The hyperscalers built nuclear deals nobody can finance. Here's what happens when reality hits.

In October 2024, Google and Kairos Power announced a master development agreement to deploy 500 megawatts of advanced nuclear capacity by 2035. Two days later, Amazon and Energy Northwest announced plans for small modular reactor deployment in Washington state. By mid-2026, hyperscalers had signed agreements covering 9.8 gigawatts of nuclear generation, nearly all of it from reactors that do not yet exist.

The announcements generated headlines about clean energy and AI sustainability. The capital commitments tell a different story. TerraPower raised $650 million in June for a Natrium plant estimated to cost $9.4 billion. X-energy secured $500 million in October 2024 for reactors projected to require multiples of that sum. Zero SMRs are powering data centers as of 2026, and the earliest first-of-a-kind units target 2030 for initial operation.

What the agreements actually commit

Amazon's announcement with Energy Northwest promised to "fund the initial feasibility phase of an SMR project" in exchange for "the right to purchase electricity from the first project (4 modules), which is expected to generate 320 megawatts of energy capacity." The agreement grants Amazon an option, not an obligation. If generation costs exceed market rates, Amazon can decline to purchase power.

The Kairos Power and Google agreement describes selling "energy, ancillary services, and environmental attributes to Google under Power Purchase Agreements." Neither announcement specified pricing terms or minimum purchase volumes. Power purchase agreements function as financing commitments, signaling intent to offtake electricity at agreed rates if the plant operates, but they do not fund construction directly.

Flux Kinetics analysis identified 5 to 7 gigawatts of hyperscaler nuclear capacity scheduled for first power between 2030 and 2035. Alatirok's comprehensive mapping tallied 9.8 gigawatts across Google, Amazon, and Meta commitments. Google's Kairos units target 2030 for the first reactor and 2035 for the full 500-megawatt fleet. Amazon's X-energy partnership envisions multi-gigawatt deployment stretching toward 2039, with initial units around 2030. Meta's agreements with TerraPower and Oklo cluster in the 2032 to 2035 window.

The construction capital gap

TerraPower's Natrium plant carries a $9.4 billion cost estimate for 345 megawatts of electrical output. The company raised $650 million in June from investors including NVIDIA's NVentures, Bill Gates, and Hyundai. That sum represents 6.9 percent of the estimated construction cost. X-energy secured approximately $500 million in October 2024 from a consortium including Amazon's Climate Pledge Fund, Citadel founder Ken Griffin, Ares Management, and the University of Michigan. A subsequent round in 2025 increased total financing to $700 million.

NuScale's Idaho project was cancelled in 2023 when its $9.3 billion cost estimate for 462 megawatts proved too expensive for electricity purchasers. The Natrium estimate sits at roughly the same total cost for 25 percent less capacity. High-temperature gas-cooled reactors like X-energy's Xe-100 carry even higher cost estimates, with projected electricity costs double or more those of large nuclear plants.

Amazon and Google announced data center capital expenditures Their nuclear investments represent a fraction of one percent of annual data center spending. NextEra Energy CEO John Ketchum described SMRs as "an opportunity to lose money in smaller batches."

Public funding fills the gap

In September 2024, Constellation Energy and Microsoft announced an agreement to restart the Three Mile Island reactor, claiming it was "entirely a private agreement" involving "no public funds." The Washington Post revealed that Constellation had applied to the Energy Department for a $1.6 billion federal loan guarantee. In November 2025, the Energy Department loaned Constellation $1 billion, and in June 2026, it announced $17.5 billion in American Nuclear Supply Chain Loans to accelerate deployment of ten large-scale commercial reactors.

The loan programs address a structural problem: electricity from new nuclear reactors costs approximately three times the corresponding energy from solar or wind plants per unit. BWRX-300 SMRs are estimated to produce power at $140 per megawatt-hour for the first reactor, declining to $80 per megawatt-hour for subsequent units. Wind and solar range from $40 to $86 per megawatt-hour. Cost trends diverge over time: nuclear becomes more expensive and renewables cheaper.

Dennis Wamsted, co-author of a 2024 IEEFA report titled "Small Modular Reactors: Still too expensive, too slow and too risky," said: "I still think that's one of the best-titled reports we ever wrote. I think it's just as overhyped as it was a few years ago. There is a shiny new object mentality to SMRs. They're going to work perfectly right out of the box."

Timeline slippage and regulatory reality

In 2020, the Energy Department declared a goal of having the Natrium reactor and Xe-100 reactors operational "within 5-7 years." In November 2025, a TerraPower executive promised a "commercial operation delivery date, power on the grid" by "2031." TerraPower's Natrium reactor cleared its final NRC safety evaluation in December, but commercial operation now sits a decade beyond the original 2020 projection.

No SMR is powering a data center anywhere in the world as of 2026. The earliest first-of-a-kind units target around 2030 for Google's Kairos Power reactors and some Oklo units, with most larger deployments landing in the early to mid-2030s. Forty-five gigawatts of nuclear capacity sits under contract with zero online. The contracts represent financing structures and offtake commitments, not operating assets.

There are no small modular reactor projects currently under construction in the United States. Estimated times of arrival have been pushed back. The gap between announcement timelines and regulatory, construction, and commissioning reality continues to widen.

What is actually meeting data center demand

Within the United States, renewable energy output grew from 547.7 terawatt-hours in 2020 to 926.9 terawatt-hours in 2025, excluding large hydro plants. The increase of 379.2 terawatt-hours outpaced growth in demand from data centers. Nuclear energy output declined marginally from 831.5 terawatt-hours in 2020 to 826.1 terawatt-hours in 2025. Nuclear power has played no role in fulfilling increased energy demand from data centers during this period.

If output from existing nuclear reactors, such as the Three Mile Island restart, is directed toward data centers, that electricity must be diverted from the general public or other traditional consumers, not added as new generation. The only source of energy growing faster than data center demand is renewable energy. Solar and wind installations continue to scale at rates that SMR construction timelines cannot match.

The hyperscaler nuclear announcements function primarily as public relations, according to analysis from the Bulletin of the Atomic Scientists. The investments, typically containing adjectives like "safe," "reliable," "clean," and "affordable," divert attention from the negative environmental impacts of data center operations. Such commitments are more likely funded from public relations budgets than operational capital expenditures.