Fractile Secures 220M To Mass-Produce Revolutionary In-Memory Compute Chip

Fractile Raises $220 Million to Bring Its In-Memory-Compute Inference Chip into Production

London-based chip startup Fractile has raised $220 million to take its innovative in-memory-compute inference chip from design to production. The round was led by Accel, with former Intel CEO Pat Gelsinger participating as an angel and operating adviser.

This investment not only reinforces Fractile’s position as a leader in the European chip market but also underscores the growing interest in alternative architectures for inference computing. Fractile’s technology challenges the conventional approach to AI acceleration, where separate compute and memory dies are used, resulting in high energy consumption and latency.

The company’s design integrates matrix multiplications inside SRAM cells alongside compute logic, effectively removing the need for DRAM dependence, a significant constraint on inference cost. This approach enables Fractile’s chips to run frontier models up to 100 times faster and 10 times cheaper than current GPU setups.

However, it is essential to note that the company has only disclosed simulation and small-silicon results, rather than at-scale benchmarks against deployed GPU clusters. The first commercial chip is not expected until 2027, and the $220 million investment is aimed at taking the design through tape-out, software-stack build, and early customer integration.

The timing of this round is significant, as Fractile has been in early discussions with Anthropic to become a customer. If the relationship formalizes, Fractile would become Anthropic’s fourth named compute supplier, joining Nvidia, Google’s TPUs, and Amazon’s Trainium and Inferentia parts. This partnership would further validate Fractile’s technology and position it as a key player in the inference market.

Fractile is part of a growing group of European chip startups that argue the inference market is structurally distinct from training and can be won by specialized architectures tuned for throughput and energy per token rather than peak FLOPs. This thesis is gaining traction, with companies like Groq, Etched, Cerebras, and SambaNova also working on transformer-specific silicon or language-processing units.

Google itself is assembling a four-partner inference-chip supply chain with Broadcom, MediaTek, and Marvell to challenge Nvidia at the inference layer. Fractile’s claim of winning on watts per useful token is critical for cost-sensitive inference, an area where the company believes its in-memory architecture excels.

The investment round follows Fractile’s February announcement of a £100 million ($132 million) three-year expansion of its London and Bristol operations, including a new hardware-engineering site in Bristol. This move underscores the wider UK sovereign-AI push, which has produced partnerships between BT, Nscale, and Nvidia data centres.

Fractile’s founder and CEO Walter Goodwin has been the public face of the pitch, drawing engineers from Graphcore, Nvidia, and Imagination Technologies to build its software stack alongside the silicon. Tape-out and customer integration are the next visible milestones, which will test Fractile’s technology in real-world scenarios.

The competitive landscape for inference chips is becoming increasingly crowded, with multiple startups vying for market share. However, Fractile’s innovative approach and growing investor backing position it as a major player in this space. As the company moves towards production, its technology will be closely watched by industry observers, who are eager to see whether its claims hold up under real-world conditions.

The $220 million investment in Fractile highlights the growing interest in alternative architectures for inference computing. As the AI market continues to evolve, it is likely that we will see more innovation and competition in this area, with companies like Fractile pushing the boundaries of what is possible.

The EU tech scene can now look forward to seeing Fractile’s technology take shape. With its innovative approach and growing investor backing, Fractile is well-positioned to make a significant impact in the inference market.

Original Source

• Read the full article here