Microsoft executives spent the back half of 2017 and early 2018 worried that OpenAI would walk to Amazon if Redmond didn't write a big enough check, according to internal emails unsealed this week in the Musk v. Altman trial. Sam Altman's opening ask to Satya Nadella was roughly $300 million at Azure list prices, pitched days after OpenAI showed a bot beating a Dota 2 professional in the summer of 2017. Microsoft CTO Kevin Scott warned Nadella in January 2018 about the "PR downside" of letting OpenAI "storm off to Amazon in a huff and shit-talk us and Azure on the way out." A month later, Microsoft announced its $1 billion investment.
The exchanges, surfaced in court documents on May 8, 2026, are the clearest record yet of how reluctantly the partnership came together. Altman's $300 million Azure proposal landed shortly after the Dota 2 demo, and OpenAI was already burning through the Azure credits Microsoft had given it. Azure chief Jason Zander pushed back in an August 2017 email to Nadella, writing that "for those numbers to make sense we'd have to be generating significant incremental revenue directly due to the deal ($500 million+) that couldn't be gained in a more efficient way."
Altman tried a second route. According to Brett Tanzer, now VP of Azure solutions and ecosystem, OpenAI came back with a proposal to "create a partnership with Xbox around gaming, and an open offer to share their technology and IP in exchange for expanded sponsorship for their Dota research." The Xbox team was open to "exploring collaboration opportunities" but wouldn't carry the research bill alone.
Key facts
- 01Sam Altman pitched Satya Nadella on roughly $300M of Azure compute at list prices in summer 2017 to fund OpenAI's next research phase.
- 02Azure chief Jason Zander told Nadella in August 2017 the deal needed $500M+ in incremental revenue to pencil out.
- 03Microsoft CTO Kevin Scott warned Nadella in January 2018 about OpenAI defecting to Amazon and 'shit-talking' Azure on the way out.
- 04Microsoft announced its $1B OpenAI investment a month after Scott's email, kicking off a nearly seven-year partnership.
- 05OpenAI told staff last month its Microsoft deal limited access to AWS and Amazon Bedrock, where many enterprises run.
That left the decision sitting with Scott. In his January 2018 note to Nadella, Scott said he wasn't sure what Microsoft was "going to get out of" the deal and didn't see how the Dota work would translate into Microsoft revenue. The competitive risk is what tipped him. "They are building credibility in the AI community very fast, recruiting well, and are going to be an influential voice," Scott wrote. "All things equal, I'd love to have them be a Microsoft and Azure net promoter. Not sure that alone is worth what they're asking."
“The PR downside of us not funding them, and having them storm off to Amazon in a huff and shit-talk us and Azure on the way out.”— Jaeden Schafer
Scott also conceded, in a later email to Nadella and Bill Gates, that he had been "highly dismissive" of the AI work coming out of OpenAI and Google DeepMind when the two were competing on game-playing stunts. He warmed up once OpenAI shifted toward natural language processing and started worrying Microsoft would fall behind Google. The $1 billion investment followed shortly after.
That investment, announced in 2019, became the foundation of Microsoft's AI strategy and gave Azure exclusive cloud rights to OpenAI's frontier models for years. The framing in Scott's emails — that the worst outcome was OpenAI becoming an AWS reference customer rather than an Azure one — turned out to be prescient about the stakes, if not the timing. The exclusivity held for nearly seven years.
It also held until it didn't. OpenAI has spent 2026 renegotiating its Microsoft deal to route its API, Codex, and other products through AWS, and last month told employees that the original arrangement had "also limited our ability to meet enterprises where they are — for many that's [Amazon] Bedrock." That is, almost word-for-word, the kind of public pivot Scott flagged in 2018.
The trial context matters. The emails were filed as exhibits in Elon Musk's lawsuit against Altman and OpenAI, which we covered last week alongside Mira Murati's deposition on the 2023 board ouster. Musk's broader claim is that OpenAI's transition from nonprofit research lab to commercial enterprise was engineered through deals exactly like the Microsoft one, and the discovery process has produced a steady drip of unflattering internal communications from both sides.
None of the 2017–2018 emails reveal wrongdoing on Microsoft's part — they show normal corporate due diligence on a large compute commitment. What they do undercut is the tidy public narrative that Nadella saw the AI wave early and bet boldly. The internal version is closer to a reluctant hedge: Microsoft wasn't sure OpenAI was worth $300 million in Azure credits, and ultimately wrote a $1 billion check primarily to keep the lab away from Amazon and Google.
For Microsoft, the irony of the unsealed emails is that the scenario Scott most wanted to avoid is now the one playing out — just on a delayed timeline and with Microsoft still holding equity. OpenAI on AWS Bedrock is exactly the "net promoter for Amazon" outcome Scott described. The $1 billion bought roughly seven years of Azure exclusivity and a stake in the most valuable AI company in the world, which is a defensible trade. But the emails make clear that Microsoft's leverage in this partnership has been eroding from the start, and the renegotiated 2026 terms are the bill coming due.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
