- Tech Rundown
- Posts
- 💣 OpenAI Is Trying to Blow Up Its Microsoft Deal
💣 OpenAI Is Trying to Blow Up Its Microsoft Deal
Behind closed doors, OpenAI is building the infrastructure to cut Microsoft out.

Microsoft invested $1 billion into OpenAI back in 2019 when ChatGPT was just a twinkle in Sam Altman's eye. Over time they’ve plowed another $12 billion into the company, with increasing conviction. In exchange, they got 49% ownership, exclusive distribution rights, 20% of revenues up to $92 billion, and crucially, access to all of OpenAI's IP. It seemed like a reasonable deal at the time. Microsoft got the inside track on the AI revolution, and OpenAI got the capital and cloud infrastructure to build their models at scale.

That was then. Now OpenAI is reportedly considering the "nuclear option" of accusing Microsoft of anticompetitive behavior. They're trying to carve out their $3 billion Windsurf acquisition so Microsoft can't touch the IP. They're building their own $500 billion data center project. And oh, they just signed a $200 million contract directly with the Department of Defense.

What happens when your biggest investor becomes your biggest competitor? You get creative with the divorce proceedings.
The Windsurf situation perfectly illustrates the problem. OpenAI wants to acquire this coding startup, but under their current deal, Microsoft gets access to all OpenAI IP. The issue? Microsoft owns GitHub Copilot, which directly competes and has been lagging behind the coding agent startups. So OpenAI is essentially saying: "We'd like to make this acquisition, but we'd prefer our largest shareholder not get the technology that would help them compete with us." Microsoft's response appears to be: "That's not how partnerships work."
It's worth stepping back and understanding what OpenAI is actually trying to build. Most people think of them as an AI model company. But having the best model isn't really their strategy. Their strategy is becoming the AI superapp for everyone. They're building personalization through memory features, context integration that pulls data from places like HubSpot, meeting transcription tools, and workflow interfaces like Deep Research and Canvas. The goal isn't just to have the smartest AI; it's to have the most useful AI for actual work.

This puts them in direct competition with Microsoft across multiple fronts. Microsoft has their own AI assistant (Copilot), their own productivity suite (Office), their own coding tools (GitHub), and their own enterprise relationships. The partnership that once seemed synergistic now looks very bloody.
The Department of Defense contract is particularly telling. Microsoft has long been the DoD's cloud provider, recently getting Azure OpenAI Service approved for classified use. Now OpenAI is going around them entirely. The $200 million isn't huge money for either company, but it signals something important: OpenAI doesn't want to be dependent on Microsoft for any category of customer.
Microsoft seems to understand what's happening. CEO Satya Nadella has started talking about how AI models will become "commoditized," with the real value shifting to AI-enabled applications. Translation: if OpenAI's models become just another input, Microsoft's competitive advantages in productivity software and enterprise relationships become more important. They've also started diversifying away from OpenAI models, presumably hedging against exactly this scenario.

The restructuring negotiations reveal just how messy this is getting. OpenAI needs Microsoft's approval to convert from a non-profit to a for-profit structure by the end of the year, or they risk losing billions in funding from other investors like SoftBank. Microsoft knows this, which gives them significant leverage. But OpenAI also knows that Microsoft needs their technology through 2030 under existing contracts, which gives them leverage too.

The revenue sharing arrangement suddenly looks less appealing from OpenAI's perspective. Twenty percent of revenues up to $92 billion sounded fine when they were burning money on research. Now that they're forecasting $100 billion in revenue by 2027, that's potentially $18.4 billion they're handing over to Microsoft. For context, that would represent about 7% of Microsoft's total 2024 revenue of $245 billion. Meaningful for Microsoft, but not existential.

The compute infrastructure question is equally complex. Right now, OpenAI wants to drive down the cost of AI to capture market share. They don't necessarily want their margins to erode, but they want control over pricing rather than being subject to Microsoft's Azure markup. Project Stargate, their plan to build $500 billion worth of data centers, isn't just about capacity. It's about not giving Microsoft $0.20 for every $1 they make.
What's remarkable is how aggressive OpenAI is being about this. They're simultaneously trying to maintain their partnership with Microsoft while systematically reducing dependencies across infrastructure, distribution, and IP access.
Microsoft could theoretically walk away from the restructuring negotiations and rely on their existing contracts through 2030. But that would essentially guarantee that OpenAI becomes a direct competitor rather than a strategic partner. The question, as one person close to the discussions put it, is "what does Microsoft get in return for giving up the right to that revenue?"
The answer seems to be based on how aggressive Microsoft are willing to be. Supposedly senior leaders and/or Satya Nadella himself have threatened that they will prevent OpenAI being able to convert to a for-profit which means OpenAI would lose tens of billions of dollars of funding from SoftBank and would find it much more challenging to go public. Seemingly the question is how favourable a set of terms can Microsoft demand to give up some its clauses (IP, revenue share, exclusive distribution, for-profit org).

The irony is that Microsoft's investment was so successful that it created a competitor they can't control. OpenAI now has the scale, technology, funding and ambition to go head-to-head with Big Tech companies. Including the one that made it all possible.
I guess that's what happens when you fund a revolution. Sometimes the revolutionaries decide they don't need you anymore.