Fractile: The future of AI inference
Summary
- AI is shifting from generating short answers to performing sustained, multi-step work involving reasoning, coding, & iteration.
- Inference has become a critical infrastructure constraint as models spend more time processing at runtime.
- While training previously dominated infrastructure discussions, inference now determines speed, cost, & latency for usable AI output.
- Fractile, founded in 2022 by Oxford PhD Walter Goodwin, is building chips & systems optimized specifically for emerging inference workloads.
- The company aims to improve AI efficiency by rethinking compute, memory bandwidth, & software from first principles.
- Inference optimization is essential for advancing agentic coding, scientific discovery, enterprise automation, & long-horizon research.
- Fractile's team combines deep expertise in AI research, chip architecture, systems design, software, & manufacturing.
- The article announces the co-leading of Fractile's Series B funding to support its mission of making inference faster, cheaper, & more scalable.
AI is evolving from short answers to sustained work.