Nvidia Cuts OpenAI Ohio Deal to $120B as Anthropic Revenue Soars

What happened

Nvidia has scaled back a financial commitment tied to OpenAI's planned Ohio data centre, cutting its guarantee from $250 billion to just under $120 billion, following pushback from investors concerned about concentration risk in its AI infrastructure bets. The revision comes as Anthropic reports a sharp jump in quarterly revenue, from $4.7 billion to $11.5 billion, a figure that has complicated the broader narrative around an AI spending bubble.

The two developments, reported together by The Decoder, sit on opposite sides of the same question: whether the scale of capital flowing into AI infrastructure and model providers is sustainable. Nvidia's decision to reduce its exposure to a single, large-scale commitment suggests shareholders are demanding more caution around mega-deals, even as one of OpenAI's principal rivals posts growth that argues against a demand slowdown.

Why it matters

For leaders tracking AI adoption and infrastructure investment, this is a signal that the market is starting to differentiate between AI enthusiasm and AI financing discipline. Nvidia's willingness to renegotiate a headline-grabbing guarantee shows that even the most dominant supplier in the AI stack is not immune to investor scrutiny over how concentrated its bets on any one customer or project should be.

At the same time, Anthropic's revenue trajectory indicates that demand for frontier AI capability, particularly among enterprise and developer customers, remains strong enough to support aggressive growth without needing speculative infrastructure commitments to justify it. Together, the two data points suggest the AI market is entering a phase where capital discipline and genuine commercial traction are becoming more important than headline valuations or infrastructure announcements.

By the numbers

• $250 billion — Nvidia's original guarantee tied to OpenAI's planned Ohio data centre.
• $120 billion — Nvidia's revised guarantee, a cut of roughly half.
• $4.7 billion — Anthropic's revenue in the prior quarter.
• $11.5 billion — Anthropic's revenue in the most recent quarter, more than doubling.

The Renascence take

The headline framing treats this as a binary: bubble or no bubble. The more useful lens is behavioural — what these two moves reveal about how investors and operators are recalibrating trust in AI commitments after a period of near-unconditional enthusiasm.

Most coverage will read Nvidia's retreat as bearish and Anthropic's growth as bullish, but the real story is about risk architecture, not sentiment. Investors aren't losing faith in AI; they're losing patience with single-counterparty exposure, which is a governance correction, not a demand correction. Anthropic's number, meanwhile, is a reminder that usage-based revenue is the only metric that should anchor infrastructure decisions — not headline capacity guarantees announced years ahead of actual utilisation. Any organisation building its own AI or digital transformation roadmap should take the same lesson: commit capital in step with proven demand, not in step with the size of the announcement you want to make.