Anthropic is reportedly in discussions to procure high-performance AI chips from Fractile, a London-based start-up. The agreement aims to diversify the hardware supply chain of the Claude creator, namely away from Nvidia.
The major AI company Anthropic is exploring a potential partnership with the British semiconductor firm Fractile to secure a steady supply of chips for custom inference and reduce the significant overheads associated with current semiconductor solutions.
ADVERTISEMENT
ADVERTISEMENT
According to reports, these talks represent a strategic effort by the San Francisco-based firm to decrease its dependency on Nvidia whilst enhancing the speed and efficiency of its current and next-generation models.
As the global demand for generative AI capacity continues to climb, the financial burden of the hardware required to run these systems has become a primary hurdle for developers.
Anthropic, which has received multi-billion-dollar investments from both Amazon and Google, currently relies heavily on Nvidia’s H100 units alongside custom processors provided by its cloud partners.
However, the high market price and limited availability of these industry-standard chips have squeezed profit margins, prompting firms to look elsewhere.
According to industry analysts, a deal with a specialised firm like Fractile could allow Anthropic to exert greater control over its technical infrastructure.
This strategy reflects a broader trend among tech giants, including Microsoft and Meta, who are increasingly moving away from general-purpose chips in favour of internal or boutique designs.
A shift in memory architecture and a boost for British technology
Founded in 2022 by Oxford PhD Walter Goodwin, Fractile has gained significant attention for its unconventional approach to processor design.
Unlike standard chips that must constantly shuttle data between the processor and separate memory modules, Fractile’s "memory-compute fusion" architecture keeps data directly on the chip using static random-access memory, or SRAM, which does not need to be refreshed.
According to the British start-up, this method can run large language models up to a hundred times faster than existing hardware while lowering operational costs by 90%.
While these performance claims are impressive, the technology is still in the development phase.
Fractile has not yet launched a commercial product, and its specialised chips are not expected to be ready for full-scale data centre deployment until 2027.
Despite the long timeline, the start-up is reportedly in negotiations to raise $200 million (€170.5m) in funding at a valuation exceeding $1 billion (€853m).
The potential partnership highlights the growing significance of the UK’s semiconductor sector on the world stage. If a formal agreement is reached, Fractile could become Anthropic’s fourth major chip supplier, joining the ranks of Nvidia, Google and Amazon.
According to market reports, the discussions remain at an early stage and no binding contract has been signed.
However, the interest from a major player such as Anthropic suggests that in the AI race, the ability to deliver faster and cheaper compute power is the defining factor.
