This is SoftBank putting physical AI to work on real job sites, because Gravis Robotics is building autonomy for heavy equipment and the cheque is big enough to force a global rollout.
SoftBank has agreed to invest EUR 200 million (reported as $200 million) in Gravis Robotics in a Series A round, according to the company and multiple media reports. Gravis, a Zurich-based startup and ETH Zurich spinout founded in 2022, develops software and hardware for earthmoving and construction machinery, positioning itself at the intersection of robotics, AI controls and infrastructure execution.
Several outlets described the financing as the largest Series A in construction robotics history, and the scale stands out even within a hot ConTech funding cycle. Reporting also noted that Gravis had previously raised USD 23 million in November 2025, making the jump to a USD 200 million Series A unusually steep for a company operating in a capital-intensive, field-deployment-heavy category.
Why SoftBank is leaning in
SoftBank framed the investment as part of its broader AI strategy, with the group pointing to Gravis’ AI-based machine controls. In industry coverage, SoftBank executive Dai Sakata described physical AI as central to its next phase of AI. That matters: this is not a generic “AI for construction” bet, but an attempt to move autonomy from demos to fleet-level deployments where utilisation, safety and productivity can be measured.
Gravis itself said the capital will support global rollout and fleet deployments, which is where construction autonomy tends to succeed or fail. The technology has to operate across variable sites, operators, machines, and regulatory regimes. Funding rounds at this size are effectively a commitment to solve those messy last-mile problems, not just improve models.
A with-trend deal in ConTech, but at an outlier size
The round lands in a market that has been tilting hard toward AI-enabled construction solutions. Cemex Ventures reported that in 2025, 77% of ConTech capital went to AI-enabled solutions, and that robotics attracted USD 476 million that year. Memoori reported that AI construction startups raised USD 616 million in H1 2026, nearly double the prior year’s full-year total.
Against that backdrop, Gravis looks less like a one-off and more like an aggressive scaling play within a well-funded theme: investors want AI that touches physical productivity, not only software workflows.
Switzerland’s deep-tech pipeline gets another proof point
Gravis’ ETH Zurich roots reinforce Switzerland’s role as a deep-tech formation engine, particularly in robotics and applied AI. Reports also positioned Gravis within a broader Swiss construction-robotics and physical-AI ecosystem, with Zurich again serving as the base for a major robotics funding event.
Early commercial traction will be scrutinised as the company expands. The Japan Times reported that Gravis has been selected for an USD 8 million UK government infrastructure retrofit project, a signal that its approach can travel beyond Switzerland and into large public-sector procurement environments.
Execution risks: deployment reality beats model quality
The strategic logic is clear, but construction autonomy is unforgiving:
- Integration and interoperability: Autonomy layers must work across different OEMs, machine configurations and retrofit constraints.
- Safety and liability: Any autonomy claims have to survive real-world incident scrutiny, not just controlled pilots.
- Scaling fleets: The hard part is repeatable deployment, maintenance and operator adoption across multiple sites.
SoftBank’s capital gives Gravis runway to industrialise deployment, but it also raises expectations. With a Series A of this magnitude, the market will look for evidence of repeatable unit economics and a credible path to multi-region scaling, not just technical milestones.
For European robotics, the message is simpler: investors are still willing to underwrite physical AI at scale, but only for platforms that can plausibly move from prototypes to machines in the dirt.