Anthropic is weighing building its own artificial intelligence chips, sources say - Taipei Times

Artificial intelligence (AI) lab Anthropic is exploring the possibility of designing its own chips, three sources said, as the company and its rivals respond to a shortage of AI chips needed to power and develop more advanced AI systems.

The plans are in early stages and the company might still decide to only buy AI chips and not design any, said two people with knowledge of the matter and one person briefed on Anthropic's plans.

The company has yet to commit to a specific design or put together a dedicated team to work on the project, one of the sources said.

Demand for its AI model Claude has accelerated this year, with the start-up's run-rate revenue now surpassing US$30 billion, up from about US$9 billion at the end of last year, Anthropic said.

Anthropic uses a range of chips, including tensor processing units (TPUs) designed by Alphabet's Google and Amazon's chips to develop and run its AI software and chatbot Claude.

Earlier this week, Anthropic signed a long-term deal with Google and Broadcom, which helps design the TPUs. That deal builds on the company's commitment to invest US$50 billion in bolstering US computing infrastructure.

Anthropic's discussions mirror similar efforts underway at large tech companies that are seeking to design their own AI chips, including Meta and OpenAI.

Designing an advanced AI chip could cost roughly half a billion dollars, as companies need to employ skilled engineers and spend to make sure the manufacturing process has no defects, industry sources said.

Separately, US Secretary of the Treasury Scott Bessent and Federal Reserve Chair Jerome Powell convened an urgent meeting with bank CEOs this week to warn of cyber risks posed by Anthropic's latest AI model, two sources familiar with the matter said.

Anthropic launched the powerful Mythos model earlier this week, but stopped short of a broad release, citing concerns it could expose previously unknown cybersecurity vulnerabilities.

The company has said the model is capable of identifying and exploiting weaknesses across "every major operating system and every major web browser."

Anthropic said it was in ongoing discussions with US government officials about the model's "offensive and defensive cyber capabilities."

A third source close to the matter reiterated Anthropic's outreach, saying the company briefed senior US government officials and key industry stakeholders on Mythos' capabilities ahead of its release.

The US Treasury-hosted meeting in Washington on Tuesday was aimed at ensuring banks are aware of the risks posed by Mythos and similar models, and are taking steps to defend their systems, one of the sources said.

Access to Mythos would be limited to about 40 technology companies, including Microsoft and Google, the start-up has said.