Partnerships

Idemitsu to Fuel a 600 MW Texas AI Data Center Campus

mm
Add Unite.AI to your preferred sources on Google

A Texas data-center developer has lined up its fuel before it has lined up a tenant. Hedgehog USA said on July 21, 2026 that it signed a long-term energy supply agreement with Japanese energy group Idemitsu to feed a planned campus in Navarro County, south of Dallas, engineered to support up to 600 megawatts of data-center capacity by the end of 2027. The site is built to run on natural gas burned on-site — a “behind-the-meter” setup that skips the yearslong wait for grid power that now governs large data-center projects in Texas.

What is actually signed is a supply deal, not a finished data center. Idemitsu will provide long-term energy and, ahead of the campus going live, line up the pipeline capacity and firm gas transportation to feed it. What the announcement does not include is an anchor tenant, a signed grid-interconnection agreement, a construction permit, or the deal’s financial terms. The 600-megawatt figure is a design ceiling for a “development-ready” site, not capacity running today, and Marathon Capital, the project’s exclusive financial adviser, is marketing the whole project for sale.

Why on-site gas skips the queue

In Texas, the binding constraint on new AI compute is no longer chips or capital — it is a spot in line for power. The state’s grid operator, ERCOT, is tracking more than 438 gigawatts of large-load requests, nearly 90% of it data centers, or more than five times the record demand the state has ever drawn at once. Anything pulling more than 75 megawatts at a single site counts as a “large load,” and in June 2026 state regulators approved a new batch-study process, dubbed “Batch Zero,” to work through a queue that had grown too long to evaluate one project at a time. Timing, not price, has become the differentiator: the largest hyperscalers have all pressed ERCOT for a faster path to Texas capacity.

A behind-the-meter campus sidesteps most of that. By generating its own electricity from gas turbines on the property, a developer can offer tenants power on a schedule the interconnection queue cannot promise, and a site that never draws from the public grid falls outside ERCOT’s interconnection process entirely. For a developer racing to stand up training clusters, firm power on a fixed date is worth more than a cheaper site with none.

The constraint does not disappear, though; it moves. A gas-fired campus lives or dies on pipeline capacity and firm fuel delivery, one of a set of workarounds developers are betting on to get out ahead of the grid. Securing that pipeline infrastructure, Idemitsu said, is “essential to creating a truly shovel-ready development.”

A Japanese oil major’s American gas bet

For Idemitsu, one of Japan’s largest energy companies, the agreement extends a young but deliberate push into American data-center power. Founded in 1911 and built for more than a century on oil, fuels, and lubricants, the company is now standing up a nationwide US natural-gas-supply business aimed squarely at AI sites. It has also said it is exploring adjacent services for data centers — immersion cooling, diesel backup, and battery storage — the kind of full-stack offer that turns a fuel supplier into an infrastructure partner.

This is its second such arrangement in under a year. In November 2025 Idemitsu partnered with Overwatch Capital to supply up to 1 gigawatt of on-site gas generation for data centers across ten states, from Texas to Pennsylvania. Marathon Capital advised that deal too, and both follow the same template: pair Idemitsu’s gas with a “powered land” developer, a firm that buys the land, secures the electricity, and sells a shovel-ready site to a hyperscaler that wants compute online fast.

Hedgehog says it owns its land outright, mineral and surface rights included, and commits to long-term gas contracts before it engages a single customer. In a market where power, not real estate, is the scarce input, that sequence is the product, and with Marathon shopping the entire project, it is one Hedgehog intends to sell rather than operate.

What still has to line up

Whether the model delivers 600 megawatts by the end of 2027 now turns on the unglamorous work behind the release: turbines ordered, pipeline laid, permits cleared, and a buyer signed. On-site gas also carries a cost the grid queue does not. Texas already leads the country in proposed gas plants tied to data centers, and behind-the-meter generation swaps an interconnection delay for on-site emissions and the developer’s own generation risk. For now, Hedgehog has the land and the fuel. The compute, and the customer, come later.

Theo Nash is an AI-generated specialist at Unite.AI, covering AI infrastructure, compute, and the hardware systems that power modern artificial intelligence. His work focuses on the technical foundations behind large-scale AI workloads, including data centers, accelerators, networking, and the software stacks that tie them together.

With an analytical and engineering-driven perspective, Theo examines how advances in GPUs, custom silicon, memory architectures, and distributed systems enable new generations of AI models. He pays particular attention to performance trade-offs, energy efficiency, scalability, and the practical constraints that shape real-world deployment of AI infrastructure.

Articles authored by Theo Nash are AI-generated and reviewed by Unite.AI’s editorial team to ensure technical accuracy, clarity, and responsible coverage of the rapidly evolving AI compute landscape.