HIVE bolts a 'silicon brain' onto existing heavy machines instead of building new robots. Vikafjillett mountain road, Yara and Norwegian road authority Statens Vegvesen test one shared autonomy layer across equipment.
On a Norwegian mountain road at Vikafjillett, a wheel loader (the heavy front-bucket tractor used on avalanche and road work) pushes debris off the pass. The cab is empty. The operator watching the work sits miles away in a control room.
HIVE, the London-headquartered physical AI startup, built its pitch around that scene. The company announced a $15M (€13.1M) pre-Series A round on 7 July 2026, with offices in Norway and active expansion into the United States. HIVE describes its product as a 'silicon brain': a sensor, compute, and software stack that bolts onto existing industrial machines so they can perceive, decide, and act autonomously under remote supervision.
'Physical AI for industrial machines' is shorthand for one of two architectural bets. The louder one builds new robots from the chassis up: humanoid pilots from Figure, Agility and 1X, plus industrial quadrupeds from Boston Dynamics and its peers. HIVE is making the quieter wager. Keep the machine. Retrofit the autonomy. A heavy wheel loader already does the work; bolting on sensors and compute, then supervising from a control room, costs a fraction of replacing the machine itself. The second wager is on cross-machine reuse: one shared autonomy layer can supervise a wheel loader on a mountain, a fertilizer spreader in a Yara field, and a road-maintenance vehicle on a Norwegian highway, rather than training a new model per machine class.
Vikafjillett runs for Presis Vegdrift, the contractor working for Statens Vegvesen, the Norwegian road authority. A separate partnership with Yara, the Norwegian fertilizer and crop-nutrition major, covers field operations. CEO Christoffer Jørgensvaag told Sifted that talent build-out and live deployments are the company's proof points of traction.
The number HIVE is asking investors to underwrite is an 80% reduction in productive machine-hour cost. That is a company-stated projection, not an independently audited figure, and it is the metric that, if verified, separates the retrofit thesis from the rest of the physical AI field. It is also the claim most likely to break if 'silicon brain' requires dedicated engineers at every site rather than a shared autonomy layer that travels with the machine.
The retrofit-versus-build-new debate in physical AI has been largely theoretical until the past year. HIVE's cross-sector deployments are the first operational evidence that one shared autonomy layer works across heterogeneous equipment, but the evidence is also thin. Two of the three named deployments, avalanche clearance and field agriculture, are outdoor, slow-speed and GPS-rich environments. Warehouses and production lines, the third leg of the pitch, are indoor, fast-moving and dominated by existing automation vendors. The premise that learns on a mountain pass has to transfer to a fulfillment center without rebuilding the model.
If the retrofit thesis holds, the implication for industrial operators is larger than the round suggests. Capital-allocation playbooks for industrial fleets assume a 10- to 15-year replacement cycle. A retrofit that adds autonomy without scrapping the machine extends that cycle, lowers the breakeven on autonomous operation, and lets a buyer spread autonomy spend across the existing fleet rather than committing capex to a new chassis category. For customers like Statens Vegvesen or Yara, operators with hundreds of similar machines in similar roles, that is the more important number than any funding round.
Three things to watch over the next two quarters: independent verification of the 80% productive machine-hour claim, with the next standard being a third-party benchmark against a comparable, human-operated control site; a second cross-sector deployment beyond Vikafjillett and Yara, because the breadth-versus-depth claim is load-bearing; and whether the company discloses an OEM or industrial-fleet partner for the U.S. expansion (EU-Startups and Sifted both cover the round but leave both questions open).