Gravis Robotics, a Zurich company that retrofits existing excavators with autonomy hardware, announced on 17 August that it has raised $200 million in a Series A from SoftBank. SoftBank is the only investor named in the round. Dealroom and Forbes both put the post-money valuation at roughly $1 billion; Gravis’s own announcement does not state a valuation, and the company has not confirmed one publicly that we could find.
The number that makes this interesting is not $200 million. It is $23 million — the round Gravis closed in November 2025, co-led by IQ Capital and Zacua Ventures with Pear VC, Sunna Ventures, Armada Investment, Imad Ventures and Holcim participating. Nine months later a single investor put in nearly nine times that amount. Gravis describes the round as the largest Series A in construction robotics; we have not independently verified that against a complete dataset, and it is the company’s characterisation rather than ours.
What Gravis Actually Sells
The product is a retrofit kit. The Gravis Rack — the company calls it “the brain and backbone of our autonomous excavators” — is sensors, compute and controls that bolt onto a machine a contractor already owns. Alongside it sit Gravis Copilot, an assisted-operation mode with real-time guidance and hazard detection, and Slate, the worksite interface.
Per the company’s announcement, the system has been installed on machines from Caterpillar, Case, Develon, John Deere, JCB, Hitachi, Sumitomo, Yanmar and Volvo. That list is the whole strategic argument in one line: the fleet already exists, it is mixed across brands on almost any sizeable job, and nobody is going to replace it.
The headline performance claim is “up to a 30% boost in jobsite productivity compared to peak manual operation.” That is an unaudited company figure, and “up to” against “peak manual operation” is doing a lot of work — the comparison is against a good operator having a good day, which is the hardest baseline to beat and also the one least representative of a full shift.
Gravis’s website lists armasuisse, Boskalis, Holcim, Morgan Sindall and Taylor Woodrow among the organisations it works with. The company also leads an $8 million UK government-backed CAM Pathfinder project with Flannery Plant Hire, which Gravis describes as the UK’s largest operated heavy-equipment rental provider.
The Two Strategies Are Now Both Funded
Gravis was founded in 2022 as a spinout of ETH Zurich by CEO Ryan Luke Johns and CTO Dominic Jud, with offices in Zurich, Austin and Oxford. “Our machines are built for the messy, unscripted reality of live jobsites that breaks traditional automation,” Johns said in the announcement. Dai Sakata, a SoftBank managing director, framed the cheque in portfolio terms: “Physical AI is central to SoftBank’s vision for the next phase of AI.”
The timing is the part worth sitting with. Gravis’s round landed the same day Bedrock Robotics announced its own excavators running fully autonomously on live customer sites. Bedrock has raised $350 million and is chasing large earthmoving contractors with a full-stack system. Gravis has disclosed $223 million across its two announced rounds and is aiming, per Johns’s comments to Construction Dive, at small and mid-sized contractors with something priced “competitive with premium machine control systems.”
That is not the same market, and it is not the same bet. One says autonomy is a capability large contractors will buy as a service on big dirt jobs. The other says autonomy is an option package — closer to GPS grade control than to a robotaxi — that any contractor with a twenty-tonne excavator can tick.
Both now have enough capital to be wrong for several years.
What the Announcement Does Not Say
Gravis has not disclosed how many machines are running its kit, how many hours they have logged, or what share of that work happens without a human in the cab. Its own product framing spans “AI-assisted operation” through to full autonomy, which is honest — but it also means “deployed across four continents” is compatible with a very wide range of realities.
This newsroom wrote in June that Gravis’s commercial footprint looked early, with deployments proving the concept rather than demonstrating fleet-scale operation. Nothing in the 17 August announcement changes that assessment. A $200 million Series A is a statement about the next three years, not a disclosure about the last three.
The retrofit thesis itself is not novel — Xpanner is selling automation licences for existing hardware and TerraFirma raised $115 million for remote operation on adjacent logic. What has changed is the size of the cheque behind it.
What to Watch
Whether a single investor is an advantage or an exposure. A $200 million Series A from one balance sheet means no syndicate to price the next round against, and a company whose fortunes are now tied to one firm’s appetite for physical AI. SoftBank’s conviction in a category has historically been directional rather than gentle.
Whether the SMB pitch survives contact with servicing. Selling to small and mid-sized contractors means selling into fleets that are older, more varied and further from a dealer network. A retrofit kit that needs a specialist to install and calibrate is a different business at 50 customers than at 5,000.
Whether the productivity claim ever gets a public baseline. “Up to 30% versus peak manual operation” is a marketing number until someone publishes cycle times from a real job with the machine and against it. The first contractor willing to do that will be more persuasive than the round.
