Gerstner on OpenAI’s Nvidia and AMD deals: the best chips will win

Investor Brad Gerstner urged caution after OpenAI announced new chip agreements with Nvidia and AMD. He told CNBC the arrangements are announcements rather than deployments and said results, not headlines, will decide who leads. He said winners will be decided by delivery at scale and real-world performance,“the best chips will win.”

His comments followed reports of a large AMD pact as OpenAI seeks more capacity. Gerstner said the news offers more evidence that the world will remain compute-constrained even as producers add new fabs and data centers. Demand for AI across sectors is climbing so fast that shortages are likely to persist. This view echoes a broader belief that access to accelerators now shapes product roadmaps, capital spending, and national strategy.

The partnerships also reflect an intensifying AI arms race with geopolitical stakes, especially competition between the United States and China. China’s DeepSeek raised alarms last year by touting a lower-cost model than leading U.S. systems and has continued to release open-source models built on domestically made chips. A recent government report, cited by Axios, warned of national security risks and said DeepSeek’s outputs more often align with Chinese Communist Party viewpoints than comparable U.S. models.

For OpenAI, working with AMD alongside Nvidia is seen as a way to diversify supply, ramp production, and attempt more complex systems. The company’s president told CNBC’s “Squawk on the Street” that demand across the entire ecosystem, not just at OpenAI, will create absolute compute scarcity, and urged industry collaboration to meet customer needs. The message aligns with Gerstner’s view that success will require the right mix of silicon, software, and execution.

Until deployments materialize, Gerstner views the agreements as early signals. He expects the market to reward actual performance and availability as providers and chipmakers race to turn plans into capacity. In the meantime, the scramble for compute, the push to lower costs, and the geopolitical stakes suggest the AI contest will intensify through the next wave of model building and deployment.