state laws, aligning it with preemption arguments in Colorado and Texas litigation.
• Sovereign AI: Microsoft has marketed "sovereign cloud" products to the EU and other national governments, positioning Azure as a substrate for sovereign AI strategies.
• Export controls: supportive in principle; Microsoft has invested heavily in compliance with the BIS AI Diffusion Rule regime.
• Content provenance: founding member of C2PA; signatory to the 2024 AI Elections Accord.

In a Wall Street Journal interview published June 21, 2026, CEO Satya Nadella argued that a small group of companies should not be allowed to capture the value of AI while justifying unlimited data-center expansion with warnings about job losses and safety risks, saying "You can't say, hey, all white-collar jobs are gone and this could even be a weapon and we will use all the power to build data centers." Without naming OpenAI, Anthropic, or Google, Nadella positioned Microsoft against a future dictated by frontier model-builders, pointing to the company's rollout of low-cost models and its Copilot Cowork agent — which lets users pick cheaper models for long-running tasks — as a push toward lower prices and greater user control (Source: wsj.com). See AI Power Concentration.

In a Fortune interview published July 9, 2026, Brad Smith said U.S. AI policy amounts to "regulation without transparent or complete rules," criticizing the Commerce Department's June invocation of export-control law to pull Anthropic's Fable 5 and Mythos 5 models and its pressure on OpenAI to delay GPT-5.6's public rollout (Source: fortune.com). See AI Pre-Release Vetting.

In a July 12, 2026 post, Nadella warned enterprises about the data cost of using external AI: companies "pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal," urging them to keep data, evaluations, and adapted model weights inside a hard "trust boundary" (Source: techcrunch.com).

Safety and responsible AI

Microsoft maintains a Responsible AI Standard and a Responsible AI Office; it publishes transparency reports, uses red-teaming on Azure OpenAI deployments, and has contributed to the NIST AI RMF development process. Kevin Scott (CTO) is a public-facing voice on long-horizon safety questions. Sébastien Bubeck (see Sébastien Bubeck) led the Microsoft Research team behind Sparks of AGI before moving to OpenAI in 2024.

On May 5, 2026, Microsoft joined Google and xAI in agreeing to give the U.S. Center for AI Standards and Innovation (CAISI) early access to evaluate frontier models for national-security risks before public release, joining OpenAI and Anthropic (Source: bloomberg.com; see AI Pre-Release Vetting, NIST CAISI (Center for AI Standards and Innovation)).

In June 2025 Microsoft added a "safety" category to the Azure Foundry model leaderboard, using ToxiGen (hate speech) and CAIS WMDP (weapons misuse) benchmarks, becoming the first major cloud provider to rank models by safety as a purchase-influencing metric (Microsoft AI Safety Leaderboard (June 2025)). Microsoft is also assessed alongside Google DeepMind as a competitor in FLI AI Safety Index Winter 2025.

Microsoft's installed base has also been a target of AI-assisted security research: the NSA has been testing Anthropic's Mythos specifically to find vulnerabilities in Microsoft products and other widely used software (Source: bloomberg.com).

On June 4, 2026, Nadella publicly rebuked an internal memo by another Microsoft executive that advocated making users "addicted" to AI agents such as Scout, drawing a line against engagement-maximization framing for the company's flagship agent (Source: theinformation.com; nypost.com; see AI Companions).

Workforce and research

Microsoft opened the first voluntary retirement program in its 51-year history on April 25, 2026, offering long-tenured US employees a financial payout plus extended healthcare, with full terms scheduled for May 7 (Source: geekwire.us2.list-manage.com); this set against the broader 2026 hyperscaler labor pattern, including Meta's 8000 layoffs. The program resulted in a $900M charge in the quarter (Source: geekwire.us2.list-manage.com). Reporting on July 1, 2026 said Microsoft was preparing a further round of cuts the following week spanning its Xbox, sales, and consulting units and affecting less than 2.5% of its roughly 220,000-person workforce; the company plans to spend more than $100 billion on AI infrastructure in the fiscal year while holding down operating costs, and its shares had fallen about 19% over the prior month amid questions about the profitability of those investments. The wider technology sector recorded more than 123,000 job cuts in 2026, largely attributed to AI restructuring (Source: geekwire.com; see AI Labor Disruption).

The cuts were announced on July 6, 2026: 4,800 jobs, just over 2% of the workforce, including about 1,600 immediate Xbox positions and roughly 3,200 total Xbox cuts (about 20% of that division) planned for the fiscal year. Xbox CEO Asha Sharma called it the biggest restructuring in Xbox history, citing margins 3–10× lower than comparable Microsoft businesses, and four game studios are to be spun off. The announcement followed the July 2 launch of the $2.5 billion Microsoft Frontier Company and a 30% stock slide that erased about $1.2 trillion in market value over nine months (Source: geekwire.com). The cuts were communicated in a memo from chief people officer Amy Coleman and concentrated in sales and Xbox, as the company shifts focus toward AI products (Source: bbc.com). In a July 6 interview, Sharma said of the Xbox retrenchment that "we simply spread ourselves too thin," confirming the cuts affect about 20% of Xbox staff and that four studios will be divested (Source: fortune.com).

LinkedIn, a Microsoft subsidiary, appointed Daniel Shapero CEO on April 25, 2026, replacing Ryan Roslansky after six years in LinkedIn's first leadership change in that span; Roslansky remains a Microsoft EVP (Source: geekwire.us2.list-manage.com).

Microsoft published its 2026 Work Trend Index on May 5, 2026, surveying 20,000 AI users in 10 countries. Findings included: 65% fear falling behind, but only 13% are rewarded for experimenting with AI; 49% of analyzed Copilot interactions involved cognitive work; agents on Microsoft 365 grew 15× year-over-year (18× in large enterprises); and 16% of users qualify as "Frontier Professionals" who routinely run multi-step agent workflows, with 80% of those saying they produce work they could not have a year earlier (Source: geekwire.com; see AI and Productivity, AI Labor Disruption).

Key people

• Satya Nadella (CEO): architect of the Azure-centric pivot and the OpenAI partnership.
• Brad Smith (President and Vice Chair): principal public policy voice; author of Tools and Weapons.
• Mustafa Suleyman (Microsoft AI CEO): co-founder of DeepMind and Inflection; hired March 2024.
• Kevin Scott (CTO): long-time Nadella deputy on AI strategy.
• Jacob Andreou (EVP of Copilot): former Snap product lead; promoted by Nadella in March 2026 to lead the Copilot product family.
• Sarah Bird: Chief Product Officer, Responsible AI.

Litigation and regulation

• NYT v. OpenAI/Microsoft — co-defendant in the NYT's copyright suit over training data; see AI Copyright Litigation — Analysis.
• FTC 6(b) AI Partnerships Study — subject of the FTC's January 2025 staff report on Big Tech AI investments.
• Broader AI copyright litigation — named in multiple suits as OpenAI's cloud partner and distributor.
• UK CMA antitrust probe (May 14, 2026): the UK Competition and Markets Authority launched an antitrust probe into Microsoft's business-software ecosystem and its bundling of Windows, Word, and Teams (Source: reuters.com).
• UK Azure+Copilot bundling lawsuit (April 22, 2026): Microsoft was hit with a £2.1B ($2.8B) UK competition-court lawsuit over Azure+Copilot bundling (Source: ft.com).
• EU DMA gatekeeper designation for cloud (June 25, 2026): the European Commission said the cloud-computing units of Microsoft and Amazon should fall under the "gatekeeper" obligations of the Digital Markets Act, extending the framework toward core AI-infrastructure providers (Source: reuters.com).

Relationships

• depends-on: OpenAI — commercial and technical dependency
• related: Anthropic (competitor via Azure Foundry multi-model strategy), Google DeepMind, Meta AI, Nvidia & TSMC — AI Compute Infrastructure
• instance-of: AI Race Dynamics, Semiconductor Supply Chain (principal demand node)
• supports: AI and National Security, Export Controls (AI) (Smith position)
• contradicts: strong state-level AI regulation (Smith has publicly preferred federal preemption)