Key Points:
- NVIDIA CEO Jensen Huang criticizes OpenAI, exposing rising tensions between former partners.
- OpenAI explores custom AI chips, threatening Nvidia’s dominance in hardware.
- AI industry shifts from collaboration to rivalry as cost pressures and competition intensify.
NVIDIA CEO Jensen Huang publicly criticized OpenAI this week, signaling rising tensions between former partners as the AI boom drives customers to build custom chips, redraw alliances, and challenge NVIDIA’s dominance in computing hardware.
The comments mark a rare public fracture in a relationship that helped fuel the generative AI surge, with OpenAI relying on Nvidia’s powerful graphics processing units while Nvidia benefited from soaring demand driven by large language models.
According to sources, NVIDIA CEO Jensen Huang expressed dissatisfaction with OpenAI’s recent strategic direction, reflecting concerns inside Nvidia that key customers are moving to reduce reliance on its chips. Neither company disclosed details of private discussions.
The shift underscores a broader industry trend in which AI developers, cloud providers, and chipmakers increasingly compete across once-clear boundaries. What began as cooperation is now evolving into rivalry as companies seek more control over costs, performance, and long-term strategy.
Jensen Huang Flags Strain as Partners Turn Potential Rivals
For years, Nvidia and OpenAI operated in a mutually reinforcing cycle. OpenAI used Nvidia GPUs to train and run models such as GPT-4, while Nvidia’s data center business expanded rapidly on the back of AI demand.
That balance is changing. OpenAI has explored developing custom AI chips, a move that would lessen dependence on Nvidia’s H100 and upcoming B100 processors. Similar efforts are underway at Google, Amazon, and Meta, all aiming to bring more of the AI stack in-house.
“Huang’s concern is not personal but structural,” said Patrick Moorhead, an independent semiconductor analyst. “When your biggest customers start designing their own silicon, it’s a clear signal the market is shifting.”
OpenAI’s close partnership with Microsoft, which has invested more than $13 billion in the company, adds another layer of complexity. Microsoft both relies on Nvidia hardware and competes with Nvidia-backed customers in AI services.
Custom Chips and Cost Pressures Reshape AI Competition
The financial stakes are enormous. NVIDIA CEO Jensen Huang reported data center revenue of $130.5 billion in 2025, driven largely by AI workloads, making it the world’s most valuable semiconductor company by market capitalization.
That dominance has also encouraged customers to seek alternatives. Training and operating frontier AI models costs hundreds of millions of dollars, creating strong incentives for efficiency gains through custom silicon or optimized software.
OpenAI’s shift from a nonprofit research lab to a capped-profit company has sharpened its focus on economics, analysts said. More efficient models and hardware could reduce computing needs and curb future demand growth for Nvidia’s highest-end chips.
“Cost control is existential for AI labs at this scale,” said Stacy Rasgon, a Bernstein analyst. “Custom chips are one lever, even if they take years to materialize.”
Industry Faces Fragmentation as Stakes and Scrutiny Rise
The dispute reflects deeper forces reshaping the AI industry. Massive capital spending, tighter supply chains, and rising regulatory scrutiny are pushing companies toward vertical integration and consolidation.
At the same time, open-source AI models from companies such as Meta and Mistral are gaining traction, often running on a wider range of hardware. That trend could weaken pricing power across both models and infrastructure.
NVIDIA is responding by expanding its software ecosystem, including CUDA and AI development tools, to make switching away from its platform more difficult. It is also diversifying its customer base beyond large AI labs to enterprises and governments building domestic AI capabilities.
Despite the tension, analysts say Nvidia and OpenAI remain interdependent. Custom chips take years to develop, and Nvidia cannot afford to alienate a marquee customer that helped validate its AI strategy.
“The relationship is strained, not broken,” Moorhead said. “But it shows the era of easy collaboration in AI is ending.”
The public nature of NVIDIA CEO Jensen Huang’s remarks suggests the industry’s power structure is still in flux, with competition intensifying as the AI market matures and partnerships give way to guarded coexistence.
