Jensen Huang has thrown shade on Sam Altman's semiconductor ambitions.
Huang attended the 2024 World Governments Summit in Dubai, UAE, on February 12. At the summit, he spoke with the UAE's AI minister, Omar Al Olama, about the future of AI.
Al Omar began the conversation by asking Huang, "How many GPUs can we buy for US$ 7 trillion?" Al Omar's pointed question referenced reports that OpenAI's CEO Sam Altman hopes to raise US$7 trillion to address the global supply of semiconductor chips.
After a restrained laugh, the Nvidia CEO quipped, "Apparently, all the GPUs." Huang continued to comment on Altman's astronomic investment.
Huang told Al Omar that such a substantial investment is unnecessary to establish an alternative semiconductor supply chain solely for AI and added that the focus should remain on innovating GPU architecture to enhance performance.
Huang said Nvidia had already achieved a one million-fold increase in AI performance over the past decade. "If you just assume that computers never get any faster, you might come to the conclusion [that] we need 14 different planets and three different galaxies and four more Suns to fuel all this," added.
Altman's global fundraising efforts have made waves in the industry. The Wall Street Journal reported that he is seeking $7 trillion in funding from the Middle East to support the company's semiconductor project, entering competition with Nvidia.
The reports have garnered a range of responses. Some industry watchers supported Altman's reported ambitions, while others were more skeptical.
U.S. AI expert Gary Marcus wrote a post on Substack discussing Altman's reported investment dream titled "Seven reasons why the world should say No to Sam Altman." Referencing a scene from the American comedy Happy Days, he wrote, "In asking for $7 trillion, it is pretty obvious to most that Sam Altman has jumped the shark."
Altman has not suffered in silence. In a post on X, he clapped back at critics, "You can grind to help secure our collective future, or you can write Substacks about why we are going [to] fail."
The AI chip market is largely dominated by Nvidia, and the company's H100 high-end AI GPU is recognized as an essential GPU for training large language models (LLMs). However, alternative plans have emerged from companies such as OpenAI and Meta due to the GPU's high price and limited production capacity.
Looking at the capital expenditures of advanced process chip foundries such as TSMC and Samsung Electronics, which amount to tens of billions or even hundreds of billions of dollars annually, Altman reportedly believes that significant investments are required in research and development, facilities, and professional talent to make a mark in this field.
In a February post on X, Altman wrote, "the world needs more ai infrastructure--fab capacity, energy, datacenters, etc--than people are currently planning to build."
Nvidia, which heavily relies on TSMC's advanced process production capacity, has never been rumored to have a similar plan to build a global network of chip foundries like OpenAI. It continues to maintain close cooperation with supply chains worldwide, similar to the close collaboration between AMD and TSMC.
