IT Home, October 28 — According to Business Insider's report early this morning (28th), Anthropic is using multiple types of chips to handle AI computing tasks, diversifying its risks. This strategy could, in the long term, challenge NVIDIA's market dominance.
On the 26th, Anthropic’s filing (Form S-1) revealed that the company consolidates computing power provided by multiple strategic partners into a unified resource pool and uses software to allocate computational tasks to the most efficient chips.
IT Home learned from the report that Anthropic uses four categories of chips: Amazon Trainium, Google TPU, NVIDIA GPU, and AMD GPU. To achieve cross-platform scheduling, the company has spent years and significant resources.
Analysts believe this capability could enable major AI companies to gradually reduce their dependence on NVIDIA and gain more leverage in price negotiations.
Patrick Moorhead, founder and CEO of Moor Insights & Strategy, believes that Anthropic is not just beginning to use multiple chips, but is now leveraging this capability as a major advantage to attract investors. “The real impact may be on NVIDIA’s pricing power. Customers who can flexibly allocate computing workloads across four platforms are in a completely different negotiating position than those who cannot.”
IDC analyst Ashish Nandkarni noted that adopting multiple chips also helps mitigate the impact of supply fluctuations in NVIDIA chips.
However, this strategy poses limited threat to NVIDIA. Migrating tasks between different chip platforms remains difficult, especially for model training, which is more challenging than model inference. Only the largest AI companies have the resources to achieve cross-platform scheduling.
Anthropic is also investing billions of dollars in NVIDIA's computing power.
Jill Luria, Head of Technology Research at D.A. Davidson, believes that large enterprises like OpenAI are likely to adopt similar solutions, but for smaller companies, investing significant funds and effort into this may not be cost-effective.
Brad Gastwirth, Head of Global Research and Market Intelligence at Circular Technology, noted that Anthropic’s strategy will not pose a significant threat to NVIDIA in the short term. Having more chip options does not mean NVIDIA’s advantages in performance, network interconnectivity, and software will disappear.
If AI computing power becomes increasingly commoditized, NVIDIA will face even broader risks.
Ken Griffin, CEO of Citadel Investments, warned in a recent Bloomberg interview that if hash power supply exceeds demand, rental prices could fall: “GPUs, like airlines, have very low marginal operating costs. Once there’s excess capacity, prices crash.”