Helm.ai Reports $70 Million in Signed Contracts Over 12 Months and Names No Customer
Hardware / news
Helm.ai Reports $70 Million in Signed Contracts Over 12 Months and Names No Customer
The Redwood City driving-software company says it is on a path to break even. It has published no revenue figure, no customer list and no contract sizes.
Helm.ai said it has signed $70 million in commercial contracts for its physical-AI foundation models over a 12-month period, according to The Robot Report on Oct. 8. The Redwood City, California company named none of the buyers and gave no revenue figure.
The number is vendor-supplied. Nothing in the report separates what has been recognised as revenue from what is committed for later.
What the company has said
The customers are described as global automotive OEMs, Tier 1 suppliers and industrial automation companies. Automotive is the flagship market, with the software spanning SAE Level 2 through Level 4 programs. Helm.ai also sells to mining and construction customers, and The Robot Report says its perception software is on a production track for heavy industry.
Chief executive and founder Vladislav Voroninski said "we're now on a path to break even, which is rare in this space." On the maturing programs he said, "Those kinds of projects coming to that level of maturity are super exciting for us." The report gives no break-even date.
The scale of the figure
Spread evenly over 12 months, $70 million is about $5.8 million a month. Contracts are rarely paid evenly, and a multi-year development deal signed in the period would count in full in a signed-contract figure while paying out over several years. That is arithmetic on the reported total, not a statement about Helm.ai's books.
The most recent priced round found in this reporting is a Series C of $55 million on Aug. 17, 2023, led by Freeman Group, per Maginative. ACVC Partners and Amplo took part, as did strategic investors Honda Motor, Goodyear Ventures and Sungwoo Hitech. The valuation was not disclosed. The 12-month contract total is therefore $15 million above the entire 2023 round, which is a comparison of two different kinds of number: one is cash received, the other a promise to pay.
Helm.ai's newsroom listing also carries an Oct. 14, 2025 press release about an additional Honda investment. The listing gives no dollar amount, and the contract total is not tied to any named investor or customer.
How the technology is described
The company says its models are trained with an unsupervised method it calls deep teaching, which separates understanding an environment from acting in it. It says the same perception stack works across driving, open-pit mining and other industrial settings, and across sensor arrangements for vehicles, drones and humanoids. These are company claims. The report cites no benchmark and no independent test.
That breadth is what the robotics side of the site's coverage keeps returning to. A system that generalises across sensors still has to survive the per-action arithmetic in why 99 percent per-action success finishes 37 percent of jobs. Safety validation is the subject of the SafeWorld seed round.
| Item | Reported | Disclosed? |
|---|---|---|
| Signed contracts, 12 months | $70 million | Total only |
| Customers | OEMs, Tier 1s, industrial automation | No names |
| Break-even | Path to, per the CEO | No date |
| Last priced round | $55 million Series C, Aug. 2023 | No valuation |
What to watch
Three disclosures would turn the figure into something checkable: a named production program with a start-of-production date, the share of the $70 million that is already invoiced, and the length of the contracts behind it. Helm.ai has given none of the three, and the report does not say when it will.
Sources
More in Hardware
- 01Two Yandex Data Centres in Two Days: What the Sasovo and Kaluga Strikes Took OfflineReuters reports 58,268 user complaints and a 3% share drop after drones hit a site that Yandex says holds two of its three AI supercomputers.
- 02Mitsubishi's ASX VR-e Puts 342 kW on a 57.7 kWh Battery, and Has Not Said How Far It GoesThe Foxtron-built GSR claims 3.9 seconds to 100 km/h, but Mitsubishi has published neither a range nor a price for the Australian launch.
- 03AWS Physical AI Toolchain Prices a Robot Training Run at $79, but Its Edge Step Is Still PlannedThe Apache-2.0 repo lists per-run cost estimates from $2 to $93 and Cosmos 3 at $37 an hour. Its README says edge deployment is not available yet, though the launch blog lists it as a stage.
- 04Why a Robot With 99 Percent Per-Action Success Finishes Only 37 Percent of JobsTDK Ventures president Nicolas Sauvage lays out the compounding arithmetic behind humanoid demos, and the figures a buyer should ask for instead.