Water, Not Power, Is the Real Constraint for Texas Data Centers

At the ERCOT Innovation Summit near Austin yesterday, a panelist said something I've heard at every datacenter infrastructure conference for two years straight: "Power is the single most important gating factor in siting a data center."

It's a comfortable claim. Power is one of the biggest constraints, and it's the one that makes headlines, fills conference agendas, and drives billion-dollar procurement deals.

When the Texas pipeline has 40 GW of datacenter-designated private generation capacity in various stages of development, and equipment lead times stretch to 107 weeks for Caterpillar and Cummins generators, power feels like the constraint that matters most.

But power has alternatives. Lots of them.

You can build a gas plant. You can contract for wind and solar. You can install 813 modular generators behind the meter, the way Meta is doing for its 366 MW El Paso campus at a capital cost of $473 million. You can co-locate on pre-interconnected land with a 20-year deal, the way Google and AES structured their Wilbarger County project.

You can go fully off-grid with 1.5 GW of Caterpillar engines on 4,000 acres, the way Joule Capital Partners is doing in Utah. You can contract with Boom Supersonic for some of their 42 MW natural gas turbines when they come to market.

Even outside of gas, wind, and solar there are more options, including hydroelectric or geothermal generation, though those are highly site-specific. There's one promising hydrogen possibility, and if you have the time to wait, you even have nuclear options available.

Power is hard. Power is expensive. But power has many alternatives for how it can be engineered, financed, contracted, and built on a timeline.

Water doesn't. Particularly in Texas.

How much water do Texas data centers use?

A single megawatt of datacenter capacity consumes between 1 million and 6.7 million gallons of water annually for cooling, depending on cooling technology, climate, and rack density. In Texas summers, where temperatures push evaporative systems toward peak consumption for months at a stretch, facilities cluster toward the high end of that range.

Put together, all Texas datacenters consume roughly 25 billion gallons of water annually, according to the Houston Advanced Research Center (HARC). That sounds like a huge number, but it's about 0.4% of the state's total water use.

By 2030, HARC projects that number could reach 29 to 161 billion gallons, depending on how much of the pipeline gets built and which cooling technologies are deployed. The University of Michigan's Science, Technology, and Public Policy program puts the upper bound even higher, at 399 billion gallons.

To put that in context: Texas uses roughly 5 trillion gallons of water a year. Agricultural irrigation accounts for 55% of it. Texans use an estimated 700 to 800 billion gallons a year for lawn irrigation. Even the most extreme 2030 datacenter projection is half of what we use to water our lawns.

So this isn't, and won't be, a statewide water crisis. It's a siting issue. Datacenters don't draw 0.4% of Texas water evenly across the state. They draw millions of gallons a day from the same aquifers that supply their neighboring ranchers and municipalities.

There are 6.5 GW of datacenter capacity under construction in Texas right now. Another 22 GW sits in the pipeline, according to the Hyperscale News Texas Datacenter Pipeline Tracker. Behind each of those gigawatts is a cooling system that needs water — and the water has to come from somewhere specific.

Google's global datacenter fleet consumed 8.1 billion gallons of water in 2024. Its single facility in Council Bluffs, Iowa, used 1 billion gallons by itself. Meta used 813 million gallons across its portfolio in 2023, with 95% going to datacenter cooling. Both figures predate the current wave of AI-driven capacity expansion and represent a floor, not a ceiling.

Where is water available for Texas data centers?

The regions most attractive for datacenter development — cheap land, abundant wind and solar, fewer density constraints — are the same regions where water is already spoken for. West Texas sits on the Ogallala Aquifer, which is depleting at an annual volume equivalent to 18 Colorado Rivers.

The Texas Water Development Board's State Water Plan projects a 25% decline in groundwater availability between 2020 and 2070. In the thirteen-county region centered on Lubbock, half the available groundwater will be gone by 2040. Hale County expects to lose two-thirds of its groundwater by then.

Texas has 101 groundwater conservation districts, each with independent authority to set pumping limits, grant or deny permits, and define "beneficial use." These are local bodies, often staffed by people whose families have farmed the same land for three generations, making real-time decisions about who gets water and who doesn't.

Four new datacenters are planned for the Texas Panhandle, in Amarillo, Turkey, Pampa, and Claude. The developers who engage these districts early, demonstrate responsible water sourcing, and build community benefit-sharing into their proposals will get permitted. The ones who show up with a power plan and no water strategy won't.

The Texas Commission on Environmental Quality is now reviewing water use plans for all new datacenters exceeding 5 MW. That's a signal: get your water story right before you file, not after.

What happened when Arizona datacenters ran into water limits?

Texas isn't the first state where water has shaped datacenter strategy. Arizona got there first. Mesa passed the Large Customer Sustainable Water Allowance ordinance, creating a water "budget" for any facility projecting more than 500,000 gallons of daily consumption. Phoenix and Avondale passed similar caps.

Federal officials declared a Tier 2a shortage for the Lower Colorado River Basin, cutting Arizona's allocation by 21%.

The result: operators who planned for water early are running without disruption. The ones who treated it as an afterthought are retrofitting cooling systems and renegotiating municipal agreements on someone else's timeline.

Texas has a different water law framework — the rule of capture still governs groundwater in areas without conservation districts — but the planning lesson is identical.

Why is water harder than power for data center siting?

Both power and water are real constraints. Denying that power is difficult would be absurd when 30-50% of pipeline projects face delays from hardware bottlenecks and 107-week equipment lead times are standard.

But the constraints aren't symmetrical.

Power has a deep menu of alternatives. Natural gas, wind, solar, battery storage, behind-the-meter generation, grid interconnection, co-location with existing plants, aeroderivative turbines. ERCOT's installed capacity is approximately 155 GW across every fuel type. The state added 13.9 GW of battery storage entering 2026, nearly doubling in a single year. The supply side of power is competitive, innovative, and well-capitalized.

Water has a much shorter list. You can recycle produced water. You can switch from evaporative to air-cooled systems, trading energy efficiency for water independence. You can adopt liquid cooling or direct-to-chip systems that reduce (but don't eliminate) water needs. Microsoft has moved to zero-water designs for new builds.

But here's what matters: zero-water cooling technology exists in the lab and in a handful of next-generation facilities. It is not what's being deployed across the 6.5 GW under construction or the 22 GW in the pipeline. The vast majority of projects currently permitted or in development in Texas are spec'd for conventional evaporative cooling. The constraint isn't theoretical availability of better technology. It's what's actually being built right now, at scale, with water requirements already baked into the engineering.

And even the demand-side solutions only reduce how much water you need. They don't create water that doesn't exist.

You can't build an aquifer. You can't permit a river into existence. You can't finance groundwater recharge on a 107-week timeline. Statewide, datacenter water use is a rounding error. In a specific water-stressed basin where four new facilities are competing for the same declining aquifer, the constraint isn't capital or engineering. It's geology.

What do Texas regulations require for data center water planning?

The PUCT's proposed interconnection rule for large loads (Project 58481, implementing SB 6) requires developers seeking 75+ MW connections to disclose plans and progress on "geotech, water, wastewater, and gas" studies as part of the intermediate agreement. An officer attestation is required. Quarterly updates are mandatory.

Water studies are baked into the regulatory process for connecting to the grid. The state already treats water as a material siting factor for large load interconnection.

But for obvious reasons, no panel at the ERCOT Innovation Summit discussed water in any depth. The conference agenda was wall-to-wall power: generation capacity, transmission planning, interconnection queues, financial security requirements. There was plenty of awareness of and discussion on gas and behind-the-meter electricity generation, so it wasn't an issue of staying only in ERCOT's responsibility area. Water just wasn't seen as directly relevant or applicable.

That gap between what the regulatory framework actually requires and what the industry talks about is significant. The developers who understand that gap have a structural advantage. The ones who don't will discover it during permitting, when it's expensive to discover anything.

What are zero-water cooling and water-first site selection?

The operators who'll build the most durable datacenter portfolios in Texas are the ones who started their site selection with a water plan instead of a power map.

Microsoft's next-generation datacenter design, launched in late 2024, uses a closed-loop system that consumes zero water for evaporative cooling. It's a meaningful shift from the industry standard, and it happened because Microsoft saw their global water consumption rise 34% between 2021 and 2023, driven almost entirely by datacenter expansion. Their Arizona facilities still run a WUE of 1.52 liters per kilowatt-hour. Their newer designs target near-zero.

Liquid cooling and direct-to-chip systems are gaining adoption in high-density AI workloads, where heat loads per rack have outpaced what traditional air cooling can handle anyway. These systems use dramatically less water, but they require purpose-built facilities. Retrofitting an existing evaporative-cooled campus is neither cheap nor fast.

The site selection calculus for the most sophisticated operators (based on conversations I've had with developers active in the Texas market) now runs in a specific order: water availability first, then power, then land, then permitting timeline.

That's a reversal from how most of the industry still operates, and it's why certain East Texas locations with abundant surface water and Trinity Aquifer access are attracting serious interest despite higher power costs and transmission constraints.

How do data center developers secure water rights in Texas?

The developers who treat water as an afterthought are building on a foundation they haven't tested. The ones who secure water rights early, invest in water-efficient cooling, and build relationships with groundwater conservation districts before filing permits are the ones whose projects stay on timeline. Water planning isn't a just a regulatory burden. It can be a competitive moat. The developer who locks in sustainable water access for a 20-year operating life has an advantage that can't be replicated by a competitor who shows up two years later to the same basin.

Power is a constraint dealt with through capital and engineering. Water is a constraint that requires more foresight. Both are solvable. But water rewards the developers who move first, because the supply is finite and the rights don't replenish on the same timeline as a generation queue.

The 40 GW of power coming to Texas will get built. The developers who pair that power with secured water access, efficient cooling technology, and strong local relationships will operate for decades. That's the conversation that should've been on stage at ERCOT yesterday — and the policy question underneath it is whether ERCOT or the PUCT should require verified water availability disclosure as a precondition for interconnection study approval, before another 22 GW of projects enter the queue without anyone asking where the water comes from.

---

*Hyperscale News publishes daily news and weekly recaps of water, energy and political changes impacting the Texas datacenter pipeline, including water availability tracking by site for developments in the interconnection queue. Explore the Texas Datacenter Siting Map at hyperscalenews.com/map, or subscribe at hyperscalenews.com.*

---

*For datacenter developers navigating water strategy, groundwater conservation district engagement, or regulatory positioning in Texas, contact JD Key Consulting at jdkey.com.*

---

Sources & Data References

ERCOT Innovation Summit:

1. Author attended, Austin, TX, March 31, 2026

Texas Datacenter Pipeline:

1. 40 GW datacenter-designated private generation (Hyperscale News weekly recaps, Feb 2026)
2. 6.5 GW under construction, 22 GW in pipeline (Hyperscale News weekly recaps)
3. 107-week equipment lead times for Caterpillar, Cummins, Rolls-Royce (Hyperscale News)
4. Meta El Paso: 813 modular generators, 366 MW, $473M (Meta Platforms SEC filings and project announcements, 2024-2025)
5. Google/AES Wilbarger County: 20-year co-located deal (AES Corporation project disclosures, 2024)
6. Joule Capital Partners Utah: 1.5 GW, 4,000 acres, off-grid (company announcements, 2025)
7. 30-50% project delay rate (industry surveys and developer interviews)

ERCOT Grid:

1. ~155 GW total installed capacity (ERCOT Capacity, Demand, and Reserves Report, 2025)
2. 13.9 GW battery storage entering 2026 (ERCOT generation interconnection data)

Water Consumption:

1. 1M-6.7M gallons per MW per year for cooling range (U.S. Department of Energy; Lawrence Berkeley National Laboratory datacenter cooling efficiency studies)
2. Texas datacenters current: ~25B gallons, ~0.4% of state total (Houston Advanced Research Center, 2026)
3. Texas datacenters 2030: 29-161B gallons, up to 2.7% of state total (HARC whitepaper)
4. Texas datacenters 2030 upper bound: 399B gallons (University of Michigan STPP — single academic estimate)
5. Texas total water use: ~5 trillion gallons/yr; agriculture 55%, municipal 31% (TWDB 2020 estimates)
6. Texas lawn irrigation: ~700-800B gallons/yr estimated (TWDB seasonal water use report; 58% of residential use is outdoor, 80-90% of outdoor is lawns)
7. Google: 8.1B gallons in 2024, Council Bluffs 1B alone (Google Environmental Report 2024)
8. Meta: 813M gallons in 2023, 95% to datacenter cooling (Meta Sustainability Report 2023)
9. Microsoft global water consumption growth: 34% increase 2021-2023 (Microsoft Environmental Sustainability Report 2023, 2024)
10. Microsoft zero-water design: launched late 2024 (Microsoft datacenter infrastructure announcements)
11. Microsoft WUE: 1.52 L/kWh Arizona (Microsoft Environmental Sustainability Report 2024)

Texas Water:

1. Ogallala Aquifer depletion: 18 Colorado Rivers annual equivalent (USGS and Kansas Geological Survey)
2. State Water Plan: 25% groundwater decline 2020-2070 (TWDB 2022 State Water Plan)
3. Lubbock 13-county region: half groundwater gone by 2040 (TWDB regional planning data)
4. Hale County: two-thirds groundwater gone by 2040 (TWDB regional planning data)
5. ~100 groundwater conservation districts (TWDB registry)
6. TCEQ reviewing water plans for datacenters >5 MW (Texas Commission on Environmental Quality)
7. Four Panhandle datacenters planned: Amarillo, Turkey, Pampa, Claude (developer filings and local reporting, 2025-2026)

Arizona:

1. Mesa Large Customer Sustainable Water Allowance: 500K gal/day threshold (City of Mesa ordinance)
2. Federal Tier 2a shortage: Arizona -21% (Bureau of Reclamation, Lower Colorado River Basin operations)

Regulatory:

1. PUCT Project 58481 (16 TAC 25.194): water study disclosure requirement (PUCT rulemaking docket)
2. SB 6 (89th Texas Legislature): large load interconnection standards (PURA 37.0561)

Site Selection:

1. Water-first prioritization: developer interviews conducted by author, 2025-2026

Hyperscale News Data Products:

1. Texas Datacenter Siting Map: active/planned sites overlaid with water availability timelines, groundwater conservation district boundaries, and political jurisdictions (hyperscalenews.com/map)