Newshot The "Brake" on 黄仁勋's Most Aggressive AI Bet: Interim CEO 克雷斯 Once Pushed to Cut a $250 Billion OpenAI Wager
IT Home, November 13 — The tech media outlet The Information published a post yesterday (November 12), citing people familiar with the matter, that 英伟达 interim CEO 科莱特 · 克雷斯 (Colette Kress), while serving as the company's finance chief, played the role of a financial "brake," repeatedly raising reservations about high-risk deals 黄仁勋 was pushing hard.
The report says 克雷斯 repeatedly resisted deals 黄仁勋 championed, and over the past several months also focused on finding ways to keep 英伟达's financial risk from becoming overly concentrated in a handful of companies.
The most representative case involved a proposed arrangement for OpenAI that would provide a $250 billion (IT Home note: roughly 1.67 trillion yuan at current exchange rates) credit backstop, aimed at helping OpenAI fill 8 gigawatts of data center capacity with 英伟达's latest chips.
克雷斯 objected to the proposal — not out of outright opposition to financing support for OpenAI, but mainly out of concern that the guarantee amount was too large and would trigger a reaction from investors — and pushed to scale the commitment down to about $100 billion (roughly 668.507 billion yuan at current exchange rates).
The market sees 克雷斯 as a more cautious hand at the helm than 黄仁勋. Against a backdrop of delayed data center construction, canceled cloud computing orders and doubts about returns on AI infrastructure investment, she is seen as more inclined to limit the scale of guarantees, investments and customer financing, avoiding the buildup of excessive risk in 英伟达's "circular financing" model, in which it is both chip supplier and funder of its customers.
The outlet interprets this as follows: now that 克雷斯 has become 英伟达's interim CEO, will this long-time finance chief — the "brake" — slow the most aggressive AI capital wager of the 黄仁勋 era?
Related Articles
- 英伟达 open-sources its most powerful transcription AI model, Nemotron 3.5 Transcribe: supports 40 languages and transcribes accurately even when 8 people talk over each other
- Report: 英伟达 in talks to acquire / invest further in Reflection AI, doubling down on open-source models
- AI inference MLPerf 6.1 released: AMD sets a record with 512 MI355X chips, 英伟达 Vera Rubin makes its first appearance
- 英伟达 showcases DSX MaxLPS technology, boosting AI factory token throughput per megawatt by up to 40%
- The largest AI supply chain leak of 2026: 195TB of data scraped in 40 minutes from about 2,500 companies worldwide, including 英伟达
- Performance edge above expectations: AMD says Venice processors are about 20% faster than 英伟达 Vera