At a company event on Wednesday, AMD unveiled new data-center products, including updated microprocessors, an AI accelerator and the Helios server rack.<\/p>\n
AMD said its new offerings compare favorably with rival products.<\/p>\n
Anthropic Deal Expands AMD\u2019s AI Reach<\/p>\n
AMD also announced a strategic partnership with Anthropic and said it may invest up to $5 billion in the AI company in the future.<\/p>\n
Under the deal, Anthropic will deploy 2 gigawatts<\/a> of AMD Instinct MI450 Series GPUs in AMD Helios rack-scale systems, with the first gigawatt expected in the first half of 2027, Reuters reported on Wednesday.<\/p>\n Tom Brown, co-founder and compute head at Anthropic, told Reuters on Wednesday that access to compute remains central to keeping Claude competitive and meeting customer demand.<\/p>\n The deal also adds to AMD\u2019s broader AI customer momentum. AMD previously announced a multiyear OpenAI agreement that could generate tens of billions of dollars in annual revenue and gave OpenAI the option to buy up to about 10% of the chipmaker.<\/p>\n Helios Moves Into Production<\/p>\n Su said AMD sees the high-performance AI computing market reaching a $2 trillion opportunity over the next four to five years.<\/p>\n She said AMD\u2019s Helios rack-scale system and Venice processors are now in full production, marking a major step-up in AI capability.<\/p>\n She said the newest Helios systems deliver more than 30 times the performance of the previous generation.<\/p>\n Customer Scale And Efficiency<\/p>\n Su said OpenAI and Meta have discussed multi-gigawatt-scale MI450 deployments, while Anthropic plans to deploy up to 2 gigawatts of MI450 in Helios systems starting in the first half of 2027.<\/p>\n She said customers are focused on tokens per dollar as AI usage grows. Su said Helios offers more performance than competing systems, with a 10% to 15% performance advantage and up to a 30% benefit in tokens per dollar, helping customers improve productivity while expanding AI compute.<\/p>\n NVIDIA Competition Intensifies<\/p>\n AMD continues to challenge Nvidia, which still dominates AI accelerators and generates far more AI data-center revenue.<\/p>\n The company projected the AI accelerator market could reach $1.4 trillion by 2030 and said its broader 2030 opportunity could reach $2 trillion across AI accelerators, CPUs and related data-center markets, Bloomberg reported<\/a> on Thursday.<\/p>\n Partnerships Broaden Opportunity<\/p>\n OpenAI executive Sachin Katti told Bloomberg on Thursday that the company expects to deploy Helios at massive scale as it expands AI infrastructure.<\/p>\n AMD also announced a partnership with Cerebras Systems Inc. (NASDAQ:<\/a>CBRS<\/a>) to combine Helios and Cerebras servers for AI customers focused on faster response times.<\/p>\n Cerebras CEO Andrew Feldman told Bloomberg that the companies aim to move ahead of Nvidia in that segment, with the first deployment expected in Cerebras data centers in the fourth quarter before broader commercial availability.<\/p>\n