Texas Hits Pause on Data Centers. The 20 Percent Crisis Narrative Doesn't Add Up.

I've been very surprised that the market is treating the Texas data center moratorium as an infrastructure catastrophe. The Bloomberg NEF headline that 20 percent of the U.S. data center pipeline is at risk of delay sounds alarming on paper. But the closer you look at the actual numbers behind that claim, the more it reads as a false narrative — one that confuses speculative interconnection requests with committed capital, and mistakes a procedural delay for a structural crisis.

Here's what actually happened. On August 3, Governor Greg Abbott directed the Public Utility Commission of Texas and ERCOT to pause all new data center grid interconnections pending a comprehensive project-by-project audit. The audit will verify each project's power consumption, water usage, tax incentives, ownership structure, and community impact before any project advances. Non-compliant projects will be denied grid connection.

That is a regulatory speed bump, not a road closure. And the scale of what Bloomberg NEF is calling a crisis is built on numbers that deserve decomposition before you act on them.

The ERCOT interconnection queue currently contains approximately 474 gigawatts of power connection requests. About 90 percent of those requests come from data centers. That figure — more than five times Texas' record peak demand of 91.3 gigawatts — is where the drama comes from. But the queue is not a pipeline. It is a wish list. Companies file interconnection requests at minimal cost, and most never materialize.

Bloomberg NEF itself provides the corrective context that gets lost in the headline. Of the roughly 50 gigawatts of Texas data center projects that BNEF actually tracks as credible, about 70 percent are in the early stage. BNEF forecasts only 8.25 gigawatts of new data center capacity coming online in ERCOT through 2030, bringing the total to just over 17 gigawatts. The gap between 474 gigawatts in the queue and 8.25 gigawatts expected through 2030 tells you everything you need to know about the speculative nature of this number.

BNEF's own estimate is that the audit could delay 49.8 gigawatts of compute load — representing nearly 20 percent of the U.S. 253-gigawatt development pipeline. The revenue loss in a one-quarter delay scenario, assuming 60 percent of delayed capacity is AI-related, is about $8 billion by the first quarter of 2027. In a worst-case full delay, that rises to $15 billion. Those are not trivial sums. But they are also not the systemic collapse the headline implies — they are cost estimates for a delay of indeterminate duration applied to a queue where most projects were never going to connect.

There is also a structural escape hatch that the false narrative ignores entirely. The moratorium explicitly does not apply to behind-the-meter on-site power generation. Texas already leads all U.S. states in announced behind-the-meter data center capacity at 40 gigawatts, driven by Permian Basin natural gas, pipeline infrastructure, and permissive local regulations. If the moratorium pushes out longer than developers expect, the logical response is not to abandon Texas — it's to build on-site power, which is already the dominant trend for hyperscale AI facilities nationwide. The timeline for a moratorium extension, potentially through mid-2027, is simply too short for that shift to become a panic play. But it is long enough to reinforce a direction the market was already moving.

That brings me to the companies investors actually own.

NRG Energy has been the darling Texas data center play this cycle — a stock that's up 87.6 percent over two years, built on the thesis that AI-driven electricity demand in ERCOT will fuel a generation build-out. NRG acquired LS Power for $12 billion in January 2026, adding 18 natural gas plants and 13 gigawatts of capacity. They've reserved 5.4 gigawatts of GE Vernova turbines for data center builds through 2032. The bull case was clear: Texas data center demand is secular, and NRG sits at the tap.

But look at the numbers that actually matter. NRG's total debt is $35.09 billion against $4.86 billion in equity — a debt-to-equity ratio of 479 percent. That is extraordinarily leveraged for a power generator. Free cash flow over the trailing twelve months is $348 million, down 75 percent year-over-year. Their dividend payout ratio is 207.8 percent — meaning the company is paying out more in dividends than it earns on an adjusted basis. And the stock has fallen 25.8 percent year-to-date, down 31.5 percent over the last 120 days, trading at $118 against a 52-week high of $190.

The moratorium doesn't cause those numbers. The LS Power acquisition and the aggressive build-out do. But the pause is useful because it forces the question NRG's bulls have been dodging: if the Texas data center build-out slows, what happens to the revenue assumptions supporting a $47.9 billion enterprise value on a company generating $348 million in free cash flow?

I rate NRG Energy as a Sell. The debt load is unsustainable if the demand ramp it was built for gets delayed, and the dividend — while growing for 13 consecutive years — is not supported by current free cash flow. In my opinion, the moratorium is the catalyst that exposes the leverage risk NRG investors have been willing to overlook.

On the data center REIT side, the picture is different but not as rosy as the sector rally suggests. Digital Realty trades at a market cap of $71.7 billion with a forward P/E of 32x and a 2.35 percent dividend yield. Its debt-to-equity ratio is a manageable 63 percent, and the payout ratio sits at 78.5 percent. The stock is up 25.3 percent year-to-date and 14.6 percent on a rolling annual basis. But its trailing twelve-month free cash flow is negative $507 million — still improving year-over-year, but negative nonetheless. Digital Realty is a diversified operator with significant exposure outside Texas, which insulates it from this specific pause.

Equinix is the larger name and the more stretched valuation. At $102.9 billion market cap, it trades at a forward P/E of 72x — nearly triple Digital Realty's forward earnings multiple. Its free cash flow has collapsed 860 percent year-over-year to negative $2.6 billion, driven by $6.5 billion in capital expenditures. The dividend payout ratio is 132.6 percent, meaning the company is paying out more than it earns. The stock is up 36.1 percent year-to-date, and the market has rewarded that momentum despite the cash flow destruction.

Equinix's global diversification — it operates in 24 countries across North America, Europe, Asia-Pacific, and Latin America — means the Texas moratorium is a regional footnote, not a systemic threat. But the valuation tells you how much future success is already baked in. At 72x forward earnings, Equinix needs its data center build-out assumptions to play out flawlessly across every market for the next several years.

I rate Equinix as a Hold. The global footprint protects it from Texas-specific delays, but the valuation and negative free cash flow leave no room for a slowdown in AI spending or a broader interconnection backlog. In my opinion, the stock is priced for perfection across a cycle that is now encountering its first meaningful regulatory friction.

I rate Digital Realty as a Buy relative to Equinix, for its lower leverage, sustainable dividend payout, and more defensible valuation. It still generates positive operating cash flow of $3 billion and carries a more reasonable balance sheet. If the broader data center build-out continues outside Texas — and I believe it will — Digital Realty is the safer vehicle for that growth.

The broader lesson here goes beyond Texas. We are entering a phase where the AI data center boom meets real-world constraints: grid capacity, local community pushback, regulatory audits, water availability, and the physical limits of transmission infrastructure. New York already paused hyperscale data center approvals for a year. Illinois froze incentive agreements. Florida restricted cost-shifting. Texas just joined the trend.

This is not a bear case on AI infrastructure demand. It is a reality check on the timeline. The false narrative that Texas has "broken" the data center pipeline is wrong. The pipeline was never as deep as the interconnection queue suggested. But the regulatory friction is real, and it will separate the operators with committed capital and diversified footprints from the speculative projects that filed interconnection requests on a whim.

For investors who can tolerate the volatility of a sector facing regulatory headwinds, Digital Realty offers the better risk-reward balance. For those still holding NRG Energy based on a Texas data center demand thesis that is now under regulatory scrutiny, it may be time to cut losses and reconsider what the balance sheet can actually support.