In early 2026, Nvidia and OpenAI restructured their relationship with Nvidia committing a $30 billion equity investment into OpenAI and binding hardware commitments for 3 gigawatts of inference capacity and 2 gigawatts of training compute on Nvidia’s Vera Rubin systems, creating a durable, milestone-free arrangement.

Previously, through late 2025 into early 2026, there was no definitive agreement and Nvidia’s data center guidance did not include the $100 billion commitment, as negotiations had effectively halted.

OpenAI’s finances remain a risk: despite surpassing $20 billion in annualized revenue by late 2025, the company projects a cash burn near $14 billion in 2026 and does not expect cash flow positivity before 2030, while diversifying chip sourcing with AWS Trainium and partnerships with Cerebras and Broadcom.

Investors should reassess the valuation to reflect the confirmed $30 billion equity stake and 5 gigawatts of Vera Rubin compute, rather than the earlier milestone-based framework.

For semis, energy, cooling, and infra investors, the long-term hardware demand from the OpenAI relationship remains meaningful and growing due to the confirmed compute commitments and planned deployments, regardless of which AI model dominates.

The new structure lowers execution risk because the equity is pre-funded and hardware commitments are binding, offering clearer demand visibility through 2027 as Vera Rubin reaches full production in early 2026 and partner deployments start in the second half of 2026.

The original September 2025 plan of a $100 billion, milestone-contingent deal was never binding and was ultimately abandoned.