AI research and safety company Anthropic is reportedly in early-stage negotiations to purchase advanced AI inference chips from UK-based startup Fractile. The potential deal, first reported by The Information, aims to secure a supply of these specialized chips when they become available to the market in 2027.
This strategic move by Anthropic underscores the intense competition among leading AI developers to secure critical computing infrastructure. As AI models grow in complexity and scale, the demand for high-performance inference chips—designed to efficiently run trained AI models—is rapidly increasing. Proactive agreements like this are becoming essential to mitigate future supply chain risks and ensure continuous innovation.
Fractile, a lesser-known but emerging player in the AI hardware space, focuses on developing inference accelerators. While details about their chip architecture remain sparse, a significant deal with a major AI entity like Anthropic could position Fractile as a notable contender in a market currently dominated by a few established players. Securing a long-term supply agreement years in advance allows Anthropic to plan its future AI deployments and service offerings with greater certainty regarding hardware availability.
The reported talks highlight the broader trend of AI companies seeking to diversify their hardware partnerships beyond conventional suppliers, as they look for specialized and potentially more efficient alternatives to meet their unique computational demands.