Gravis Robotics has raised $200 million in Series A funding from SoftBank, giving the Zurich-based construction robotics company substantial new capital to expand its autonomous heavy machinery technology across global infrastructure projects.
The company describes the financing as the largest Series A in construction robotics to date. The round comes as investment in physical AI increasingly extends beyond humanoid robots and warehouse automation into industries where machines must interact with complex, changing environments.
For Gravis, that environment is the construction site. Rather than building new heavy equipment from scratch, the company develops hardware and AI software designed to retrofit existing excavators and other earthmoving machinery, effectively adding varying degrees of autonomous operation to fleets contractors already own.
The new funding will be used to accelerate Gravis’ international rollout, expand its engineering team, and put its autonomous systems into more construction and infrastructure projects.
Bringing Physical AI to Heavy Construction
Founded in late 2022 as a spinout from ETH Zurich, Gravis emerged from research into robotics and autonomous control systems. The company is led by CEO and co-founder Ryan Luke Johns and CTO and co-founder Dominic Jud, with robotics researcher Marco Hutter serving as a co-founder and board member.
Construction presents a different challenge from many of the environments where autonomous machines have gained traction.
A self-driving vehicle, for example, generally attempts to understand and safely navigate an environment without changing it. An excavator is expected to do the opposite. Every bucket movement alters the terrain the machine needs to understand next.
The machine may also encounter soil, rocks, changing slopes, underground resistance and other conditions that cannot be fully predicted in advance.
Gravis is attempting to address that uncertainty through learning-based robotic control systems that combine data from machine hydraulics with LiDAR, cameras and Global Navigation Satellite System (GNSS) positioning. Its software continuously interprets the machinery’s surroundings and the physical forces affecting it rather than simply executing a predetermined sequence of movements.
The company says its models are also trained extensively in simulation, allowing the AI to encounter large numbers of virtual excavation scenarios before being deployed to physical machinery.
Turning Existing Excavators Into Autonomous Machines
A central part of Gravis’ strategy is avoiding dependence on a single equipment manufacturer.
Construction fleets are typically composed of machines from multiple brands, often accumulated over years of purchases and rental agreements. Asking contractors to replace those fleets with purpose-built autonomous machinery could make adoption considerably more difficult.
Instead, Gravis has developed the Gravis RACK, a modular autonomous control system that can be installed on existing excavators and wheel loaders. The company says the platform has already been adapted to equipment from manufacturers including Caterpillar, John Deere, JCB, Hitachi, Volvo, Yanmar, Case, Develon and Sumitomo.
The rooftop system combines cameras, 3D LiDAR, GNSS RTK positioning and automotive-grade edge computing. Because processing takes place on the machine, Gravis says autonomous functions can continue operating even when reliable connectivity is unavailable, an important consideration on remote or unfinished construction sites.
Full LiDAR coverage and sensor fusion are used to continuously scan surrounding terrain. That information can support autonomous excavation while simultaneously producing 3D site data, cut-and-fill visualizations and records of completed work.
Gravis’ perception system can also identify dump trucks and coordinate where excavated material should be placed, allowing autonomous workflows to extend beyond digging into activities such as loading and material handling.
Autonomy Without Removing the Operator
Gravis is not positioning autonomy as an all-or-nothing transition.
Its Slate tablet interface allows contractors to move between several operating modes depending on the task. Operators can remain inside the cab and use AI-assisted guidance, step away while the machine performs longer autonomous tasks, or supervise equipment remotely.
In its in-cab Copilot mode, operators can use tap-to-dig controls and augmented visual guidance. Jobs can be defined through Computer-Aided Design (CAD) or Building Information Modeling (BIM) geometry, physical reference points or coordinates. The interface can then visualize how much material needs to be removed or added while the machine works.
At the other end of the spectrum, remote orchestration allows an operator to supervise and remotely operate one or more machines using live video and site information.
That hybrid approach may prove important to adoption. Construction sites rarely consist of repetitive tasks performed under perfectly controlled conditions. Keeping humans within the operational loop gives contractors a way to introduce autonomy gradually rather than redesigning an entire jobsite around fully driverless equipment.
A Software Layer Across Mixed Construction Fleets
The broader ambition is effectively to create an intelligence layer that sits above the fragmented heavy-equipment market.
If the same autonomous software can operate machinery from numerous manufacturers and across multiple machine sizes, contractors would potentially be able to deploy automation without standardizing their entire fleet around one vendor.
Gravis says its learning-based control system adjusts to the physical characteristics of different machines rather than requiring each excavator to be programmed independently.
That distinction could become increasingly important as physical AI moves from demonstrations into commercial deployment. A robotic system that performs well on one carefully configured machine is considerably less useful to large contractors than software capable of adapting to the heterogeneous fleets already operating around the world.
Gravis says its terrain-aware excavation technology can increase throughput by as much as 30%, although actual gains will inevitably depend on the machinery, job and site conditions.
Moving From Research Projects to Active Jobsites
Gravis had already begun expanding commercially before the SoftBank investment.
In November 2025, the company announced $23 million in fresh funding alongside partnerships and deployments involving companies including Holcim, Taylor Woodrow, HD Hyundai and Flannery Plant Hire. At the time, Gravis said its systems were live in seven countries spanning the UK, European Union, United States, Latin America and Asia.
One notable deployment involved autonomous excavation on an active Taylor Woodrow infrastructure project at Manchester Airport. The company has also worked on autonomous quarry-material handling and established a partnership with Flannery designed to make excavators equipped with Gravis technology available through the equipment rental market.
That commercialization history helps distinguish the latest financing from physical-AI investments centered primarily on future prototypes. Gravis is raising capital while its systems are already being tested and used alongside conventional construction crews.
UK Project Will Put Autonomous Excavators Into a Larger Trial
The expansion is also receiving government support.
Gravis and Flannery Plant Hire were recently selected for a project under the UK’s CAM Pathfinder connected and automated mobility program. The initiative will trial a fully autonomous earthmoving system across six excavators, with applications including trenching, bulk excavation and truck loading.
The project is particularly relevant because equipment rental could become an important distribution channel for construction autonomy.
Rather than requiring contractors to purchase autonomous machines or permanently retrofit their own fleets, rental companies could provide autonomy-equipped machinery for individual projects. That could lower the financial and operational barriers to experimenting with the technology.
It would also expose autonomous systems to a much wider variety of jobsites, machinery configurations and ground conditions, potentially generating valuable operational data for improving the underlying AI models.
Why Construction Is Emerging as a Major Physical AI Market
Much of the attention surrounding physical AI has focused on humanoid robots. Heavy machinery represents another potentially significant opportunity, and one where the economic use case may be easier to define.
Excavators, loaders and other machines already perform enormous amounts of physical work. The challenge is not inventing a new mechanical form factor but adding intelligence to equipment that is already deployed at scale.
That creates a different route to commercialization. Instead of asking customers to determine what a new type of robot should do, companies such as Gravis are targeting tasks that contractors already pay skilled workers and expensive machinery to perform every day.
The remaining challenge is reliability. Construction environments are messy, dynamic and safety-critical, which means autonomous equipment needs to operate consistently across conditions that are far less predictable than factory floors or warehouses.
Gravis’ focus on simulation, machine telemetry, sensor fusion and adaptable control models reflects just how difficult that problem is.
SoftBank Funding Puts Gravis Into a New Phase of Expansion
The $200 million Series A gives Gravis considerably more resources to tackle that challenge.
Rather than simply funding further research, much of the opportunity now lies in deployment: installing systems across more equipment brands, accumulating experience from active construction projects and proving that autonomous machinery can deliver consistent economic benefits outside controlled demonstrations.
The retrofit strategy could be especially important. Construction companies have enormous amounts of capital tied up in existing machinery, and a platform capable of adding intelligence across those assets could scale differently from competitors requiring customers to purchase entirely new robotic fleets.
If that model proves reliable, construction may become one of physical AI’s more consequential markets. The machines are already there, the work is already defined, and demand for infrastructure continues to grow.
Gravis Robotics now has another $200 million to demonstrate that AI can do more than understand and navigate the physical world. It can help reshape it.
