TAR, a Texas-based green energy startup, has closed a $27 million seed round to build behind-the-meter modular power systems for AI data centers, according to Business Insider coverage published June 18, 2026. The pilot system delivers roughly 10 MW of constant on-site power, and TAR plans to field more than 200 MW of steady load capacity before 2027. Co-founder Pat Becker frames the plug-and-play model around speed and independence from the grid. The systems combine solar, wind, battery storage, and limited natural gas backup, prefabricated and pre-tested in a factory for rapid deployment.
The US grid interconnection queue can run five to seven years or longer for a large facility. TAR sells around that wait, and the trade is a fixed power envelope set by the site itself. A campus that owns 10 to 200 MW sizes its thermal architecture against a self-generated budget, with no later negotiation to add grid capacity. That budget caps cooling capex before a single rack lands. The same grid wall driving this pivot is the one we traced in power grid constraints on data center development.
When power is the scarce resource, every watt spent moving air is a watt stolen from compute. Operators working inside a self-imposed budget push hard toward direct-to-chip liquid cooling and closed-loop water to protect that budget. Higher rack densities extract more compute per watt of self-generated power. Air handlers that consume 30 to 40 percent of facility load do not survive that math, which is the physical reason liquid is now the default at high density, as we laid out in why liquid cooling became mandatory at AI rack densities.
Becker's prefabricated approach mirrors the comparables TAR cites: Aligned Data Centers shipped a 31 MW battery project in 2025, and Redwood Materials completed a solar and EV battery microgrid in under four months that same year. Speed on the power side forces speed on the cooling side. A coolant distribution unit, cold-plate manifold, and closed loop have to be procured and commissioned on the same compressed timeline the gensets and batteries arrive on.
On-site generation also collapses the distance between heat source and heat sink, the same calculus driving the off-grid gas campuses we covered in the private off-grid power build that skips public review. A site with limited natural gas backup still concentrates two thermal loads on one parcel. The closed-loop, low-water design that protects a self-generated power budget is the same design that survives the water scrutiny a fast-permitted campus invites. TAR's $27 million bet on speed sets the clock, and the cooling order book now moves on that clock: pre-engineered liquid skids, factory-integrated CDUs, and minimal external water draw.