Fractile, a London-based startup that was set up in 2022 by by 28-year-old artificial intelligence PhD Walter Goodwin, has exited stealth with \u00a312 million in seed funding, to develop a chip to deliver exponential performance improvements for AI models.<\/p>\n\n\n\n
The company has already developed a radically different approach to the design of chips for AI inference, which it claims can deliver transformational improvements in performance for cutting-edge AI models in deployment. <\/p>\n\n\n\n
The funding round was co-led by Kindred Capital, NATO Innovation Fund, and Oxford Science Enterprises, with participation from Cocoa and Inovia Capital, alongside angel investors including Hermann Hauser (co-founder, Acorn, Amadeus Capital), Stan Boland (ex-Icera, NVIDIA, Element 14, and Five AI), and Amar Shah (co-founder, Wayve). To date, Fractile has raised \u00a314 million in total funding.<\/p>\n\n\n\n
Today\u2019s chips and highly developed tools and libraries are well-optimised for training large language models (LLMs), but they are unsuited to inference, which is the process of running live data (input tokens) through a specific model with learnt parameters to produce results (in LLMs, a series of output tokens).<\/p>\n\n\n\n
This means that models are very expensive to provision and run at scale, while AI performance is inhibited, and there is limited opportunity for AI model providers to drive differentiation.<\/p>\n\n\n\n
Fractile says there are two paths available to a company attempting to build better hardware for AI inference: specialisation \u2013 focusing on very specific workloads and building chips uniquely suited to those specific requirements, which is problematic because model architectures evolve rapidly in the world of AI \u2013 or fundamentally changing the way computational operations themselves are performed. <\/p>\n\n\n\n
The latter involves creating entirely different chips from new building blocks and building massively scalable systems on top of these. This is Fractile\u2019s approach. <\/p>\n\n\n\n
CEO, Dr Walter Goodwin, said: \u201cIn today\u2019s AI race, the limitations of existing hardware \u2013 nearly all of which is provided by a single company \u2013 represent the biggest barrier to better performance, reduced cost, and wider adoption. Fractile\u2019s approach supercharges inference, delivering astonishing improvements in terms of speed and cost. This is more than just a speed-up \u2013 changing the performance point for inference allows us to explore completely new ways to use today\u2019s leading AI models to solve the world\u2019s most complex problems.”<\/p>\n\n\n\n
The company claims that a Fractile system will achieve astonishing performance on AI model inference \u2013 initial targets are 100x faster and 10x cheaper \u2013 by using novel circuits to execute 99.99% of the operations needed to run model inference.<\/p>\n\n\n\n
A key aspect is a shift to in-memory compute, which removes the need to shuttle model parameters to and from processor chips, instead embedding computational operations directly into memory.<\/p>\n\n\n\n
Fractile has already built a world-class team with senior hires from NVIDIA, ARM, and Imagination, and has filed patents protecting key circuits and its unique approach to in-memory compute.<\/p>\n\n\n\n
The company is already in discussions with potential partners and expects to sign partnerships ahead of production of its first commercial AI accelerator hardware. Fractile will use the funding to continue to grow its team and accelerate progress towards its first product.<\/p>\n\n\n\n
Dr Goodwin concluded: \u201cWe\u2019re thrilled to have raised our funding from investors with a wealth of experience in the AI and chip industries, continue to grow our world-class team and further our technological development and partnerships.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"
Fractile, a London-based AI chip company, has exited stealth with \u00a312M in seed funding. Founded by AI PhD Walter Goodwin, Fractile is developing innovative in-memory compute technology to achieve 100x faster and 10x cheaper AI model inference.<\/p>\n","protected":false},"author":4,"featured_media":307,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[10,27],"tags":[119],"class_list":["post-306","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","category-software","tag-london"],"yoast_head":"\n