Microsoft CEO Satya Nadella has spent much of the past month making an unusually pointed public case against enterprises betting their entire AI strategy on a single provider — including, implicitly, Microsoft’s own longtime close partner OpenAI. In a blog post and a series of public comments, Nadella argued that every AI model is “substitutable,” and warned that companies relying wholly on proprietary AI labs for their AI needs may ultimately not survive as competitive businesses. He described a specific risk: that companies using models like OpenAI’s or Anthropic’s are effectively “paying twice” — once for the tokens that do the actual work, and again by handing those labs the proprietary business knowledge needed to make the models useful, knowledge those same labs could later use to build competing products.
Nadella’s alternative pitch is Microsoft’s own Azure AI Foundry platform, which he says now offers customers the broadest model catalog of any cloud provider — more than 11,000 models available, spanning OpenAI’s latest releases, Anthropic’s Claude family, Mistral, xAI’s Grok, and Microsoft’s own in-house MAI model family. The underlying strategic logic is that Microsoft profits from AI infrastructure and platform fees regardless of which specific model a customer ultimately chooses to run on Azure, giving Microsoft less incentive than a pure model developer to lock customers into any one AI provider — including OpenAI, despite Microsoft’s own multibillion-dollar investment in and close partnership with that company.
A Genuinely Strong Quarter Behind the Warning
Nadella’s public positioning comes from a position of real financial strength. Microsoft’s most recent quarter brought in $90 billion in revenue and $35.8 billion in net income, and for the full fiscal year ending June 30, 2026, the company reported $331.8 billion in revenue and $133.7 billion in net income — among the largest annual profit figures any company has ever reported. Copilot, Microsoft’s AI assistant product embedded across its software lineup, has crossed 30 million paid seats, a figure Nadella has highlighted directly as evidence that Microsoft’s own AI products are scaling successfully in parallel with its model-agnostic platform pitch to enterprise customers.
The Real Tension Behind the Multi-Model Pitch
Nadella’s public case for model diversity is easier to understand against the backdrop of Microsoft’s own increasingly complicated relationship with OpenAI. The two companies’ original exclusive partnership ended following a major restructuring and recapitalization deal that concluded in late 2025, after OpenAI’s growing computing needs outpaced what Microsoft alone could supply and OpenAI began pursuing infrastructure partnerships with Microsoft’s own rivals, including a separate, roughly $500 billion data-center project with Oracle and SoftBank. Under the restructured agreement, Microsoft holds a 27% ownership stake in OpenAI — down from an estimated 32.5% before the deal, reflecting new investors and OpenAI’s own ballooning valuation — while OpenAI is now free to strike infrastructure deals with other cloud providers, and Microsoft is free to build and promote models from other AI labs. Nadella’s public “every model is substitutable” messaging is, in effect, Microsoft formally adjusting its own public posture to match a partnership that has already become considerably less exclusive in practice. Microsoft has, in parallel, continued investing heavily in its own MAI model family specifically so it is not left solely dependent on OpenAI’s roadmap regardless of how that underlying relationship continues to evolve. Azure’s own cloud revenue growth, which Microsoft has continued reporting as accelerating even as the OpenAI relationship has grown more complicated, is the clearest evidence that Nadella’s bet on model diversity rather than exclusive dependency has not come at the cost of Microsoft’s own core cloud business. Whatever tension exists between the two companies behind closed doors, both sides have continued to publicly describe the relationship as an ongoing partnership rather than a clean break — a distinction worth noting given how often the two companies’ arrangement gets described in shorthand as a ‘breakup’ by outside commentary.
What This Means for Philippine Founders
Nadella’s “substitutable model” argument is directly relevant to any Philippine startup currently building its entire product around a single AI provider’s API — his warning that a single-vendor dependency creates real business risk applies just as much to a startup relying entirely on one model as it does to the large enterprises he’s speaking to directly. Building at least some model-switching flexibility into a product’s architecture from the start, rather than hard-coding a single provider’s API throughout, is a genuinely low-cost form of insurance against the exact risk Nadella is describing — a risk that applies whether or not a founder ever plans to build at Microsoft’s own scale. His comment about AI labs potentially using customer data and proprietary knowledge to build competing products is also worth taking seriously when negotiating any enterprise AI contract: understanding exactly what a vendor can and cannot do with a company’s own data and usage patterns is a genuine, practical due-diligence item, not just a theoretical concern.
Share this article