NVIDIA CEO Jensen Huang has indicated that the company's recent $30 billion investment in OpenAI is likely to be its "last" before the AI startup proceeds with its public offering. Huang made these remarks on Wednesday at the Morgan Stanley Technology, Media, and Telecom Conference.
Huang also mentioned that NVIDIA's $10 billion investment in OpenAI's competitor, Anthropic, is also "very likely" to be the final one. The investment in OpenAI, part of a $110 billion funding round announced by the startup last Friday, may represent NVIDIA's last capital injection ahead of OpenAI's anticipated IPO around the end of this year.
According to Huang, a previously discussed opportunity for a $1 trillion investment, promoted in September last year as part of a large-scale infrastructure deal, is likely "not in the plans." He explained, "The reason is they are going public."
Huang's comments come amid months of speculation regarding the scope of NVIDIA's relationship with OpenAI. In a quarterly filing in November, the chipmaker disclosed that the previously announced $1 trillion deal might not materialize, and reports in January suggested the agreement had been "shelved." NVIDIA's February quarterly filing contained similar language, noting that there is "no assurance" the company will finalize an "investment and cooperation agreement" with OpenAI.
Last Friday, OpenAI confirmed that as part of the agreement, it secured 3 gigawatts of dedicated inference computing capacity and 2 gigawatts of training capacity on NVIDIA's Vera Rubin systems for AI data centers. The September transaction between the two companies had previously shaken the tech industry and spurred a series of subsequent infrastructure agreements. That deal outlined a structure where NVIDIA would invest in OpenAI over several years as new supercomputing facilities came online. In contrast, the $30 billion investment is not tied to any deployment milestones.
NVIDIA has been one of the biggest beneficiaries of the AI boom, as it produces the graphics processing units essential for AI companies to train models and handle large workloads. However, demand from AI firms has shifted from training to inference—a type of processing that allows AI models to respond quickly to user queries—posing some pressure on the company. Reports indicate that NVIDIA is developing a new chip specialized for inference, with OpenAI expected to be one of the largest customers for the new processor.
In February, OpenAI announced it would purchase "dedicated inference capacity" in large volumes from NVIDIA. The company is also heavily investing in inference-optimized chips from Amazon and utilizes Google's tensor processing units.
According to a source familiar with the matter who requested anonymity due to the private nature of the details, OpenAI CEO Sam Altman is scheduled to speak at the Morgan Stanley conference on Thursday.
Comments