AMD secures multi-billion-dollar Anthropic AI server deal, challenging Nvidia’s dominance

0

Advanced Micro Devices (AMD) has signed a multi-year agreement worth tens of billions of dollars with artificial intelligence startup Anthropic to supply next-generation AI servers, marking one of the chipmaker’s most significant enterprise contracts to date.

‎‎The deal, first reported by The Wall Street Journal, sent AMD shares up about 2 per cent in midday trading as technology stocks rebounded.

‎‎Under the agreement, Anthropic will purchase up to two gigawatts of AMD’s forthcoming Instinct MI450 AI accelerators, with deliveries expected to begin in the first half of 2027. AMD will also invest up to US$5 billion in Anthropic as deployment milestones are achieved, representing the company’s first direct equity investment in the AI developer.

‎‎AMD Chief Executive Officer Dr Lisa Su said the partnership reflects the company’s ambition to become a core supplier of Anthropic’s computing infrastructure.

‎‎”We have very much wanted to be a major part of their infrastructure,” Su said, adding that engineering teams from both companies had been working together behind the scenes for several months.

‎‎The agreement is widely seen as a significant step in AMD’s efforts to challenge Nvidia’s leadership in the rapidly expanding AI chip market.

‎‎Industry analysts said the transaction signals a shift in the way large-scale AI computing agreements are being negotiated. Barclays analyst Tom O’Malley noted that, unlike several previous contracts in the sector, AMD did not need to issue stock warrants to secure Anthropic’s commitment.

‎‎He said the deal suggests equity incentives are becoming less important in long-term chip supply agreements as demand for AI computing capacity continues to outstrip supply.

‎‎O’Malley also observed that the value of AI infrastructure contracts continues to rise, with pricing per gigawatt now exceeding the roughly US$15 billion seen in earlier agreements. He attributed the increase to growing demand for more powerful server processors and expanding compute requirements.

‎‎Although he expressed some surprise at the initial share price volatility, given that AMD’s Analyst Day is scheduled for the following day, O’Malley said the agreement strengthens confidence in the company’s long-term growth prospects and supports expectations that the server central processing unit (CPU) market could approach a total addressable market of US$200 billion.

‎‎For Anthropic, the agreement secures access to critical computing resources at a time when demand for AI infrastructure continues to exceed available supply.

‎‎The company currently relies on a combination of Google’s Tensor Processing Units (TPUs), Amazon’s Trainium chips and Nvidia’s graphics processing units (GPUs) to train and run its AI models. Under the new agreement, Anthropic will deploy AMD’s Instinct MI450 chips in its own data centres while also leasing additional computing capacity through major cloud providers and specialised cloud infrastructure companies.

‎Su stressed that developing AI infrastructure on such a scale requires long-term planning.

‎‎”You can’t just wake up one morning and say, ‘Oh, I want a gigawatt of compute tomorrow.’ You actually have to plan 12, 18, 24 months in advance,” she said.

‎‎AMD is also exploring options to provide financial support for Anthropic’s future data centre leasing commitments, according to a person familiar with the discussions. The move reflects a broader trend in the AI industry, where semiconductor companies are increasingly helping leading AI developers finance the enormous infrastructure investments required to build and operate frontier AI models.

‎‎The agreement underscores intensifying competition in the AI semiconductor industry as AMD seeks to expand its presence in a market long dominated by Nvidia, while providing Anthropic with the computing capacity needed to support its next generation of AI systems.

LEAVE A REPLY

Please enter your comment!
Please enter your name here