# Gritt

> Source: https://aiwiki.ai/wiki/gritt
> Updated: 2026-10-02
> Fact-checked: 2026-10-02
> Categories: AI Companies, AI Energy, Physical AI, Robotics Companies
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Gritt." aiwiki.ai, 2 Oct 2026. https://aiwiki.ai/wiki/gritt
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**Gritt**, also known as Gritt Robotics, is a [robotics](https://aiwiki.ai/wiki/robotics) company that develops automation for outdoor construction, initially for utility-scale solar installations. Its systems combine robotic arms with AI software and attach to equipment such as skid steers and forklifts. The first commercial application is picking up and positioning solar panels.[1][2]

Gritt describes its approach as [physical AI](https://aiwiki.ai/wiki/physical_ai): machines perform construction work while collecting information that can train subsequent models. Its public materials distinguish solar installation from proposed expansion into other infrastructure and site management.[12]

## History and leadership

Obvious Ventures dates Gritt's founding to 2023.[3] The co-founders are chief executive Puneet Puri and chief technology officer Vishal Dugar. According to the investor's account, they met as researchers at [Carnegie Mellon University](https://aiwiki.ai/wiki/cmu)'s Robotics Institute. Puri previously developed inspection drones and an avocado-picking robot; Dugar worked on autonomy for heavy trucks.[4]

Gritt's team page identifies Anirudh Vemula as head of AI and Felipe Roz as head of mechanical engineering. It describes Vemula's background in robotics research at Carnegie Mellon and Dugar's work on autonomous trucking and aircraft.[5]

In September 2024, Gritt Robotics was one of two winners of Round 7 of the U.S. Department of Energy's American-Made Solar Prize. The department announced the results on September 10 at the RE+ conference. Each winning team received $500,000 and $75,000 in technical-support vouchers. The official record identified Gritt's location as Belmont, California, and its project as converting standard construction equipment into robots for solar construction.[6]

Gritt announced its public launch on July 21, 2026, with $32.4 million raised across pre-seed financing and a $26 million Series A. Obvious Ventures led the Series A, with Union Square Ventures and Active Impact Investments participating. Earlier investors included First Round Capital, Climactic, Congruent Ventures, and VSC Ventures. The $32.4 million figure is the cumulative financing announced at launch, rather than the size of the Series A alone.[1]

## Equipment and solar installation

Gritt builds modular systems from commercially available components that can be fitted to construction vehicles already used or rented by contractors. Its founders describe this as a way to collect operating experience on active sites without requiring customers to adopt a proprietary vehicle fleet. They also describe distributing updated model weights to deployed machines over the air.[12]

For solar work, Gritt's product page describes automated panel pick-and-place and compatibility with multiple panel and tracker configurations. It lists purlin assembly, component transport, and oversight and verification as areas of expansion.[2]

The solar product page reported the following cumulative figures when accessed on October 2, 2026:[2]

| Measure | Company-reported figure |
|---|---|
| Panels installed | More than 35,000 |
| Solar capacity installed | More than 18 MW |
| Solar construction contracted | 2.8 GW |

The contracted capacity represents a different measure from installations completed. Union Square Ventures also reported the 2.8 GW contract figure in its July 21 investment announcement, attributing the work to contracts with U.S. construction firms.[10]

On October 1, Puri reported that Gritt was installing 1 MW of solar capacity per day and compared the rate with four times the speed of a typical installation crew. The post referred to an active site. It was a company operating claim, without a published comparison protocol or a breakdown of crew size and working hours.[11]

## Models and training

In a September 2026 technical article, Dugar described the Gritt Foundation Model, or GFM, as a shared representation of construction objects, environments, and interactions. It is pretrained on labeled and unlabeled images and video. Gritt then trains task-specific policies or perception systems, including [image segmentation](https://aiwiki.ai/wiki/image_segmentation) and [pose estimation](https://aiwiki.ai/wiki/pose_estimation), while also updating the underlying representation.[7]

The company describes a process that reconstructs field observations as simulations. Images and videos supply scene descriptions; CAD models and reconstructed objects form simulated episodes with domain randomization. Gritt uses those episodes for training and checks results against the original field event.[7]

Domain randomization is an established method for reducing the gap between simulated and physical environments. Tobin and colleagues' 2017 study varied the rendering of synthetic images and transferred an object-localization system to a physical robot. That research provides context for the method; it did not evaluate Gritt's system.[8]

For deployment, Gritt says it validates candidate actions with uncertainty-aware observation systems, tests new models in shadow mode, and retains takeover and recovery mechanisms. Its technical article also describes transferring capabilities to rebar tying and cinderblock construction. These are developer descriptions, not a published independent evaluation of construction performance.[7]

## Contractor deployments

On August 25, 2026, Burns & McDonnell announced a partnership with Gritt to evaluate and deploy robotics on utility-scale solar projects. The engineering, procurement, and construction company said the partners had spent the preceding year testing Gritt's technology at multiple solar sites. This provides a named contractor's account of field use.[9]

Burns & McDonnell described the robots as assisting crews with repetitive installation work and emphasized continued human operation and management. Its announcement confirms the collaboration and multi-site testing, but does not publish controlled measurements of installation speed, cost savings, or injury reduction.[9]

## Development scope

Gritt's founders identify [data centers](https://aiwiki.ai/wiki/data_center), energy projects, and other large infrastructure as target markets beyond solar. They also propose using observations from deployed machines to support construction planning, verification, and management.[12] The company presents broader site decision-making as a development objective, while its solar page specifically identifies several installation and oversight functions as expansion work.[2]

## See also

- [Cosmic Robotics](https://aiwiki.ai/wiki/cosmic_robotics)
- [Monumental](https://aiwiki.ai/wiki/monumental)
- [Embodied AI](https://aiwiki.ai/wiki/embodied_ai)

## References

1. Gritt. [Gritt Launches with $32.4M to Build Physical AI for Accelerating the World's Infrastructure Buildout](https://www.gritt.ai/news/gritt-launches-physical-ai). July 21, 2026.
2. Gritt. [Solar](https://www.gritt.ai/solar). Accessed October 2, 2026.
3. Obvious Ventures. [Portfolio: Gritt](https://obvious.com/portfolio/?pillar=planetary-health). Accessed October 2, 2026.
4. Andrew Beebe. [Why We Invested in Gritt](https://obvious.com/ideas/why-we-invested-in-gritt/). Obvious Ventures, July 21, 2026.
5. Gritt. [Team](https://www.gritt.ai/team). Accessed October 2, 2026.
6. U.S. Department of Energy. [American-Made Solar Prize Round 7](https://www.energy.gov/cmei/systems/american-made-solar-prize-round-7). Accessed October 2, 2026.
7. Vishal Dugar. [How Gritt works: Practical intelligence for building the real world](https://www.gritt.ai/news/how-gritt-works-practical-intelligence). Gritt, September 2026.
8. Josh Tobin, Rachel Fong, Alex Ray, Jonas Schneider, Wojciech Zaremba, and Pieter Abbeel. [Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World](https://arxiv.org/abs/1703.06907). arXiv:1703.06907, March 20, 2017.
9. Burns & McDonnell. [Burns & McDonnell and Gritt Partner to Advance the Future of Solar Construction](https://www.burnsmcd.com/news/partnership-advances-solar-construction). August 25, 2026.
10. Albert Wenger. [Gritt](https://www.usv.com/writing/gritt). Union Square Ventures, July 21, 2026.
11. Puneet Puri. [Solar installation rate update](https://x.com/puneetp/status/2105464649609367640). X, October 1, 2026.
12. Puneet Puri and Vishal Dugar. [The world can't build enough infrastructure. We intend to fix that](https://www.gritt.ai/news/the-world-cant-build-enough-infrastructure). Gritt, July 2026.