Huang Renxun: NVIDIA's investments in OpenAI and Anthropic will come to a close

Nvidia CEO Jensen Huang recently stated at a Morgan Stanley conference that the company’s recent $30 billion investment in AI startup OpenAI may be its last before the company’s planned IPO at the end of the year. This announcement confirms that the widely discussed $100 billion infrastructure partnership from September last year will not materialize. Meanwhile, Nvidia’s expected $10 billion investment in another AI company, Anthropic, is also coming to an end.

OpenAI Funding and Nvidia’s Investment Strategy Adjustment

Nvidia recently participated in a new funding round for OpenAI totaling $110 billion, with Nvidia investing $30 billion, Amazon pledging $50 billion, and SoftBank committing $30 billion. According to CNBC, Huang said at the Morgan Stanley meeting that since OpenAI plans to go public by the end of the year, the previously discussed $100 billion infrastructure partnership is no longer feasible. This indicates Nvidia is adopting a more cautious capital allocation approach, and Huang announced that the $10 billion investment in Anthropic has also been put on hold.

(Huang Huang responds to WSJ leak in Taipei: No dissatisfaction with OpenAI, no promise to invest $100 billion)

Securing Computing Resources and Infrastructure Deployment

In this agreement, OpenAI secured exclusive inference computing capacity of 3 GW on Nvidia’s Vera Rubin system, along with 2 GW of training capacity, which is critical for its data center operations. Unlike the framework from September last year, which depended on the progress of new system deployment, this investment is not tied to specific deployment milestones. This reflects a shift in industry demand from infrastructure planning to more flexible, immediately applicable solutions.

Industry Shift: From Training to Inference

As technology matures, the hardware needs of the AI industry are transitioning from “model training” to the faster-response “inference” phase. To meet this trend, Nvidia is developing new chips optimized for inference, with OpenAI expected to become a major customer. At the same time, OpenAI is also adopting Amazon’s inference-optimized chips and Google’s TPU, indicating that leading AI companies are actively diversifying supply chains to reduce reliance on a single hardware supplier.

This article, Jensen Huang: Nvidia’s investments in OpenAI and Anthropic will come to an end, first appeared on Chain News ABMedia.