A powerful challenge has been thrown at the artificial intelligence (AI) semiconductor market. AMD, a semiconductor giant, has made a strategic investment of up to 7 trillion won in Anthropic, a developer of the AI model Claude, and has secured a major purchase of its latest AI chips, establishing a mega-scale strategic alliance that directly confronts Nvidia's monopoly.
On the 22nd (local time), AMD signed a contract to invest up to $5 billion (approximately 7.4 trillion won) in Anthropic in phases. Simultaneously, Anthropic agreed to purchase AMD's latest graphics processing unit (GPU) MI450 at a capacity of 2 gigawatts (GW) starting in the first half of 2027. Anthropic plans to utilize some of the purchased chips in its own data centers, while leasing the rest through cloud providers or 'neo-cloud' companies to expand infrastructure.
This deal is interpreted as AMD's all-out challenge to Nvidia, which dominates the AI chip market, and as part of a strategy to diversify chip dependency among AI developers including Anthropic. Just as Anthropic previously established a similar strategic relationship with Google by utilizing Google's Tensor Processing Unit (TPU), this alliance with AMD demonstrates the urgent situation of AI developers seeking to reduce their dependence on specific chip manufacturers and secure stable infrastructure. AMD CEO Lisa Su emphasized strengthening the relationship with Anthropic, stating that she has long anticipated becoming an important pillar of Anthropic's infrastructure.
The two companies have signed a technology collaboration agreement that goes beyond simple investment and chip purchases, signaling synergies. They have decided to conduct engineering collaboration to improve AMD's chip technology performance using Anthropic's Claude model. This provides AMD with an opportunity to strengthen its chip competitiveness and solidify its position in the AI market, while enabling Anthropic to build a mutually beneficial relationship that can improve model development and operational efficiency through optimized infrastructure.