Microsoft’s 2018 Gamble on OpenAI: Skepticism, Strategy, and the Rise of an AI Power Couple

In mid-2018 Microsoft executives exchanged emails that revealed deep doubts about OpenAI’s technical roadmap and Sam Altman’s ability to deliver AGI. Nonetheless, senior leaders, including CEO Satya Nadella, approved a partnership to keep the startup from moving to Amazon Web Services, a move later formalized with a $1 billion investment after OpenAI created a for-profit arm. (Sources 1, 6)

Wired Magazine

At the time OpenAI was a nonprofit focused on reinforcement-learning projects such as Dota 2 bots and robotic observation, and was transitioning to a “capped-profit” model that unsettled investors. The internal debate involved Satya Nadella (CEO), Brett Tanzer (business development), Jason Zander (strategy), Kevin Scott (CTO), Harry Shum (AI research), Eric Horvitz (research), and Jason Graefe, alongside OpenAI’s Sam Altman and external influencer Elon Musk. (Sources 1, 2, 3, 4, 6)

Microsoft might be all-in on OpenAI now, but back in 2018 thought it was just 'motivated by a need to show how Al can crush humans'

Azure’s list price for GPU compute was $1.15 per hour, but OpenAI paid $0.24 per hour under a 2016 discount. The 2016 contract exchanged $10 million for $60 million of services, projecting a $15 million loss over three years. Later cost analyses showed the proposed $300 million compute request would generate roughly a $150 million negative gross margin for Microsoft. The partnership grew to a $1 billion investment in 2019, a $10 billion infusion in 2023, and a total of over $13 billion, securing a 27 percent stake. (Sources 2, 3, 4, 7)

Microsoft might be all-in on OpenAI now, but back in 2018 thought it was just 'motivated by a need to show how Al can crush humans'

Internal memos framed the deal as a defensive move: executives repeatedly cited the risk of “Amazon swooping in” and emphasized Azure’s need for a differentiating AI offering. Nadella’s January 10 2018 email noted Elon Musk’s belief that OpenAI was “on the verge of some big AGI breakthroughs,” while also acknowledging that Microsoft was merely “renting/discounting HW” without a clear feedback loop. (Sources 3, 6)

Musk v. Altman Evidence Shows What Microsoft Executives Thought of OpenAI

Senior technologists voiced strong skepticism. Kevin Scott called an imminent AGI breakthrough “highly skeptical” and questioned any branding benefit. Jason Zander warned the deal would cost Microsoft roughly $150 million, labeling it a “non-starter.” Harry Shum and Eric Horvitz warned that OpenAI treated Azure as a “bucket of undifferentiated GPUs” and that the partnership could backfire if the startup migrated to AWS. (Sources 4, 3)

Microsoft might be all-in on OpenAI now, but back in 2018 thought it was just 'motivated by a need to show how Al can crush humans'

The email excerpts stem from court filings in the Musk v. Altman lawsuit, and some sources lack reliability scores, limiting verification of exact figures. While internal documents project specific losses, Microsoft has not publicly confirmed the discount terms or the projected gross-margin impact. The timing of the $1 billion investment relative to the internal debates also varies across reports. (Sources 1, 2, 6)

Musk v. Altman Evidence Shows What Microsoft Executives Thought of OpenAI

• “The 2016 deal, where they agreed to pay $10M for $60M in Azure services was projected at a $15M loss over three years, given assumed usage profile,” — Brett Tanzer, Business Development

Microsoft might be all-in on OpenAI now, but back in 2018 thought it was just 'motivated by a need to show how Al can crush humans'

• “I reviewed the GM with the team and this deal would cost us ~-$150M over the current contract which frankly makes it a non-starter,” — Jason Zander, Strategy

Microsoft might be all-in on OpenAI now, but back in 2018 thought it was just 'motivated by a need to show how Al can crush humans'

• “On the incremental investment, I'm not sure what Microsoft is going to get out of it.” — Kevin Scott, CTO

Microsoft might be all-in on OpenAI now, but back in 2018 thought it was just 'motivated by a need to show how Al can crush humans'

• “less about what they've built and more about what we prevent Amazon from accessing.” — Microsoft executive (anonymous)

Musk v. Altman Evidence Shows What Microsoft Executives Thought of OpenAI

The 2018 decision anchored Microsoft’s AI strategy, granting exclusive Azure hosting of GPT-3 and later models, fueling products such as GitHub Copilot and Bing AI. By preventing Amazon from securing OpenAI, Microsoft turned early skepticism into a cornerstone of its cloud revenue, while raising broader questions about whether strategic partnerships are driven by technology merit or competitive blockage. (Sources 6, 7)

Musk v. Altman Evidence Shows What Microsoft Executives Thought of OpenAI