Realtime AI News
Britain's Fractile raises $720M to push purpose-built AI chips
Fractile, a British startup building chips for AI, has raised a $720 million round, according to TechCrunch, making it one of the best-funded players in the race for purpose-built AI silicon. The size of the raise shows capital is still willing to bet on a path other than Nvidia.

Fractile, a British startup building chips for AI, has raised a $720 million round, according to TechCrunch, making it one of the best-funded players in the race for purpose-built AI silicon.
"Purpose-built" is the operative phrase. Unlike general-purpose GPUs, these chips are designed around specific computational patterns in an attempt to win on energy efficiency or cost, at the price of a narrower addressable workload and a thinner software ecosystem.
The size of the round suggests capital is still willing to pay for an alternative to Nvidia, even though the engineering and manufacturing risks on that path are steep.
TechCrunch places Fractile in Britain, a notable position in an AI chip map otherwise dominated by the United States and China, and one that gives the company a natural role in the story of European compute supply.
Chip companies spend money in two places: long-running design teams, and the slow validation from tape-out to volume production. The first scales with people, the second with time and yield, and neither can be bought outright with a single round.
$720 million is unlikely to build a fabrication plant of its own, but it can fund several generations of product and early customer deployments, which is the typical capital requirement at this stage.
The next test is whether the money turns into deliverable silicon, and whether performance on the first real workloads supports the argument that specialized beats general.
Why it matters
The raise puts Europe's AI chip story back in the foreground, and shifts competition toward deliverable silicon and measured performance rather than the size of funding rounds.
Nearby Updates
All11/12, 03:41
Runway open-sources Praxis-1, claiming 1.65x π0.5 on Nvidia's test
Runway has open-sourced Praxis-1, a model that reaches 1.65x the performance of π0.5 on Nvidia's test, according to FourWeekMBA. Using a widely cited robot foundation model as the reference point places the release squarely in the embodied AI comparison set.
11/12, 03:44
Google launches Gemini for healthcare, ten months behind OpenAI and Anthropic
Google has launched a Gemini offering aimed at healthcare, roughly ten months after OpenAI and Anthropic brought their own healthcare products to market, according to Techish Kenya. The gap in timing is itself the story the report tells.
11/12, 03:54
CoreWeave turns to customer prepayments and shorter contracts to fund AI growth
According to its third-quarter earnings, CoreWeave is increasingly asking customers to pay upfront, borrowing against specific contracts, and weighing shorter-term deals. TechCrunch reports that the arrangements fund expansion but could also push prices higher for customers.
11/12, 02:56
Tether's 800M-parameter AI agent reportedly beats models 2.5x its size on phone tasks
Cryptonews.net reports that an AI agent from Tether with roughly 800 million parameters outperformed models about 2.5 times larger on everyday phone tasks. The interesting part is not the absolute score but whether small on-device models can be genuinely useful assistants.