Gritt and MasTec Partner to Bring Physical AI to Solar Construction

Gritt and MasTec are expanding the use of physical AI on renewable-energy construction sites, starting with utility-scale solar projects across multiple U.S. states. The partnership pairs Gritt’s robotic systems with MasTec Renewables’ construction teams, bringing AI out of planning software and into the movement and installation of materials.

According to the announcement, MasTec evaluated Gritt on active projects where the systems helped install megawatts of solar panels. The companies now plan a pipeline of deployments and will explore applications in other infrastructure sectors. They did not disclose the agreement’s financial terms or identify individual project sites.

For construction robotics, the significance is the operating environment. A solar farm offers thousands of repetitive installation tasks, but those tasks happen amid changing weather, uneven ground, moving equipment, and people working nearby.

Robotic Attachments for Equipment Crews Already Use

Gritt’s approach combines AI with robotic attachments for common construction machinery. In its launch announcement, the company described connecting robotic arms to vehicles such as skid steers and forklifts to perform material handling and assembly. This adds automation to familiar equipment rather than requiring contractors to introduce an entirely new vehicle platform.

The company emerged publicly in July with $32.4 million in pre-seed and Series A financing, including a $26 million Series A led by Obvious Ventures. The MasTec agreement follows that launch and an earlier solar-construction partnership with Burns & McDonnell.

Gritt’s solar product page identifies panel pick-and-place as a current application, with purlin assembly, component transport, and oversight among its expansion areas. It reports more than 35,000 panels installed and more than 18 MW of solar installed. Its separate 2.8 GW contracted-construction figure describes future work under contract, not completed installations.

How Gritt Teaches Machines to Handle Changing Sites

In a technical explanation of its architecture, Gritt describes a foundation model trained on labeled and unlabeled images and video from construction environments. That shared representation supports task-specific models, including control policies, segmentation, and six-degree-of-freedom pose estimation—identifying an object’s position and orientation in three-dimensional space.

The company combines learning with checks on execution. Observation systems assess uncertainty and validate proposed actions, while new models first operate in shadow mode before taking control. Takeover and recovery mechanisms allow work to continue when the system needs help and generate examples for subsequent training.

Gritt also describes a real-to-simulation-to-real pipeline: site images and video help reconstruct scenes, which become simulated training and validation episodes. CAD models, reconstructed objects, and randomized conditions let the company revisit unusual situations without depending entirely on encountering them again on a live site.

These mechanisms address a central deployment challenge: a robot must handle unfamiliar conditions without turning every unexpected event into a damaged component or a stopped project. The engineering goal is repeatable performance under variation, not simply a successful demonstration.

Why MasTec Is a Relevant Deployment Partner

MasTec Renewables’ solar business spans development and design through construction and operations. Its capabilities include photovoltaic system modeling, materials logistics, foundation work, electrical collection systems, substations, and high-voltage connections. It installs fixed-mount systems as well as single- and dual-axis tracking configurations.

That breadth puts robotic panel installation within a much larger delivery process. A machine’s placement speed matters only if crews, materials, foundations, and electrical work remain coordinated. An attachment that performs well in isolation still has to fit the contractor’s sequencing, maintenance, and safety practices.

MasTec Renewables Vice President of Construction Rob Turner said the company’s evaluation showed an opportunity to combine its workforce’s experience with Gritt’s technology. The companies frame the rollout around supporting skilled crews, with people continuing to guide and validate work.

Worker Detection Is Part of the Technical Problem

Gritt’s published research also illustrates why outdoor robotics safety requires attention to the training pipeline. In an analysis of worker detection, the company examined how automated data-quality filters can disproportionately discard images of kneeling, crouching, or partly obscured people.

Those positions are ordinary on construction sites, even if they appear infrequently in a filtered dataset. A bounding box may correctly identify a worker while a pose estimator struggles to reconstruct the person’s joints. Rejecting the whole example can remove exactly the cases a deployed robot needs to recognize.

Gritt reported that adding simulated workers in mined working postures improved detection recall for low poses by 1.6 percentage points on real site footage. The result concerns that specific perception evaluation; it is not a measure of the MasTec deployment’s overall safety performance.

From Installing Panels to Understanding the Jobsite

Gritt’s broader ambition is to connect physical execution with site verification and planning. Its systems collect information while working, creating the possibility of comparing what has actually been built with what a project expects to complete next.

For MasTec, the immediate application remains solar construction. For Gritt, repeated deployments offer a route to adapting its technology across more tasks and conditions. The partnership’s next test is operational: whether useful automation can be introduced consistently across projects while preserving the judgment and coordination that construction crews provide.