Altman, Huang and the last-minute negotiations that sealed the $100 billion OpenAI-Nvidia deal (www.cnbc.com)

Through late-night calls and last‑minute contract tweaks between CEOs Jensen Huang and Sam Altman, Nvidia and OpenAI struck a $100 billion partnership that will see Nvidia invest in OpenAI and supply the GPUs to jointly build a global AI supercomputing footprint under the codename “Stargate.” The deal is structured as ten $10 billion tranches (the first locked at a $500 billion valuation), with Nvidia committing processors and capital as OpenAI ramps toward a planned 10‑gigawatt buildout; the first site is expected online in H2 next year. Negotiations were unusually direct — largely one‑on‑one across London, Washington and California — and the announcement followed private briefings during President Trump’s U.K. visit. For the AI community, the pact accelerates the concentration of compute, capital and influence in two dominant players and signals a shift toward vertically integrated, AI‑optimized data centers. Technically, the deal secures massive sustained access to cutting‑edge accelerators (critical for training frontier models) but also raises vendor‑concentration and supply‑chain questions: OpenAI will lease Nvidia chips while seeking additional financing (debt, other partners) and continuing ties with cloud providers (Microsoft was informed at the last minute; Oracle and SoftBank are also involved). OpenAI has floated operating its own commercial cloud within a year or two, which would reshape where large models are trained and run and intensify competition over power, permitting and global data‑center siting.