Tenstorrent CEO Jim Keller, who worked for Intel, AMD and Tesla, said in an interview with Nikkei Asia that he wants to lower the price of AI applications by designing chips that are more efficient than Nvidia's and, in the process, end the American giant’s global market dominance.
Headquartered in Toronto, Tenstorrent has offices in Austin, Texas and Santa Clara, California, as well as in Belgrade, Tokyo, Bangalore, and Seoul. The company claims to be developing AI processors that perform as efficiently as Nvidia's AI GPUs but at a lower cost.
At present, Nvidia is the global market leader when it comes to AI chips. Its AI accelerators command between 70% and 95% of the global market share, according to CNBC.
Keller also said in the interview that Tenstorrent hopes to sell its second-generation multipurpose AI chip by the end of this year.
According to Keller, Tenstorrent can achieve its objective because it does not use high-bandwidth memory (HBM), a popular type of advanced memory chip capable of transferring large amounts of data quickly. HBM is a vital component for generative AI chips and has played a significant role in the success of Nvidia's offerings.
But Keller pointed out that HBM was also one of the culprits behind the massive energy consumption and high prices of AI chips.
"Even the people that use HBM struggle with the cost of it, [and] the design time to build it,” he said.
By doing away with HBM, Tenstorrent will design a chip that can replace both GPUs and HBMs in some areas of AI development, adding that its new product will be "as cost-effective as possible".
