Basin 2.0, measured.

Why we’re not proposing floating solar

We are not proposing floating solar on Lake Mead or Lake Powell in any meaningful amount. Siting millions of panels in deep, canyon-bound reservoirs with 500-foot depths and 100-to-180-foot water-level swings runs into hard structural, anchoring, and mooring problems. And a realistic, recreation-compatible 10% coverage saves only about 60 to 75 thousand acre-feet a year per reservoir, a tiny fraction of the deficit. So we keep floating solar out of the water-conservation core.

Canal-top solar remains a pragmatic, niche alternative, such as the Gila River Indian Community's Casa Blanca canal (2024) or California's Project Nexus, which cut evaporation from the canal surface and reduce weed growth [23][24]. However, the water saved is minimal (canals are a tiny share of basin-wide evaporation), and water authorities like the Central Arizona Project have declined canal-top solar because the physical structures block crucial visual canal inspections. Floating or canal-top solar remains a minor resilience play for tribal or local canal operators, not a basin-scale solution.

Replacing lost hydropower with land-based solar and batteries

As the reservoirs fall, Hoover (2,080 → ~1,600 MW) and Glen Canyon (1,320 → ~800 MW, and it stops entirely below elevation 3,490 ft) are losing firm, cheap, dispatchable power and the grid services that come with it. Utility-scale land-based solar-plus-storage, photovoltaic panels (PV) paired with battery energy storage (BESS), is the definitive answer. By co-siting large-scale, land-based solar arrays and Battery Energy Storage Systems (BESS) co-located directly at the dams' existing transmission nodes, we can produce abundant, water-blind electricity. This reuses the dams' massive transmission lines with $0 in grid costs, fully backfills the lost hydropower, and decouples reservoir releases from peak electrical grid demands. Once solar-plus-storage carries the firm power, releases no longer have to be timed for the evening peak, allowing operators to release water purely for delivery, ecology, and downstream temperature control (the fix to the Grand Canyon bass problem) with zero power-revenue penalties.

Pumped hydro is a complement, not a substitute: it is a net energy consumer (70–80% round-trip), so it firms and shifts power but needs solar to charge it. Real proposals exist (Daybreak’s 2.2 GW Navajo Energy Storage near Powell, an LADWP project at Hoover), but open-loop designs add a new evaporating reservoir, ~3,000 acre-feet a year for a 500-acre pond, a water cost in a basin that can’t afford one. Closed-loop avoids most of that but costs more. Net: lead with PV+BESS at the dams, and use pumped hydro only closed-loop, for long-duration firming.

What the last 25 years would have looked like. We backtested it. Because Glen Canyon and Hoover releases are set by water-delivery law (a narrow ~7.5–8.2 MAF band), their generation fell not because less water moved but because the reservoirs dropped and the head collapsed, combined dam output is down about 26% since 2000, a cumulative ~39 TWh of lost generation. A flat 1.8 GW of PV on the dams’ existing interconnects (~4.6 TWh/yr, water-blind) would have covered that shortfall roughly 1.8× if built in 2010, displacing the market gas WAPA buys during drought (~1.8 MtCO&sub2;/yr). It would not have changed the volume released, that is fixed by delivery law, but it would have made every hard call of the drought era (steady flows for the Grand Canyon, cool-water bypass for the native fish, letting Powell fall below the power pool) a water decision instead of a power-revenue sacrifice.

The transmission is not empty, and that changes the play

An early version of this work claimed ~3.5 GW of stranded transmission at the dams. An interconnection review corrected that, and we are keeping the correction visible. Here is the honest version. Hoover and Glen Canyon carry firm hydropower contracted to utilities across seven states through 2057, plus the ancillary services that keep the Southwest grid stable, so their lines are not idle. But as the reservoirs fall the dams generate less, and the region has to import replacement power [26]. Solar that backfills the lost hydropower, on the order of ~1 GW across the two dams, uses the interconnection capacity the shrinking hydro freed and cuts the imported gas. Oversize the array past the line limit and route the midday excess into storage instead of clipping it, and the battery carries it through the evening peak, the hours the strained grid needs most and exactly when the dams used to be dispatched. That part is real, though it is non-firm and six-hour storage does not cover a deep overnight or a winter lull. What is not real is the larger 3.5 GW figure, which counted capacity that is not free. Beyond the backfill, the one repurposable corridor is the retired Navajo Generating Station’s ~800 miles of 500-kV line, already contested by a 2.2 GW pumped-storage project and a 750 MW solar project and controlled by the Navajo Nation [27][28]. So the honest play is a ~1 GW backfill at the dams, plus new solar co-located with the data-center load, competing for a seat at NGS as a partner, not a discovery of empty transmission.

The regulatory key is FERC’s surplus interconnection service (Order 845): co-locate solar and storage at a derated generator’s existing interconnection and use its headroom on an expedited study, months, not the five-plus years of the standard queue, where only ~14% of projects ever finish [29]. It has been done on retiring coal (Comanche, Sherco) and gas. The catch is honest: surplus service is often non-firm (curtailed first in congestion, which the battery softens), and it needs agreements with WAPA, the Navajo Nation, and the plant’s ownership consortium.

Sized to the load, with new interconnection on a real queue

Take it as an illustration of scale, not a shovel-ready plan. A build on the order of ~5 GW of solar with six-hour storage, roughly $12B, co-located with data-center load and new or repurposed interconnection, would throw off about 13 TWh a year. That is enough to replace the dams’ fading output (~4.6 TWh), carry ~700 MW of round-the-clock data-center load, and still leave ~2.2 TWh to pump, treat, and bank water. The catch is the one the transmission review makes plain: this capacity is not sitting free on the dams’ existing lines. It needs new interconnection on a multi-year Western queue, or co-location with the load, or a negotiated partnership at the retired Navajo plant. Energy is the easy part. Grid access is the hard part [29][28].

Seasonality is the catch, and water is the answer. Desert solar runs ~34% capacity in June and ~13% in December, and batteries smooth the day but not the season. Flexible water loads, aquifer recharge, CAP pumping into Lake Pleasant, groundwater and desalination that can ramp, soak up the summer surplus and bank it underground as water, covering the winter dip that no battery can. Water infrastructure becomes the seasonal store [30][31].

And it lands where it’s needed most. The Navajo Generating Station’s 2019 closure cost the Navajo Nation ~$40M a year and up to 80% of the Hopi general fund, and ~800–1,000 coal jobs. Rebuilding on their own transmission, on the already-disturbed mine and plant land, which sidesteps the desert-tortoise conflict and earns the IRA “energy community” 10% bonus, turns a stranded coal asset into ~28,000 construction jobs, tens of millions a year in tribal revenue, and ~5 MtCO&sub2; avoided [27][32].

And the same abundant solar can make water

Abundant carbon-free PV is not only a power source, it is a water source, water and energy substitute at the margin, and in the desert the sun is the cheap input. One gigawatt of desert PV can desalinate on the order of 876,000 acre-feet of brackish or drainage water a year (~2,500 kWh/AF), and at ~$0.10/kWh the energy cost of reuse or brackish desal falls to ~$100–250/AF [31][30][33]. Ranked by water per unit of cheap energy:

Put together, solar-powered desalination, reuse, and drainage treatment could add or free 0.3–0.5 MAF a year, the same order as the whole conservation program, with energy as the input instead of someone’s forgone water. Brine disposal is the binding inland constraint. The coastal exchange sidesteps it but needs interstate and binational deals. The binding constraints are permitting and brine disposal, not energy, and the credible near-term number is far smaller than the long-run ceiling (see the engineering reality). This is the thread that ties the hyperscalers’ energy build-out to the basin’s water: the PV they need for their own load, on transmission that already exists, throws off cheap power that backfills the dams, frees their operation for temperature and ecology, and manufactures water.

The biggest drought-proof source is the water we already threw away

The flagship use of firm solar-plus-storage is wastewater reuse and aquifer recharge, the Orange County model. OCWD purifies 134,000 acre-feet a year of sewage that used to go to the ocean, recharges it into the aquifer, and delivers it cheaper than imported water ($487–1,073/AF vs $1,130–1,600), and it is drought-proof, because wastewater flows whether or not it snows [34]. Scaled across the basin, a UCLA analysis finds lifting reuse to 40% would create ~900,000 AF a year of new supply and 50% would create 1.3 MAF. Metropolitan’s Pure Water Southern California alone targets 168,000 AF [35]. The honest catch: net-new counts only where effluent is currently discharged, Southern California’s ocean outfalls are the prize. Las Vegas already credits its return flow to Lake Mead and Phoenix already recycles 97%.

As a portfolio, the surplus energy buys a credible 1.5–3 MAF a year of new-or-saved water: reuse-to-recharge (firm power), plus cheap flexible wins that barely touch the surplus, precision and deficit irrigation (a 26–65% cut on the 70–80% of water agriculture uses, often at a net energy saving), aquifer storage and recovery (Arizona’s water bank already holds 4.4M AF of credits), and AI leak detection. Energy stops being the limit. Capture infrastructure, capital, and brine are.

One tier pays for itself, the rest is public goods

Honest finance beats a flattering headline. Conservation is the winner. The ~$400 per acre-foot recent basin deals paid is the payment to the user, not the all-in cost. With measurement, verification, and shepherding to Lake Mead, durable conserved water runs closer to $600 to $1,000 per acre-foot, still the cheapest high-integrity water in the basin, and genuinely investable if a buyer (say, a corporate replenishment commitment) anchors the offtake. The capital-heavy pieces, the Glen Canyon retrofit, aquifer recharge, ag efficiency, do not pencil as private infrastructure at any honest water value. They are public goods, and they should be funded as grants and public debt, not sold as equity promising double-digit returns. There is no 54% internal rate of return (IRR) here. There is a cheap, real conservation win and a clear-eyed public-goods bill.

Can the beneficiaries finance it, not appropriations?

Partly, and here is the honest split. Of the ~$15.7B of capital, the $12.2B on revenue-capable options (metered recharge and reuse, conservation offtake, measurement fees) can service its own debt if a buyer anchors the demand. The remaining $3.5B is public goods, principally the Glen Canyon retrofit, with no cash flow, repaid by a small beneficiary assessment [37][38]:

How it actually gets financed

Not one mega-bond, but a stack of proven instruments: WIFIA federal loans cover up to 49% of eligible cost at Treasury rates, 35-year, with a 5-year repayment deferral [38]. The balance as tax-exempt revenue and assessment bonds into a US water and sewer muni market that issued $46B in 2025 alone [39]. And Environmental Impact Bonds (the DC Water model) that pay investors on verified acre-feet saved, turning the measurement ledger into the settlement layer [40]. Labeled blue bonds already raise several billion a year, so a staged Colorado River program sits well within market appetite [37].