Gravis Raises $200M to Retrofit Excavators, Not Replace Fleets

In its August 17, 2026 funding announcement, Gravis Robotics said SoftBank is investing $200 million through a Series A to expand autonomous heavy machinery across construction sites worldwide. The capital is intended to support commercial deployment of the Swiss company’s control technology, rather than the development of a new excavator.

The central proposition is to add autonomy to equipment contractors already own or rent, not force them to replace entire fleets. Construction Dive’s August 18 coverage describes the Gravis Rack as a physical retrofit that sits on heavy machinery and controls equipment from established manufacturers, alongside Gravis Copilot for AI-assisted manual operation.

Retrofitting could shorten the deployment cycle

Construction equipment is a long-lived capital asset tied to attachments, maintenance arrangements, dealer support and operator familiarity. Installing a separate control layer allows a contractor to retain that mechanical investment while adding assistance or autonomy without waiting for the next fleet-replacement cycle.

The Gravis Rack is designed as the hardware interface between the company’s software and machines from multiple manufacturers. That mixed-fleet approach matters because contractors often operate equipment from several brands; an autonomy system tied to one new model would cover only part of the fleet and could require a different procurement decision.

Retrofit does not mean universal or immediate compatibility. Each deployment still depends on the machine’s controls, the installed sensors and computing equipment, the attachment being used, the defined task and site-specific safety procedures. Publicly available material does not provide a model-by-model compatibility list, installation time or standard price.

The stack combines sensing, machine data and AI control

ETH Zurich’s technical account describes cameras and sensors scanning the surroundings, software building a three-dimensional terrain map and the controller analysing machine data such as engine load; it also identifies digging, levelling and loading lorries as autonomous applications and says the technology was operating on dozens of commercial sites when the funding was disclosed.

The sensing layer establishes the machine’s surroundings and the changing shape of the work area. Machine telemetry supplies a second type of information: how the excavator itself is responding as the bucket meets soil, rock or resistance. The control software uses both inputs to update movements as the work changes the terrain.

Gravis separates three operating modes that should not be treated as equivalent. Copilot keeps an operator in the cab and provides terrain guidance and hazard detection; remote operation moves direct control outside the cab; autonomous mode executes a prescribed machine task without continuous input at the controls.

That distinction narrows what “autonomous excavator” means in current deployments. The system can take responsibility for a defined operation with established targets and boundaries, but the available evidence does not show a machine independently planning and managing every activity on a changing construction site.

Digging, grading and loading are the clearest applications

The best-supported applications are bounded earthmoving jobs: trenching, bulk excavation, grading or levelling, managing material piles and loading trucks. These operations have measurable target surfaces, paths or repeated cycles, making them more suitable for automation than work that requires continual improvisation around crews, utilities and moving equipment.

A March 3, 2026 Hitachi Construction Machinery demonstration plan detailed a retrofit-equipped ZX135US-7 learning a trench from operator inputs and then executing the sequence autonomously; it also listed augmented terrain and mapped-utility visualisation, precision grading, people detection, repetitive-task autonomy and remote operation.

The demonstration illustrates a practical division of labour. An operator can perform complicated or judgment-heavy work directly, establish a site plan or excavation path, and then assign the repetitive portion to the machine. Human supervision remains relevant when ground conditions, boundaries or nearby activity depart from the defined operating plan.

SoftBank’s money is aimed at commercial scale

The funding is directed toward hiring, international expansion and putting more equipped machines into commercial use. Scaling a retrofit platform requires more than duplicating hardware: Gravis must adapt and support the system across machine configurations, attachments, site conditions and regional distribution channels.

Inc.’s August 17 transaction report identifies SoftBank as the sole investor, places Gravis’s post-money valuation at $1 billion and describes plans to expand hiring, international operations and commercial deployments, including a rental arrangement through which contractors can hire excavators already fitted with the Gravis Rack.

Rental distribution could remove a second adoption barrier. Contractors would not necessarily need to purchase either a new excavator or the autonomy system outright, although the economic case will still depend on rental pricing, installation and support requirements, utilisation, reliability and the amount of supervision required on real sites.

Broader performance evidence is still missing

The disclosed material establishes the investment, the retrofit architecture and several assisted or autonomous earthmoving tasks. It does not provide audited fleet-wide productivity or safety results, standard installation schedules, pricing, downtime records or a comprehensive list of compatible machines and attachments.

Meaningful comparisons with manual operation would also need to hold machine size, operator experience, soil, weather and job design reasonably constant. Setup, supervision, rework and maintenance can alter the result even if an autonomous digging cycle performs well in isolation.

As of August 24, the confirmed development is therefore substantial but specific: SoftBank has financed Gravis’s effort to deploy autonomous controls across more existing construction machinery. Digging, grading, levelling and truck loading are supported applications; the next evidence needed is how reliably and economically the system performs across a wider range of machines, attachments and active jobsites.

Also read:

• Tufts Report: 9.3 Million US Jobs at Risk from AI in the Next 5 Years — And the Hits Are Coming for High-Tech Hubs, Not Rust Belt Towns
• TerraFirma Raises $115 Million for Infrastructure Construction Technology
• A Browser Tab Is Not the Test: What Makes a Web App—and Which Examples Fit