The largest hyperscalers operating in the desert Southwest have published water-positive pledge and a standard that credits verified conservation and reuse. The accounting already exists. Here are the projects they could fund now that also move real acre-feet toward the basin's 3 to 3.5 MAF gap.
The AI buildout is landing in the driest basin in the country. Phoenix-area data-center water use is projected to climb toward 3.7 billion gallons a year. Southern Nevada data centers used over 716 million gallons in 2024, drawn almost entirely from the Colorado River. Both figures are reported rather than measured by us, and the Phoenix number is a projection rather than an observation. That looks like the basin's newest problem. It is also its best-funded opportunity.
Every major operator has answered the backlash the same way. AWS reports it is 75% of the way to water positive as of 2025, up from 53% in 2024. These are corporate self-reports measured against each company's own definition of water positive, not independently audited outcomes, and the methodologies differ between them. And a published commitment is not a contract. None of these pledges creates a binding obligation to fund a specific project, on a specific schedule, at a specific price. Anyone underwriting against one would be underwriting a corporate intention, which is why the useful ask is a procurement conversation rather than a capital raise.[1] Microsoft replenished more than it consumed in fiscal 2025.[2] Google has committed over $500 million to water infrastructure.[3] Meta targets 200% restoration in stressed regions and already funds Colorado River Indian Tribes conservation into Lake Mead.[4] They all settle on one standard, Volumetric Water Benefit Accounting, written by WRI and LimnoTech.[5]
That standard is the bridge. It credits the cleanest benefit for two things: verified consumptive-use conservation and recycled water that offsets a fresh diversion. Both are wet water. And the price a funder would pay is not a guess. It is on this project's own transaction record. What that record prices is a short-term, federally subsidised programme, so treat the figures below as the floor of a public conservation programme rather than a market price for permanent water.
$325/AFReclamation’s 2026 system-conservation price, on the record in 2 board actions. Short-term subsidised programme, not durable water
$400/AFThe 2024–25 federal fallowing payment, across 7 MWD board actions
>1MAFFreed by a ~10% cut in basin agricultural consumptive use, the largest fundable block
75%How far AWS is toward water-positive by 2030. The last quarter is the hardest.
The full priced record is on the documented record. These are the numbers boards actually vote, not model prices.
The strongest reuse play is not an abstract idea. It is a slate of existing basin projects with a capacity, a cost, a funding gap, an operator, and a data-center cluster next door. A funder buys a slice, a treatment train or a pump station or a pipeline segment, and takes a clean credit for the diversion it offsets. Every project below is real and public.[6]
| Project | Capacity | Status | Cost / gap | Nearest cluster | Why the credit is clean |
|---|---|---|---|---|---|
| Pure Water Southern California MWD SoCal + LA County Sanitation | ~150 MGD (~168,000 AFY) | Demo operating. Final EIR certified early 2026. Deliveries ~2032. | $4.8-9B total, multi-billion gap | Inland Empire | Offsets Colorado River and State Water Project imports. Clean wet-water credit. |
| Pure Water San Diego City of San Diego | 83 MGD (~93,000 AFY) | Phase 1 ~85% built. Phase 2 planning. | $5B through 2035, plus $130M overrun | San Diego | San Diego is 85% imported water. Reuse offsets the Colorado River share. |
| Phoenix Advanced Water Purification City of Phoenix | ~14,000 AFY combined | Design phase. Online early next decade. | $179M funded, ~$926M planned (reported) | Phoenix metro | New local potable supply offsets river diversions and groundwater. |
| Scottsdale Water Campus Scottsdale Water | up to 20 MGD (~22,400 AFY) | Operating since ~1999. Upgrade needed. | ~$51M immediate compliance need | Phoenix metro | Potable reuse plus recharge. Direct offset, already proven. |
| Tucson potable-reuse demo Tucson Water | 2.5 MGD (~2,800 AFY) | Council approved Jan 2025. Online early 2030s. | $86.7M funded. Scale-up needs more. | Tucson | New local supply offsets Colorado River demand. |
| SNWA return-flow model Southern Nevada Water Authority | ~99% indoor water recycled to Lake Mead | Operating for decades | Ongoing opex and upgrades | AWS North Las Vegas ~8 mi | Every gallon returned lets Nevada withdraw another above its 300,000 AF. |
| Aurora Prairie Waters Aurora Water | 10,000 AFY now, 50,000 AFY ultimate | Operating since 2010. Expansion planned. | $653M built. ~$6.5M expansion gap. | Denver-Aurora | Indirect potable reuse. Drought-resilient offset. |
| Gila River Indian Community reuse Gila River Indian Community | ~20,000 AFY reclaimed, +125,000 AFY conserved | Federal agreements signed | $83M pipeline plus conservation funds | South of Phoenix | Direct river conservation and reuse, with tribal water-security co-benefit. |
Ranked by fundability, basin impact, and proximity: Arizona’s Phoenix, Scottsdale, and Tucson purification projects sit in the densest cluster and the most stressed lower-basin state, and Scottsdale has an immediate need near $51M. Pure Water Southern California is the largest volume and the largest gap. SNWA’s return-flow model is 8 miles from AWS North Las Vegas and is the cleanest proof that reuse means wet water put back. Figures are agency- and press-reported. Read the source before you quote a number.
Each project type scored one to five on how fundable it is by a hyperscaler water team, how many acre-feet it moves, and how ready it is to contract now. The product sorts them. The top three are the fund-it-now plays.[7][8]
Cleanest possible credit. Pay farmers to cut measured consumptive use, shepherd the water to Lake Mead. Ag is about 75% of human use, so a ~10% cut in agricultural consumptive use is about 1 MAF a year. The price is already on the record at $325 to $400 an acre-foot. Read that as the price of a SHORT-TERM, federally subsidised programme, one to three years, and not as the cost of durable water. A funder asking a district to give up optionality for twenty or fifty years is asking for a different and dearer transaction.
The cheapest, most on-brand play for AI and data companies. It does not save water by itself. It verifies every other credit and lowers the cost of the whole market. OpenET is already the standard USBR and the Upper Colorado River Commission use.
Fund infrastructure in your own backyard. Recycled water offsets a fresh river diversion, metered and clean under the accounting. The named projects below are shovel-ready and capital-hungry, sitting next to the data-center clusters.
Tangible and popular, but credit only the net basin-wide reduction, never the gross savings. Seepage from an earthen canal is often someone's return flow. Meta and CRIT already ran the model.
Vast storage. Arizona has banked over 4.4 MAF. But it is storage, not new supply, and it does not cut the structural deficit unless the recovered water retires other demand.
Real resilience value, wrong instrument for a volumetric portfolio. The downstream yield is diffuse, slow, and model-estimated.
Contested science and, in Colorado, salvaged water belongs to the river under priority. It was removed from the water math in the funder brief for exactly this reason.
The claim is narrow on purpose, because a credit a critic can pick apart is worse than none. Arizona already says no to weak-additionality paper water. Tucson's council voted down the Project Blue data center, and the developer subsequently committed to a zero-water design. The vote is the part on the record. We do not claim the commitment secured an approval. So here is the honest split.
Conservation and reuse. Verified consumptive-use conservation paid on measured ET and shepherded, and recycled water that offsets a fresh diversion. Both are wet water. Fund without asterisks.
Efficiency and recharge. Efficiency counts only the net basin-wide reduction, never gross savings. Recharge counts the avoided diversion when stored water is used in a shortage, not the act of storing. Fund with monitoring.
Removal and restoration. Phreatophyte removal and forest restoration have real co-benefits and diffuse, slow, hard-to-measure water yield. Wrong instrument for a volumetric portfolio.
Put it together. The basin’s biggest new water consumers are its most motivated funders. The standard to pay them exists. The price is on the record. And the named projects are shovel-ready, next to their own data centers. Each company offsetting only its own gallons does not close a 3 to 3.5 MAF gap. Pooled, at scale, and verified in public, it can.