Microsoft Feared OpenAI Dependence, Court Testimony Reveals

CNBC reported Wednesday that testimony in the ongoing Musk v. Altman trial has exposed deep anxiety inside Microsoft over its reliance on OpenAI — fears that surfaced well before generative AI became a global phenomenon.

Nadella’s IBM Warning

Microsoft CEO Satya Nadella voiced concern about Microsoft OpenAI dependence as early as April 2022, seven months before ChatGPT’s public debut. In an internal email surfaced through discovery, he drew a stark historical comparison. He said he did not want Microsoft to play the role IBM once played — only to watch a more agile software partner absorb the bulk of the industry’s market value. The IBM-Microsoft dynamic of the 1980s, where IBM distributed Microsoft’s operating system before losing out over time, served as his cautionary frame. Testifying Monday in Oakland, California, Nadella said it became critical for Microsoft to maintain control “at every layer of the stack,” not merely as a cloud infrastructure provider.

A Partnership That Kept Evolving

Microsoft’s first investment in OpenAI, a $1 billion commitment, landed roughly three years before that internal warning. The companies have since renegotiated their agreement multiple times. The most recent revision, completed in April, capped the revenue-share payments flowing to Microsoft and permitted OpenAI to supply its products through competing cloud providers, including Amazon and Google. A Microsoft corporate development executive testified this week that cumulative spending on OpenAI — covering investment, infrastructure, and hosting — will surpass $100 billion by June 2026.

How the Early Fears Proved Justified

The concerns Nadella recorded in private have since found confirmation in the market. OpenAI now carries an $850 billion valuation and has struck cloud agreements with rivals Oracle, Google, and Amazon. Meanwhile, roughly 45% of Microsoft’s commercial remaining performance obligations at the end of 2025 were tied to OpenAI’s continued use of Azure infrastructure. A July 2022 email from Nadella, also produced in discovery, captured his lingering ambivalence. He noted the appeal of being an investor rather than shouldering the full operational risk of building frontier AI models himself.

The Infrastructure Legacy

Microsoft’s chief technology officer Kevin Scott testified that the first supercomputer the two companies built together required about six months of work and drew on 10,000 graphics processing units filling a substantial portion of a data center. Scott said he remains proud of what that early infrastructure investment allowed OpenAI to accomplish in research and commercial products. The trial continues to offer a rare window into how two of the most consequential companies in modern AI navigated an unusually high-stakes partnership.

Read Next: OpenAI Secures Fresh Funding as Valuation Climbs Toward $300 Billion