· investment-strategies · 3 min read
Gravis Robotics Raises $200M Series A from SoftBank — Retrofit Excavator Autonomy
SoftBank is the sole investor in Gravis’s $200M Series A (Aug 17, 2026). ETH Zurich spinout sells a mixed-fleet autonomy kit — physical AI for earthmoving, not a new OEM.
Gravis Robotics announced a $200 million Series A on August 17, 2026. SoftBank is the sole investor. Company: largest Series A in construction robotics. HQ: Zurich (ETH spinout), with US/UK presence.
Key facts
| Field | Detail |
|---|---|
| Company | Gravis Robotics — physical AI for heavy construction |
| Round | $200M Series A · SoftBank sole |
| Date | August 17, 2026 |
| Prior | ~$23M (Nov 2025; IQ Capital / Zacua et al. — press) |
| Product | Gravis Rack kit + Copilot / full autonomy; mixed OEM fleets |
| Named OEMs (company) | Caterpillar, Case, Develon, John Deere, JCB, Hitachi, Sumitomo, Yanmar, Volvo, more |
| Traction (company) | Deployments on four continents; UK CAM Pathfinder (£/$ mix — company says $8M UK government-backed project with Flannery Plant Hire) |
| Productivity claim | Up to 30% vs peak manual — company |
Who uses the product — and for what job
Users: contractors and rental houses that already own mixed excavator fleets and cannot scrap iron to buy a robot brand.
Job: move earth for housing, grids, and data centers when skilled operators are scarce — without waiting for a single-OEM autonomous lineup.
Most physical AI navigates a static world. Gravis’s pitch: excavators destroy and reshape terrain. Software has to model soil and hydraulics, not just lanes.
Modes: operator in cab with 3D copilot, or supervise a fleet remotely. Every machine is also a site sensor.
Why now
- AI campuses and energy projects are earthmoving-constrained. That is the same bottleneck Hadrian sells from the factory side.
- SoftBank has been vocal on physical AI. Dai Sakata (SoftBank Group) is quoted on the company page.
- Retrofit beats rip-and-replace in a fragmented OEM market (company: ~two-thirds of demand outside the top three manufacturers).
Why SoftBank — portfolio fit
SoftBank writing the entire $200M Series A is concentrated governance. Founders get speed and a global LP; they give up syndicate optionality.
SoftBank is not in our fund directory — no /fund/softbank. Do not invent it.
Press (TFN) noted earlier acquisition chatter (Bloomberg-sourced, sub-$500M talk). The announced structure is primary $200M, not a confirmed buyout. Flag as unverified unless SoftBank/Gravis say otherwise.
Likely founder rationale: one check large enough to hire and kit fleets globally; OEM-agnostic story needs capital, not another $20M seed.
| Investor type | What they bring |
|---|---|
| SoftBank (sole) | $200M, physical-AI narrative, follow-on capacity — and concentrated control |
| Seed holders (IQ Capital, Pear VC, Holcim MAQER, etc.) | Diluted but still on cap table (LinkedIn/Inc.; not restated on Series A page) |
Competitive map
| Player | Lane |
|---|---|
| OEM autonomy (Cat, Volvo, etc.) | Closed fleets |
| Bedrock Robotics | Autonomous earthmovers / haulers (different stack) |
| Hadrian | Factories, not jobsites |
| Traditional machine control (Trimble, etc.) | Guidance, not full autonomy kits |
Market signal
A $200M Series A with one investor is a deployment round. The 30% productivity number will be diligence or marketing depending on whether independent jobsites replicate it.
$1B post-money is widely reported (Inc./TFN) but absent from Gravis’s own post. We do not treat it as company-confirmed.
When not to use this as a template
- Wrong if you need a diversified Series A syndicate as a governance model.
- Wrong if the product only works on one OEM.
- Wrong if you quote $1B without labeling it as press-reported.
Practical takeaway
- Founders (physical AI): Sell kit-on-installed-base, not a new yellow machine.
- Investors: Underwrite safety cases, union/jobsite process, and OEM politics.
- Operators / GCs: Ask for hours of autonomous production on your soil type, not a Zurich demo.