AI Data Centers in Texas (ERCOT)
AI Data Centers in Texas (ERCOT)
Texas has become the epicenter of the US AI data center buildout — and also the clearest case study in how fast interconnection queues can outrun a grid's ability to deliver power.
AI Data Center Buildout in Texas
Cheap land, deregulated power markets, favorable tax treatment, and a business-friendly permitting environment have made ERCOT territory a top destination for hyperscale and AI-focused data center development. That demand has shown up almost entirely as large-load interconnection requests rather than completed projects — which is exactly where the risk lives for investors and developers.
Power Demand vs Grid Upgrades in ERCOT
The scale of the mismatch is stark. ERCOT's large-load interconnection queue grew from about 63 gigawatts at the end of 2024 to roughly 226 gigawatts by November 2025, and to over 410 gigawatts by April 2026 — nearly five times ERCOT's current peak demand of roughly 85 gigawatts. Data centers now make up an estimated 87–95% of that queue. Yet delivered reality is a fraction of that: ERCOT reported only about 1.8% of queued capacity was actually operational as of late 2025, and just 2,168 megawatts — less than 0.5% of the queue — was approved to energize in the twelve months ending March 2026.
Texas lawmakers responded with Senate Bill 6 (2025), which shifted ERCOT away from a pure first-come, first-served model toward one requiring large loads over 75 megawatts to show proof of financial commitment, site control, and interconnection readiness. By early 2026, that requirement had cleared roughly 40% of speculative projects out of the queue. ERCOT has also moved to a new "Batch Zero" large-load interconnection process, with an initial transmission plan expected in fall 2027 and a second batch opening for applications in summer 2027 — a reminder that even "fixed" queue processes still run on multi-year timelines.
Texas Interconnection Queue Risk
For any Texas-sited AI data center, the queue position and batch-study status matter far more than the announced timeline. Roughly 30,000 megawatts of ERCOT large-load demand is currently considered "study-backed" through 2030, while only about 6,000 megawatts has actually been observed taking service — a useful reality check against the 400+ gigawatts of total requests in the pipeline.
Investor and Utility Use Cases
Investors underwriting Texas data center exposure — whether through REITs, utilities, power generators, or AI infrastructure equities — need to know whether a specific project sits in study-backed capacity or speculative request territory. Utilities and transmission providers like Oncor, which alone has received large-load requests exceeding Texas's entire system peak demand, are using this same queue data to prioritize capital spending and communicate realistic timelines to customers.
Where to Go Deeper
The Powering AI report includes a dedicated ERCOT section tracking queue depth, Batch Zero progress, and SB 6 implementation so you can separate real Texas data center capacity from the noise.