The speed of it is the strange part. A company founded by a robotics PhD student in 2022 was valued at about US$1 billion after a US$220 million round, and then, within about three months, reported to be raising at US$6.5 billion.
Walter Goodwin was the student. He worked at the University of Oxford's Robotics Institute when he started Fractile, a chip company with a deliberately narrow target. He was not trying to out-train the biggest models. His bet was about what happens afterwards, when people actually use them.
He worked at the University of Oxford's Robotics Institute when he started Fractile, a chip company with a deliberately narrow target.
What almost nobody outside the chip world noticed was a plain observation. Running a trained AI model, known as inference, is limited by how fast data moves between memory and compute. Fractile's design puts compute and memory on the same die using SRAM, rather than fetching data from separate memory. It is a bet that the bottleneck is not arithmetic but waiting.
That is also a risky bet for a young founder. Chips take years and large sums to prove, and public sources do not describe Goodwin's own doubts, family background or early setbacks, so none of that is claimed here. What is documented is the money: a small early round of roughly US$15 million, then a US$220 million Series B from Accel, Factorial Funds and Founders Fund.
The turn came in 2026 with a US$250 million chip deal to supply inference hardware to Anthropic. Reports in August said Fractile was in advanced talks to raise at US$6.5 billion, and a later aggregator listing put the October Series C at US$7.1 billion. These figures come from reports and one aggregator, and Dealroom shows a lower range of US$2.5 to 5 billion.
The price is the part founders should read carefully. The chips do not ship until 2027, and the contract is still in its initial phase. A valuation this high is a promise, not a result. Goodwin now has to make silicon that does what the slides said.
The lesson is not that a student can become a billionaire-adjacent founder in a season. It is that a narrow, technical claim about where the real bottleneck sits can be worth more than a broad one, if one customer believes it enough to sign. After that, the hard part starts.