Pat Gelsinger, the former CEO of Intel, has announced his investment in Fractile.ai, a British AI chip startup founded in 2022 by Oxford Ph.D. graduate Walter Goodwin. The company has raised $15 million in funding from prominent investors, including the NATO Innovation Fund, Kindred Capital, and Oxford Science Enterprises.

A Revolutionary Approach to AI Hardware

Fractile.ai is pioneering compute-in-memory technology, which combines processing and memory into a single component, addressing key bottlenecks in AI inference. This approach eliminates the reliance on traditional architectures, such as GPUs with separate high-bandwidth memory (HBM), significantly reducing power consumption and latency.

Gelsinger highlighted the importance of this innovation, stating: “We need orders of magnitude cost and power reductions to make AI useful at scale. Fractile.ai's in-memory compute tackles the memory bottleneck holding back GPUs while slashing power consumption, the biggest constraint for scaling up data centers over the next decade.”

Game-Changing Potential

According to simulations, Fractile.ai's chips could deliver AI inference performance 100 times faster and 10 times cheaper than NVIDIA's GPUs, as reported by Fortune. While test chips have yet to be fabricated, the company's technology has already attracted attention from key industry players.

Gelsinger's Post-Intel Move

After stepping down as Intel's CEO in late 2024, Gelsinger has shifted his focus to investing in cutting-edge AI technologies. Despite Intel's struggles to compete with NVIDIA in AI hardware, Gelsinger sees Fractile.ai as a disruptive force capable of reshaping the field.

Future Outlook

With a team of former NVIDIA and Arm engineers, Fractile.ai is well-positioned to push the boundaries of AI hardware. Its compute-in-memory technology offers a path to scalable, efficient AI systems, addressing growing demands for lower costs and energy consumption. As Gelsinger noted, “This is one of the most exciting approaches to AI acceleration I've ever seen.”

Fractile.ai's success could mark a turning point in AI hardware, unlocking new possibilities for large-scale AI deployment.