342
Productivity & Workflow347
Software Development244
Automation & Workflow240
Marketing & Growth200
AI Infrastructure & MLOps191
Photography & Imaging155
Writing & Content Creation194
Design & Creative170
Data & Analytics166
Customer Support152
Education & Learning136
Sales & Outreach129
Voice & Speech130
Operations & Admin119
London-based Fractile raised $720m and says its AI chips with built-in memory could speed up chatbot responses. It plans a first product next year.
In short: London-based chip start-up Fractile has reached a $7bn valuation after raising $720m to build AI chips that focus on faster memory.
Fractile, a four-year-old company headquartered in London, has raised $720m from a group of well known investors. They include Thrive Capital, Founders Fund, Conviction, CapitalG (Alphabet’s venture arm), and Lightspeed Venture Partners. The round values Fractile at $7bn including the new money, up from about $1.1bn in a funding round in May, according to people familiar with the deal.
Fractile is working on its own chips for AI. The company says its design could make “inference” faster, which is the step where an AI system answers a question (like when a chatbot replies). Founder Walter Goodwin argues that many AI chips have gained computing power faster than they have improved access to memory, which is where data is stored for quick use.
Goodwin described today’s AI models as “drinking through a straw,” meaning the chip may be powerful, but it can be slowed down by how quickly it can pull data from memory. Fractile says it is building custom DRAM memory (a common kind of computer memory) and stacking it directly on top of its chip, like putting a pantry right next to the chef instead of across the street.
Fractile has not released a production chip yet, and it plans to launch its first product next year. The Financial Times reported that Anthropic has shown interest in being an early customer.
AI services are getting more expensive to run, and chips are a big reason why. If companies can answer questions faster using less electricity and fewer chips, that can lower costs and reduce delays for everyday tools people use at work and at home.
Source: Financial Times