Archetype AI was founded in 2023 by a team of scientists, designers and engineers from Google, several of them veterans of its Advanced Technology and Projects (ATAP) group. Chief executive Ivan Poupyrev led technology development at ATAP, and earlier at Disney and Sony; his co-founders include ATAP's former sensing lead, Jaime Lien, now chief scientist. The company is American and small — just over 40 people as of 2025, working in distributed fashion, but well capitalised for its size.
In November 2025 it closed a $35m Series A led by IAG Capital Partners and Hitachi Ventures, with participation from Bezos Expeditions, Venrock, the Amazon Industrial Innovation Fund, Samsung and others. Its premise is what it calls physical AI: a foundation model, named Newton, trained to interpret the raw signals thrown off by the physical world rather than text or images.
Newton ingests streams from sensors already installed on industrial sites, from radar and video to accelerometers and temperature probes, and fuses them with natural language, so that an operator can ask what is happening on an asset without writing bespoke detection code for each camera or vibration sensor. On top of the model sit packaged applications the company calls Lenses, launched with three partners in March 2025. Kajima, the Japanese contractor, used a Lens to process five terabytes of video from 27 cameras spanning four years of construction work, tracking progress against programme.
The City of Bellevue, Washington, working with the integrator Khasm Labs, deployed a pedestrian-safety Lens that processes around 43 gigabytes of camera data per intersection per day, flags incidents in real time and can adjust signal timing. In November 2025 the company added a platform and an Agent Toolkit for customers to build their own "physical agents", naming NTT DATA as a further early customer. What is proven is that named organisations run Newton on live assets. What remains a company claim is generality: the disclosed metrics describe data volumes processed rather than outcomes such as collisions avoided or delays prevented, and the Bellevue deployment needed an integration partner to stitch the model into the city's existing sensors.
The launch deployments map squarely onto problems held by UK and Irish public bodies and utilities. Bellevue's intersection monitoring is the same task facing councils pursuing road-safety targets with ageing CCTV estates; Kajima's progress-tracking Lens addresses the cost and schedule overruns endemic to large British construction programmes.
Water companies, electricity network operators and rail infrastructure owners hold exactly the sensor-rich, analysis-poor asset bases that a retrofit model like Newton is aimed at — instrumented for years, interpreted rarely. The presence of Hitachi Ventures on the cap table is worth noting given the Hitachi group's industrial interests in Britain. As yet, however, no UK or Irish deployment has been announced; the relevance is structural, not contractual.
This company is on the radar for the shape of the problem it addresses, not for evidenced UK activity — none has been verified. The confirming signal would be a named pilot with a UK council, utility or tier-one contractor, or the publication of outcome data from the existing deployments: incident reductions in Bellevue, programme savings at Kajima, rather than terabytes processed.
The undermining signals would be launch partners failing to expand beyond pilots, or evidence that each deployment requires the heavy bespoke integration that Bellevue's did, which would make Newton look less like a foundation model and more like a well-tooled consultancy. The investor list, heavy with industrial strategics, suggests where the next proof points are likely to surface.
Sources: archetypeai.io · silicon.co.uk · archetypeai.io
Entries are verified against primary sources. The bar is the frontier: first-of-kind, new technology, new materials — delivery milestones of conventional infrastructure, however large, do not qualify. The frontier is tracked globally; every entry states what it means for UK and Irish infrastructure.
Verified by GridteamAI.
