Rune raises $40M to bolt AI servers onto solar farms and skip the multi-year wait to plug into the US power grid.

Rune, a San Francisco startup, raised a $40 million Series A on Tuesday to ship modular AI servers that bolt directly onto existing solar farms. The round, led by Spark Capital, brings Rune's total funding to $53.5 million and backs a product the company calls RELIC, short for Renewable Energy Linked Intelligent Compute. Rune says RELIC can energize customers in as little as six weeks from contract (Business Wire via Morningstar, TechRadar).

RELIC sits behind the meter at a solar facility and runs natively on DC, the same form the panels produce, so the company says it skips both the AC conversion losses a normal data center incurs and the multi-year wait for a grid interconnection study. CEO and co-founder William Layden put it in plain terms: "Every solar plant is a latent data center. The power is already there, sitting idle while AI labs wait years for grid connections that may never come" (Business Wire via Morningstar).

Rune says a 200-megawatt pilot is already live in Texas on existing solar infrastructure, with no site modifications, and that a single RELIC unit can be installed in roughly 60 minutes (Datacenter Dynamics, SiliconANGLE). The company's stated savings math is the most concrete number in the announcement: Rune says RELIC cuts non-compute infrastructure cost by about 85%, or roughly $620 million saved on a 100-MW deployment (Unite.AI). Rune also says US solar plants waste up to 20% of generation, around 50+ TWh per year, the figure it is pitching as "stranded" capacity (TechRadar).

The macro case for the bet is not Rune's alone. Lawrence Berkeley National Laboratory's Queued Up 2026 edition puts the US active interconnection queue at about 2.6 TW at the end of 2025, with solar alone at roughly 773 GW. The headline number is down about 19% year over year, but the queue is still the largest in the world by a wide margin, and it sets the floor on how fast any new large load can reach the grid. S&P Global Commodity Insights reports ERCOT solar curtailment is on pace for another annual record in 2025, while Amperon documents a roughly sixfold year-on-year jump in PJM solar and wind curtailment in 2024, with 2025 already running higher. Stranded solar is not a marketing line; the queue and the curtailment data say it exists.

What is harder to verify is whether Rune's specific numbers hold up outside one site. The 85% non-compute capex reduction, the $620 million per 100 MW, the 60-minute install, and the six-week delivery window are all company-asserted and have not been independently audited in published technical literature. No third-party benchmarking of RELIC's DC-native efficiency has appeared. The "no water" claim is also a relative one: a 200-MW solar-sited cluster may use a fraction of the evaporative cooling a conventional hyperscale hall would, but absolute water use at scale depends on local climate and on whether liquid cooling is the dominant heat-rejection path. Multi-site economics, not the Texas pilot, are what would let a buyer underwrite the math.

The Series A line-up tells part of the same story. Spark Capital led, with Union Square Ventures, Lowercarbon Capital, Activate Capital, Committed Capital, Timeless Partners, and Logos Fund participating (SiliconANGLE). Lowercarbon's involvement signals the pitch is being read as a climate-and-compute wedge. That changes how a hyperscaler should price the deal: a behind-the-meter product is, in effect, a queue-skipping right, not a power-purchase agreement.

The next test for Rune is whether it can name a second anchor site outside Texas by the end of Q1 2026, and whether the public economics on that second site match the first. If they do, the queue-skipping lens the company is selling becomes a tool every other AI-infrastructure announcement this cycle has to be read against. If they do not, the six-week promise stays a press-release line, and the 2.6-TW queue keeps getting longer.