IT之家 News on October 28: According to a report by Business Insider in the early hours of today (the 28th), Anthropic uses multiple types of chips simultaneously to handle AI computing tasks in order to spread risk. In the long run, this strategy could shake Nvidia's market dominance.
On the 26th local time, Anthropic's IPO application filing (Form S-1) showed that the companyintegrates the computing power provided by multiple strategic partnersinto a unified resource pool, then uses software to allocate computing tasks to the chips where they will deliver the greatest benefit.
IT之家 learned from the report that the chips Anthropic uses fall into four major categories:Amazon Trainium, Google TPU, Nvidia GPU, and AMD GPU. To achieve this cross-platform scheduling, the company spent years and a great deal of money.
Analysts believe this capability could help large AI companiesgradually reduce their dependence on Nvidiaand gain more leverage in price negotiations.
Patrick Moorhead, founder and CEO of Moor Insights & Strategy, believes that Anthropic is not just starting to use multiple chips, but has begun to present this capability as a major advantage to attract investors. 'What could really be hit is Nvidia's pricing power. A customer who canflexibly allocate computing tasksacross four platforms is negotiating from a completely different position than one that cannot.'
IDC analyst Ashish Nadkarni pointed out that another benefit of using multiple chips isreducing the impact of fluctuations in Nvidia chip supply。
However, this strategyalso poses only limited threat to Nvidia. Task migration between different chip platforms remains difficult, especially for model training, which is harder than model inference. Only the largest AI companies have the resources to achieve cross-platform scheduling.
Anthropic is also still spending billions of dollars on Nvidia computing power.
Gil Luria, head of technology research at D.A. Davidson, believes that large companies like OpenAI will likely adopt similar approaches, but for smaller companies, investing so much money and effort in thismay not be worthwhile。
Brad Gastwirth, head of global research and market intelligence at Circular Technology, pointed out that Anthropic's strategywill not pose a major threat to Nvidia in the short term. Customers having more chip choices does not mean Nvidia's advantages in performance, networking interconnect, and software will disappear.
If AI computing power gradually becomes commoditized, Nvidia will face broader risks.
Ken Griffin, CEO of Citadel, warned in an interview with Bloomberg last week that if compute supply exceeds demand, rental prices could fall: 'GPUs, like airlines, have very low marginal costs of operation. Once there is excess capacity, prices will plummet.'