> ## Content Index > Fetch the complete content index at: https://nexi.fund/llms.txt > Use this file to discover other available public pages before exploring further. # The Gigawatt-in-Ten-Weeks Machine: Terafab V2 and the End of Manual Solar - URL: https://nexi.fund/autonomous-solar-construction-robotics-2026/ - Published: 2026-08-18T07:00:55.000Z - Updated: 2026-08-18T07:00:55.000Z - Description: Terabase's Terafab V2 uses a mobile factory and a fleet of rovers to install a gigawatt of solar in ten weeks. A $130M SoftBank-led Series C โ€” with Breakthrough Energy Ventures, Prelude, Fifth Wall, SJF and EDP โ€” funds a line that could crack the labour bottleneck. - Author: Nexi.fund Labs - Tags: Defence & Robotics, #mode-3, #hook-paradox, #track-F Solar is the cheapest source of electricity ever built. It is still assembled by hand, panel by panel, out in the sun. Utility-scale photovoltaic is the fastest, most cost-competitive generation the grid can add today. The constraint is not silicon. It is not inverters. It is not capital. It is labor, crews of workers who bolt steel and glass across thousands of acres. Terabase Energy, a robotics company in Berkeley, California, sells a machine to fix that. Its Terafab line finished field testing in March 2026 and is now shipping commercially. ๐ŸŽฏ **The short version** Terafab is a mobile field factory. It pre-assembles solar panels onto their supporting frames, checks every unit inline, and lets a fleet of rovers place them. One line runs 24/7 and installs more than 20 MW per week, roughly 1 GW per factory per year. Terabase is scaling factory capacity toward 10 GW a year. The demand side is doing exactly what the automation side needed. Data centers and AI are pushing American electricity demand to levels not seen in decades. Hyperscale operators need power in years, not in the decade a coal plant took. Solar plus batteries is the shortest path from permit to electrons. The world added a record 814 GW of wind and solar capacity, as Electrek reported in March, and the pace is still being capped by physical installation, not by panel production. As we wrote in August, robots have started showing up on construction sites, from DEWALT's data-center drilling rigs to autonomous earthmovers. Terabase is the more radical version of the same bet: move the whole assembly line onto the field. The question is whether a machine that works in a clean factory keeps working in heat, wind, mud, and dust. Terabase just bet $200 million that it does. ## The factory that moves Terafab inverts the way solar farms get built. In conventional construction, crews install steel torque tubes in the field first, then lift heavy glass panels onto them one at a time. Terafab does the assembly work in a mobile factory instead. Panels and torque tubes are joined into single units, checked by machine vision as they move down the line, and carried out to their spots. The hard part was never the picking and placing. It was doing it outside. Campbell is blunt about what a field factory has to survive: rain, hail, wind, tornadoes, dust, ants, bees, snakes, badgers, rats. Literally. Terabase spent years hardening the line so it can hold factory-grade precision in conditions a factory never sees. The throughput number is the one that matters for the investment question. A single Terafab line runs on a two-minute cycle. Continuous operation, around the clock, adds up to more than 20 MW installed per week. That is about 1 GW per factory per year. 1 GW per factory per year #### Terafab V2 installed throughput One line, 2-minute cycle times, running 24/7: more than 20 MW installed per week, about 1 gigawatt per factory per year. ยท *Terabase Energy, 2026* Compare that with the manual baseline. A utility-scale project the size of a medium plant needs hundreds of thousands of panels mounted by hand, each one lifted, aligned, and bolted by a person in the heat. Developers are already fighting skilled-labor shortages, schedule risk, and rising rates at the moment the market most needs speed. For EPC contractors, a machine that doubles installation productivity means taking on more project volume without waiting for more crews. > Every week we shave off a construction schedule means earlier revenue for project owners, lower financing costs, and faster delivery of clean electrons to the grid.โ€” Matt Campbell, CEO and co-founder, Terabase Energy ## Why the build is the bottleneck Solar panel prices collapsed over the past decade. Commodity modules are cheap, inverters are cheap, and the grid is hungry. What did not get cheap is the physical act of turning a flat field into a generating asset. Terafab attacks exactly that step, and it has a track record to show: the first-generation system has been deployed on five commercial power plant projects in the United States, installing about 40 MW of tracking solar to date. That changes the risk profile of the story. This is not a concept video. The earlier Terafab has been running on real projects, and Terabase says developers that used it reported higher productivity, better build quality, and improved safety. The next version, V2, is what went into commercial shipment in March. Terafab V2 is built around three pieces. The assembly and inspection center, erected on the site, where panels and torque tubes are joined and checked in real time. The robotic arms that load approved units. And the rovers that drive the finished assemblies out to their target positions in the array. Behind them sits a Manufacturing Execution System, software that uses AI to manage and optimize the build of the whole plant. The honest detail is in the last step. In the current version, the rovers deliver the units and workers manually fit them onto the pre-positioned mounts. Terabase says a future version, due in 2027, will automate that final attachment too. So the machine has taken over the heavy lifting, the positioning, and the inspection. The last metres are still human. Terafab V2 is also smaller and faster than the original. The line can be packed up and moved in about four hours, which is how a single fleet of factories can hop from one site to the next. The company expects the rovers to run fully autonomously in the near term, and the panel handling is evolving: the current line is tuned for First Solar Series 7 panels and Nextpower trackers, with silicon-module and other-tracker compatibility on the way. ## The money behind the robots The commercial rollout is backed by a $130 million Series C led by SoftBank Vision Fund 2, a round reported in March 2026 and largely financing the Terafab build-out. It brought Terabase's total funding past $200 million, with participation from Breakthrough Energy Ventures, Prelude Ventures, Fifth Wall, SJF Ventures, and EDP Ventures. SoftBank's partner Kentaro Matsui framed the thesis plainly: digitalizing and automating solar deployment positions the sector to scale. The point is not the technology for its own sake. It is that photovoltaic installation is one of the largest infrastructure buildouts in history, and it is still run like an outdoor assembly bench. What the capital is being spent on is factory capacity, not a demo fleet. Terabase is building out its Northern California facility over the next twelve months to support up to 10 GW of installations per year. There are two deployable factories available now, a third expected by the end of 2026, and Campbell says the company plans to have ten factories ready in the second quarter of 2027\. The machines are designed and manufactured in the United States, which makes the story double as a domestic-manufacturing play. Terabase is selling solar construction robots, but the software is a real asset too. Its Construct construction-management platform has supported more than 12 gigawatts of projects. PlantPredict handles energy modeling. The company is an engineering-services business with a robotics line attached, not the other way around, and that matters when you are sizing the revenue base beyond hardware units. ๐Ÿ’ฐ **What the $200 million buys** Total funding past $200 million, more than $130 million of it a SoftBank-led Series C focused on Terafab. Capacity plan: 10 GW per year of factory output from Northern California, ten deployable factories targeted for Q2 2027\. ## The gigawatt-in-ten-weeks math Campbell has put a specific, checkable number on the road ahead: install a gigawatt in ten weeks. The way he describes it, that means sending four or five Terafab lines to one site and running them around the clock. A gigawatt in ten weeks changes the shape of a project. A utility-scale plant of that size, built by hand, needs months of sequential crew work from groundbreaking to grid connection, and every month of schedule adds financing-carry cost to the developer's books before the asset earns a single kilowatt-hour. Shave a month off and the owner books earlier revenue on a lower interest bill. That is the mechanism Campbell is selling, stated in the company's own words, and it is why the pitch reads less like a hardware sale and more like a project-finance service. Put the numbers side by side. The world added a record 814 GW of wind and solar in the last reporting year. A single Terafab line nominally installs 1 GW per year running continuously. Even a modest slice of global utility-scale build-out, routed through automated lines, absorbs a meaningful share of the installation shortfall. The gap is structural. Terabase reports its construct platform has already supported more than 12 GW of projects, so the software side has real exposure to the market before the hardware fleet even scales. Ten factories, ten gigawatts a year of headline capacity, each line moving between sites in about four hours. The fleets hop from project to project, which is why utilization, not machine speed, will decide whether the unit economics hold. A factory that stays busy is a profitable line. A factory waiting for interconnection is a fixed cost with nowhere to go. ## The limits the press release hides The announcement is precise about throughput and quiet about the edges. That is where the sober reading lives. First, the line is specialized. Terafab V2 is optimized for one panel type and one tracker type. A machine that only handles First Solar Series 7 panels is a production line tied to a supply chain, not the universal solution the headline implies. Silicon-module versions are coming, but the timeline is open. Second, the last step is still manual. Rovers carry the assemblies, and people fit them onto the mounts. Terabase targets full roving autonomy and automated placement for 2027\. Until then, the machine removes the heavy lifting and the inspection burden but does not remove the crew from the site entirely. Third, the economics depend on utilization. A factory that moves between sites in four hours is only cheap if it stays busy. The gigawatt-per-year figure assumes continuous 24/7 operation, which in turn assumes a pipeline of projects large enough to keep ten factories loaded. Against the current pace of interconnection and permitting, that is not guaranteed. #### What still needs a crew \+ Manual unload of assembled units from rovers \+ Manual fit onto pre-positioned tracker mounts \+ Site prep, cabling, and commissioning outside the automated flow **What 2027 needs to prove:** fully autonomous rovers and automated mount fitting on a live commercial project. The labour angle cuts the other way too. Construction automation has a long history of press releases over production. Every site is bespoke, every jurisdiction different, every workforce wary of a machine that replaces shifts. Terabase argues the grid's demand curve removes the choice: the projects cannot be built by the crews available. That is the strongest argument in the pitch, and it is also the one that depends least on the robots themselves. If the machines are this good, why is most solar still built by hand? ## The counter-argument The honest sceptic case is not about robots failing in the field. The V1 system already ran on five projects. The case is about whether the machine earns its keep across the whole project, not just the peak assembly window. Terafab occupies a narrow, repetitive slice of construction: the mounting of tracking arrays. It does not pour foundations, pull cable, or hook up inverters. A developer that buys one buys faster assembly in exchange for a sizable capital commitment and a machine from a single vendor. The bet is that standardization wins, that the industry will converge on fewer building methods, and that the company that owns the fastest method owns the margin. That bet has precedent in other capital-intensive industries that industrialized their build phase. It also has recent failures, and the scar tissue is instructive: automation in construction has repeatedly been a demo-year business where the machine works and the economics do not survive the first downturn in project starts. Terabase's answer is scale. Ten factories, 10 GW a year of capacity, a gigawatt in ten weeks per cluster of sites. That goal, stated by Campbell, is falsifiable. If the 2027 version places the final mount automatically and the ten-factory fleet lands on schedule, the automated-solar thesis stops being a demonstration and becomes unit economics. If the pipeline falters, idle factories become the cost. For a robotics investor, the transferability matters most. The same autonomy stack, outdoor robotics, machine-vision inspection, and fleet coordination that Terafab needs on a solar site is the stack that defence and logistics programmes are paying for now. Dual-use is a word that gets used loosely in robotics; here it is structural. Terabase did not build for defence, but the capabilities it had to build to survive a desert summer are the ones the market is short of. The paradox of clean energy is that its cheapest input, sunlight, is free, while its costliest input per panel is a tradesperson with a torque wrench. Automation does not replace the sun. It replaces the wrench. If Terabase delivers on scheduling, the terawatt era of solar stops being bounded by how many workers want to stand in triple-digit heat. ๐Ÿ“ก **Signal: the 10 GW capacity target** Factory capacity scaling to 10 GW per year from Northern California, with ten deployable Terafab lines targeted for Q2 2027\. Watch whether shipments follow the schedule. โœ… **Signal: 2027 autonomy milestone** Fully autonomous rovers and automated mount fitting are the next milestone. Confirmation on a live project would close the last manual step in the line. โš ๏ธ **Signal: panel compatibility** The line is tuned for First Solar Series 7 and Nextpower trackers. Silicon-module compatibility will decide whether Terafab serves a niche or the whole market. ## Sources [ Terabase Energy's Next-Generation Terafab Completes Field Testing, Ready for Deployment The company's announcement covering V2's field-testing results, 2-minute cycle times, 1 GW per factory per year, and the 10 GW capacity build-out. Terabase Energy ](https://www.terabase.energy/resources/terabase-energys-next-generation-terafab-completes-field-testing-ready-for-deployment?ref=nexi.fund) Primary source for throughput figures, factory capacity, and CEO commentary. [ Terabase Energy advances automated PV construction with robotics, AI tools Independent specialist coverage including the 40 MW already installed by V1, the 2027 autonomy roadmap, and the SoftBank-led Series C financing detail. pv magazine ](https://www.pv-magazine.com/2026/04/03/terabase-energy-advances-automated-pv-construction-with-robotics-ai-tools/?ref=nexi.fund) Closest thing to an independent audit of the deployment roadmaps. [ A 24/7 solar farm-building robot just hit the market Consumer-facing write-up of the Terafab V2 announcement with the demand-side context: US electricity demand growth from data centers and AI. Electrek ](https://electrek.co/2026/03/20/a-24-7-solar-farm-building-robot-just-hit-the-market/?ref=nexi.fund) Useful for the macro framing on why speed to power matters right now.