anthropic is currently in negotiations with London-based startup Fractile to procure its inference chips, which are expected to enter mass production and be deployed in data centers as early as next year. Founded in late 2022, FrACTile's core Technology replaces the High Bandwidth MEMOry (HBM) relied upon by GPUs with Static Random-Access Memory (SRAM). This APProach minimizes data transfer between the chip and external memory, significantly reducing inference power consumption and costs. Other companies pursuing a similar path include Cerebras and Groq.
The agreement is still in its early stages, with the scale unknown and the possibility of the talks falling through. However, this potential order has become a key selling point for Fractile's latest funding round. The company is seeking over $100 million in financing at a Valuation exceeding $1 billion, with Founders Fund, 8VC, and Accel all in discussions. Previously, Fractile had rAIsed only $15 million from investors including Kindred capital, the NATO Innovation Fund, and Oxford Science Enterprises.
Anthropic has consistently pursued a strategy of diversifying its chip supply. The company already rents cloud servers from Google and Amazon and, last fall, committed $30 billion to rent NVIDIA servers from Microsoft Azure (with Nvidia investing $10 billion and Microsoft $5 billion as part of the deal). It recently also agreed to purchase google's in-house developed chips for use outside of Google Cloud. Reuters previously reported that Anthropic is also considering designing its own inference chips, a strategy similar to that of OpenAI and Meta.
High inference costs are a current pain point for Anthropic. The company's inference business gross margin fell short of internal expectations last year. Furthermore, a recent surge in demand for Claude Code has led to compute shortages, resulting in some users being rate-limited during peak hours and sparking public protests from developers. Nvidia CEO Jensen Huang recently expressed regret for not investing in Anthropic earlier, suggesting it might have prevented the company from shifting so heavily towards Google and Amazon chips.
Comments & Questions (0)
No comments yet
Be the first to comment!