Anthropic's push to run multiple types of artificial intelligence (AI) semiconductors together has raised the possibility it could challenge Nvidia's market dominance.

On Oct. 27, Business Insider reported that Anthropic, in an S-1 filing made public ahead of its initial public offering (IPO), said it manages computing resources from multiple strategic partners like a single integrated pool.

Anthropic uses its own software to assign computing by task to the chip that is most efficient. The chips it uses are four types: Amazon Trainium, Google tensor processing units (TPUs), Nvidia graphics processing units (GPUs) and AMD GPUs. The company said securing that flexibility required years of development and significant investment.

Such a multi-chip strategy could, over the long term, reduce AI companies' reliance on Nvidia and increase their pricing leverage. Companies that can use multiple semiconductor platforms can reduce dependence on a specific supplier and respond more flexibly to supply swings.

Still, switching between chips remains difficult. Changing chips during AI model training is more complex than at the inference stage, and building a multi-chip infrastructure requires massive resources and investment. Anthropic is also still spending billions of dollars to secure Nvidia computing resources, the report said. That has led to analysis that only large AI companies may be able to implement a multi-chip strategy in practice.

Nvidia's technical edge is also unlikely to disappear in the short term. Brad Gastwirth (브래드 개스트워스), head of global research and market intelligence at Circular Technology, said that even as chip options increase, the short-term threat is limited as long as Nvidia maintains its performance, networking and software competitiveness.

Over the longer term, the commoditisation of AI computing resources and oversupply could be a bigger variable. Ken Griffin (켄 그리핀), CEO of U.S. financial firm Citadel, warned that if supply outpaces demand, rental rates for computing resources could fall. He said GPUs, like airlines, have low marginal costs for additional operation, and said, "The moment oversupply occurs, prices plunge."

Anthropic's strategy does not mean Nvidia's technical advantage will disappear immediately. But the more large AI companies flexibly use multiple semiconductor platforms, the more Nvidia's pricing power and customer dependence structure could face a long-term challenge.