A small Bay Area startup has been quietly deploying utility-scale solar at data centers around the country for the last year, having sold hyperscalers on a reimagined version of the technology that leverages robots and software to turn project sites into factories in the field.
Founded in 2020, Planted Solar now has a 20-plus-gigawatt pipeline, largely composed of projects that will either be built onsite at a data center, or will directly supply one. The company has yet to publicly announce those data center customers, but the list includes some of the biggest and best-capitalized developers in the tech industry, both newcomers and legacy players.
Because those projects will be pre-designed in software and installed in large part by robots, each will go up in months, rather than the several years that solar projects typically take to move from development to power.
Speed to power, the current mantra for data center developers, is central to Planted’s value proposition. The company has installed 42 megawatts in the last six months. That includes a 28-MW behind-the-meter project for a data center in the Southeast for which development, from first contact with the data center operator to a fully completed array, took just 10 months. In 2027, Planted plans to install as much as a gigawatt of capacity, founder and CEO Eric Brown told Latitude Media.
Today, Brown explained, the constraint on solar deployment isn’t cost, but deployment capacity.
The next breakthrough in solar, he said, will be about the ability to build fast enough, in the right places, and at the scale demanded by the AI boom and electrification. Accordingly, the Planted thesis is that the only way for solar to be a meaningful part of the near-term data center buildout is to build more flexibly and more efficiently than the industry ever has before.
“The biggest feature for most of the new demand right now is reliability of deployment and speed of deployment,” Brown said. “For solar to really have a space in the new load growth requirements, you have to put the power in the right spot.” Doing so means solving for bottlenecks like land availability and EPC capacity, as well as permitting and the manual labor of installing a project. And Planted is betting that an automation-first, software-forward approach can make that happen.
“We’ve thought about this very deliberately to cut all of the steps and processes [that can be eliminated], and automated what’s left over,” Brown said. That approach, he added, lands well with the company’s tech sector customers: “Once you get in the conversation, it feels very logical to the folks who are thinking about the right way.”
Building solar for data centers
The company’s approach, though in some ways more high-tech than traditional solar, relies largely on the industry’s most basic components. Modules are mounted on posts made of commodity tubing, ordered in massive bundles from standard pipe mills, then installed at hip-height and at a higher density by following the natural terrain of a site. Planted’s only proprietary hardware in the projects is a bracket that clips modules onto the posts, which is intentionally simple, designed to work with standard modules from any vendor.
But that’s where simplicity ends. The development process starts with creating a digital twin of the site using aerial lidar, and then populating the mechanical and electrical components, from individual foundation poles to modules and wiring runs.
That plan is then executed by robots — custom-built tracked hydraulic machines that pick posts out of a toothpick-dispenser-like magazine, and install them in position based on the 3D rendering, with millimeter-level precision. The system defines exact post locations and measures each installation as it goes, flagging anomalies like buried concrete and automatically generating quality-control and rework guidance for the field crew. (This is something that would normally be covered by six to 10 different companies who specialize in various parts of development, according to Brown.)
Planted’s speed and energy density — about five times faster to build and twice as energy-dense per acre compared to traditional single-axis solar projects — is thanks in large part to its ability to build on uneven terrains.
Single-axis trackers — used in most utility scale solar projects — require relatively flat, uniform surfaces, which usually means grading large areas of land before construction starts. Even a crowned soccer field can be “too wavy,” Brown explained. But grading disturbs topsoil, slows stormwater permitting, and delays final stabilization.
Instead of grading sites, Planted relies on its digital twin of the site to design projects in dense, fixed rows that conform to existing hills and contours. The robots then drive directly over uneven ground, installing posts at the appropriate depth for the terrain in low, aerodynamic rows.
The company can therefore pack in more megawatts onto the same footprint, even on complex, uneven land that conventional trackers often can’t touch. Projects are built on top of existing vegetation, which cuts down site preparation work and shortens the permitting process, Brown explained, because it keeps the ground more stable and the soil healthier post-construction.
The approach also uses about half the land and has lower visual impact than traditional utility-scale projects.
This geographic flexibility unlocks solar’s ability to be a real solution for co-location or behind-the-meter power for data centers, Brown said. “As soon as the developer or hyperscaler has decided to do some sort of bridge power for onsite, we can really be the best renewable backbone for that stack that actually gets in the ground on the timeline that’s required, because of our ability to go faster and locate on less land overall,” he explained.
Autonomous scaling
Brown and much of the core team came from Cogenra Solar, a cleantech 1.0 startup that built high-efficiency solar panels and was acquired by SunPower in 2015. The takeaway from that experience, Brown said, was that when modules plummeted in price, the bottleneck to deployment became “everything else.” Planted was founded to target those remaining constraints.
While the bulk of Planted’s near term pipeline is for co-located data centers, the company actually has two additional verticals: community solar and utility-scale independent power producers. Hyperscalers have long been a “strong enabling factor” for the deployment of low-carbon energy, and Planted always expected them to be a significant part of its business. The AI boom didn’t necessarily change the company’s strategy, he added, but it did change the timing, pulling expected demand forward to 2026.
The “robots” installing projects this year are actually modified Caterpillar tractors with a human in the cab, ensuring each physical post is aligned with its planned location in the digital twin. Each robot has a weekly installation capacity of about 2 MW, and to date Planted’s data center projects have gone up with four robots on site at a time, working in parallel.
The company’s plans to scale from 100 MW installed this year to a gigawatt next year relies almost entirely on its ability to add more machines to the fleet, which are hauled to sites around the country on flatbed trucks. The ability to attract customers, Brown added, is not a constraint at all.
Planted’s next generation of robots — on whose shoulders the gigawatt-in-2027 plan rests — will be fully autonomous. These will drive themselves over uneven terrain, follow the digital twin without an operator in the cab, and coordinate as a fleet to install projects much faster. They are already being manufactured by a network of contractors that build industrial tracked platforms; Planted then equips them with its own robotics and control systems.
The first of these robots are already being tested at Planted’s demonstration site east of San Francisco.
A fleet of them, Brown explained, “decouples us from a lot of the constraints like skilled operators. We can make [deployment] a manufacturing problem, rather than the human capacity factor challenge that is hitting the industry across all levels right now.” And as the robots take on installation, the data they amass will train the models they run on, meaning the system is constantly learning.
“Our system operates like a factory,” Brown said. “It’s a bunch of repeatable, simple steps that you can parallelize as much as possible…We’re extending those same repetitive learning curves from project to project…so that each project actually gets faster.”
This is already playing out, he added: “Cycle times,” meaning how long it takes for a robot to install one foundation and move to the next, has already improved from around 55 seconds per pole to 47 seconds.
The next inning of data center power
Planted, and its focus on data center customers, is in a crowded field of speed-to-power solutions. Virtual power plant providers are inking “bring-your-own-capacity” deals with hyperscalers, electric vehicle charging companies and battery recycling companies alike are building behind-the-meter storage solutions, and even carbon removal companies are pivoting to power the AI boom.
But most of the players in this space are piling in with partial solutions to the challenge of putting “reliable megawatts in the right spot at the right time,” Brown said. “A lot of companies are solving for one piece of it and not the broader integrator problem, especially when it comes to renewables.”
Solutions like modular gas and Bloom Energy’s “power plant in a box” have gained traction in the data center era because they’re compact, standardized, and largely the same from site to site, he added. Solar has so many more external dependencies, including land availability, entitlements, zoning, and possibly a connector feeder or transmission line integration, that complicate things. Planted is betting that much of the next generation of gigascale data centers will be powered by hybrid microgrids that leverage multiple types of generation: largely solar balanced by storage, with grid and gas as firming and backup.
Ultimately, Brown said, the industry would benefit from a single source provider that can move quickly . “I think that’s a good reason why Google bought Intersect, because they were really leading the charge of the more nuanced grid positions,” he added. “You’re seeing that again with hyperscalers going way upstream to buy either solar projects or secure the pipeline in a way that funnels directly into their power needs.”
Planted’s spot in this crowded field is to serve as the “execution layer underneath it all,” Brown said, getting solar installed at a pace and density relevant to hyperscalers. “It’s a fun moment to be in, it feels like the first inning for a lot of people.”
