UK AI chip start-up Fractile sold about ยฃ577 million of new Series C shares between 9 September and 23 October 2026, by our sum of nine share-allotment returns it filed at Companies House. Tech Funding News, which reported the filings on 30 October, put the total at “around $774 million”, about 3.5 times the $220 million round we covered in May, by our arithmetic.
The filings show three new share classes sold at three prices. The highest, Series C-3, was priced at ยฃ1,452.35 a share, about 1.8 times the ยฃ801.54 paid for Series C-1, by our arithmetic.
Short explainer · Fractile Series C
A vertical news Short narrated from the headlines and opening lines of this piece and three related pieces. It goes public at 12:00 UTC on 30 October.
Business Pill · The Memory Wall
A short explainer on the memory wall: why moving data, not computing it, often limits chips. It teaches the general idea only and says nothing about any company in this piece.
The key insight: By our sum of nine Companies House returns, Fractile issued about £577 million of Series C shares in six weeks, and it sold the same round at three prices, with the top class at £1,452.35 a share, about 1.8 times the £801.54 paid for the first class, by our arithmetic.
Fractile’s SH01 returns, the forms UK companies file when they issue shares, list the allotments. On 11 September it issued 184,412 C-1 preferred shares at ยฃ801.54, 52,877 C-2 shares at ยฃ707.25 and 182,681 C-3 shares at ยฃ1,452.35, all paid in cash.
Further C-3 allotments are dated 9 September and 6, 7, 8, 16 and 23 October. By our arithmetic, the C-1 shares raised about ยฃ147.8 million, the C-2 shares about ยฃ37.4 million and the C-3 shares about ยฃ392.2 million, for a total of about ยฃ577.4 million. Tech Funding News reported the same three class totals.
A resolution filed on 12 October records the “Creation of series c-1, series c-2 and series c-3 preferred share classes”. Fractile’s latest statement of capital, dated 23 October, lists 2,644,736 shares in issue.
Fractile’s Series C shares issued between 9 September and 23 October 2026 by class: C-3 about £392.2 million at £1,452.35 a share, C-1 about £147.8 million at £801.54 and C-2 about £37.4 million at £707.25. Class totals are our arithmetic from nine SH01 returns. Source: Companies House.
How the Round Compares
Our May piece reported that Fractile raised $220 million to build silicon designed for inference. Tech Funding News converted the new total at about $1.34 per pound, giving around $774 million, which is about 3.5 times the May round by our arithmetic.
Bloomberg reported in August, in a headline, that Fractile was seeking a $6.5 billion valuation after an Anthropic deal. Tech Funding News says “Anthropic has placed an initial $250m chip order, according to Bloomberg”. Dealroom reported the Series C at ยฃ540 million, lower than the filings’ total.
By our arithmetic, the 2,508,843 issued shares other than deferred shares would be worth about ยฃ3.64 billion at the C-3 price. That is not a reported valuation: it leaves out any unissued option pool, and preferred classes carry different rights.
As we read the filings, the round came in waves: about £450 million of shares were issued on 11 September, by our arithmetic, and the rest in smaller C-3 tranches through 23 October.
As we read the share prices, the round was split into classes at different prices: the filings date C-2 shares at £707.25 and C-3 shares at £1,452.35 to the same day, 11 September, and do not say who bought which class.
As we read Tech Funding News, the round funds a chip that is not yet in production: it says Fractile expects to deploy in data centres in 2027.
Tech Funding News, 30 October 2026
“Fractile has not shipped a production chip yet”
Three Implications
FOR FRACTILE Its latest statement of capital lists 2,644,736 shares in issue, including 135,893 deferred shares, according to the 23 October return.
FOR NVIDIA’S CHALLENGERS Tech Funding News says Anthropic has placed an initial $250m chip order with Fractile, according to Bloomberg.
FOR INVESTORS By our arithmetic, the issued shares other than deferred shares would be worth about £3.64 billion at the C-3 price; that figure is not a reported valuation.
The analysis puts it this way: “While attention remains fixed on GPU shortages and frontier model capabilities, the binding constraint on AI scaling has quietly shifted to High Bandwidth Memory (HBM).”
As we read it, Fractile’s pitch sits on that constraint: our October piece covered its case that FLOPs are the wrong metric. The analysis does not discuss Fractile or its Series C.
What Is Not Established
The SH01 returns give amounts paid, not investor names or a valuation. The class totals, the 3.5x and 1.8x ratios and the ยฃ3.64 billion figure are our arithmetic. The dollar figure uses Tech Funding News’s exchange rate. The Anthropic order and the $6.5 billion target come from Bloomberg as relayed in headlines and by Tech Funding News.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Companies House filings show Fractile issued about ยฃ577 million of Series C shares between 9 September and 23 October, by our sum of nine returns, at prices from ยฃ707.25 to ยฃ1,452.35 a share. Tech Funding News puts that at around $774 million, about 3.5 times the $220 million round Fractile raised in May, by our arithmetic.
A note on sourcing. The share classes, prices, allotments and share count come from Fractile Ltd’s SH01 returns and filing history at Companies House, which we read directly. The dollar conversion, the Anthropic order and the deployment timing come from Tech Funding News’s report of 30 October 2026; Dealroom and Bloomberg are cited by headline. The class totals and ratios are our arithmetic. We did not contact Fractile.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
