Anthropic in talks to secure custom AI chips for Claude
What's the story
Anthropic, a leading player in the artificial intelligence (AI) space, is said to be in early-stage talks with UK-based semiconductor start-up Fractile. The discussions are centered around securing a future supply of specialized AI inference chips from Fractile. These advanced chips are designed to run trained AI models more efficiently, thereby reducing costs and improving speed.
Strategic move
Fractile's chips expected to be ready later in the decade
Fractile's chips are expected to be launched later in the decade, hinting at a long-term deal between the two companies. The specialized inference chips being developed by Fractile are specifically designed to efficiently run trained AI models. Unlike general-purpose GPUs, these chips can handle repetitive, high-throughput workloads with lower energy consumption and reduced latency. This makes them ideal for enterprise AI tools, customer service automation, and large-scale API deployments.
Supply diversification
Reducing reliance on NVIDIA
The move to partner with Fractile also comes as part of Anthropic's strategy to reduce reliance on dominant chip suppliers such as NVIDIA. Currently, NVIDIA's GPUs power much of the global AI infrastructure. However, their high cost and limited supply have become constraints for many companies. By exploring alternative hardware providers like Fractile, Anthropic could be looking at diversifying its supply chain and ensuring more predictable access to compute resources as it scales its Claude family of AI models.
Market shift
Custom silicon trend in AI
The timing of Anthropic's talks with Fractile also reflects a wider industry trend toward custom AI silicon. Major tech companies such as Google, Amazon, and Microsoft have already developed in-house chips tailored for AI workloads. These efforts are driven by the need to optimize performance, reduce long-term costs, and gain tighter control over the hardware-software stack. For independent AI labs like Anthropic, partnerships with specialized chip start-ups could offer a similar path without full in-house development.