Legendary Chip Architect Jim Keller Disagrees with Sam Altman's $7 Trillion Plan for AI Chips
Renowned chip architect Jim Keller recently responded to OpenAI CEO Sam Altman's ambitious plan to raise $7 trillion for the production of artificial intelligence (AI) chips. In a Twitter tweet, Altman suggested the need for a network of fabs to meet the increasing demand for AI chips. However, both Keller and Nvidia CEO Jensen Huang believe that the focus should be on improving the sophistication of these processors rather than increasing their quantity.
Keller, who currently works at Tenstorrent developing AI and high-performance computing (HPC) processors, expressed his disagreement with Altman's fundraising target. In a quote tweet, he confidently stated, "I can do it for less than $1 trillion." This echoes the sentiment shared by Huang, who emphasized the importance of architectural innovation in AI processors.
Altman's plan involves a significant expansion of the semiconductor supply chain, which could potentially lead to overcapacity at foundries and a decrease in the value of these chips. In contrast, Keller and Huang argue that the key is to enhance the sophistication of the processors and simplify the supply chain to reduce costs. Keller specifically suggested eliminating excessive profit margins added by various supply chain participants.
Tenstorrent has a well-defined roadmap in place for the development of AI and HPC processors. Their approach involves increasing the number of processing units with each iteration, resulting in more advanced and efficient AI processors. However, the demand for AI performance is rapidly growing, necessitating innovative solutions.
Altman's fundraising goal is significantly higher than the current valuation of the global semiconductor industry, which amounted to $527 billion in sales last year. While the industry is expected to reach the $1 trillion mark by 2030, chip manufacturers are already investing substantial amounts of money into chip fabrication equipment. In 2022 alone, they invested $99.5 billion, and this year, they are projected to allocate $97 billion for fabrication tools.
As the debate continues, it remains crucial to strike a balance between meeting the demand for AI chips and ensuring their technological advancements. Both Keller and Huang's viewpoints shed light on the importance of focusing on the sophistication of AI processors to drive cost efficiencies and improve overall performance. Only time will tell which approach prevails in the ever-evolving AI landscape.
Analyst comment
Positive news: Jim Keller and Nvidia CEO Huang’s emphasis on the need for more sophisticated AI processors is a positive development. It suggests a focus on quality rather than quantity, which can lead to improved performance efficiency. This approach aligns with Tenstorrent’s roadmap for developing advanced processors.
Market prediction: The market for AI chips is likely to see a shift towards improving processor architecture rather than increasing chip production. This may drive innovation and efficiency in AI hardware, ultimately reducing costs and increasing performance. The demand for AI performance is expected to continue growing, making this approach crucial for the market’s future success.
