TechFEATURE

Why Are Tech Giants Racing to Buy Nuclear Power? Can SMRs Solve AI Data Centers' Electricity Crisis?

N
NathanTechnology Editor · Technical Lead
Published · Updated
Facing a projected jump in global data center electricity demand from about 415 TWh in 2024 to about 945 TWh in 2030, Microsoft, Google, Amazon, and Meta have all signed nuclear power deals — but most rely on small modular reactors (SMRs) that won't deliver meaningful capacity until 2030 or later, and NuScale's 2023 project cancellation shows the economics are still unproven.

How Is Surging AI Data Center Power Demand Pushing Tech Giants Toward Nuclear?

Global data center electricity consumption is projected to rise from about 415 TWh in 2024 to about 945 TWh in 2030, more than doubling in six yearsCITE:E7. The International Energy Agency (IEA) attributes this trajectory to AI workloads, and nuclear power's appeal lies in its ability to supply round-the-clock, carbon-free baseload electricity that intermittent renewables cannot matchCITE:E7. Meta has translated that need into a concrete procurement action, issuing a request for proposals (RFP) seeking 1 to 4 GW of new nuclear generation capacity, including both SMRs and advanced reactors, with delivery targeted for the early 2030sCITE:E4.

How Are Tech Giants Partnering With Nuclear Operators? What Deals Have They Signed?

Microsoft, Google, and Amazon have each signed a distinct nuclear deal, spanning both a restarted legacy plant and multiple SMR venturesCITE:E1CITE:E2CITE:E3. Microsoft signed a 20-year power purchase agreement (PPA) with Constellation Energy to restart Three Mile Island Unit 1, renamed the Crane Clean Energy Center, adding about 835 MW of carbon-free power to the grid, with a 2028 target date; Constellation describes it as the company's largest PPA to dateCITE:E1. Google signed a master development agreement with Kairos Power to deploy a combined 500 MW of advanced nuclear capacity by 2035, with the first reactor targeted for 2030CITE:E2. Amazon led a roughly $500 million Series C-1 investment in SMR developer X-energy and plans to co-build an initial 4-unit, 320 MW project with Energy Northwest, expandable to 12 units and 960 MW, while targeting more than 5 GW of X-energy SMR capacity deployed by 2039CITE:E3. Meta's RFP for 1 to 4 GW remains at the solicitation stage rather than a signed construction dealCITE:E4.

What Are SMRs? How Do They Differ From Traditional Nuclear Plants, and What Is the Global Development Status?

Small modular reactors (SMRs) are generally defined as nuclear units with a capacity of 300 MWe or below, built around factory-based modular manufacturing intended to shorten construction schedules compared with traditional reactors, which typically exceed 1,000 MWe per unitCITE:E5. Despite the design appeal, only a small number of SMRs are currently in commercial operation worldwide, while dozens of SMR designs remain in development, and nearly all of the projects tech giants have signed onto are not expected online until around 2030 or laterCITE:E5.

When Will Tech Giants' SMRs Come Online? What Is the Scale and Timeline?

Google and Amazon have set the clearest SMR timelines so far, both centered on the early-to-mid 2030sCITE:E2CITE:E3. Google and Kairos Power target a first reactor online in 2030, building toward the full 500 MW by 2035CITE:E2. Amazon and X-energy target the initial 320 MW Energy Northwest project online in the early 2030s, with a longer-range goal of exceeding 5 GW of deployed capacity by 2039CITE:E3.

CompanyNuclear PartnerCapacityTypeTarget OnlineSource
MicrosoftConstellation Energy835 MWRestarted legacy reactor2028CITE:E1
GoogleKairos Power500 MW (first reactor 2030)SMR2030–2035CITE:E2
AmazonX-energy320 MW → up to 960 MWSMREarly 2030s (5 GW target by 2039)CITE:E3
MetaRFP stage, unnamed1–4 GWSMR + advanced reactorsEarly 2030sCITE:E4

How Risky Are SMR Economics and Licensing? What Does History Warn?

The clearest cautionary precedent is NuScale's flagship SMR project, which was terminated in November 2023 after being the first SMR design approved by the U.S. Nuclear Regulatory Commission (NRC)CITE:E6. The 6-unit, 462 MW project with utility group UAMPS was cancelled due to rising costs and insufficient subscriber commitment, with its target electricity price having risen from $58 to $89 per megawatt-hour before terminationCITE:E6. That outcome underscores the broader development picture: only a small number of SMRs are in commercial operation globally today, and dozens of designs remain unproven at commercial scaleCITE:E5.

What this means: the electricity gap the IEA projects — from about 415 TWh to about 945 TWh by 2030 — arrives on a faster clock than any of the announced nuclear deals can fillCITE:E7. Microsoft's 835 MW restart is the only project targeting delivery by 2028CITE:E1; Google's and Amazon's SMR commitments stretch to 2030, 2035, and 2039CITE:E2CITE:E3; and Meta's 1–4 GW request remains unsolicitedCITE:E4. NuScale's 2023 cancellation, following NRC design approval and a price that rose from $58 to $89 per megawatt-hour, shows that even a fully permitted flagship SMR project can still fail on cost and subscriber groundsCITE:E6.

📊 Evidence

FAQ

How Is Surging AI Data Center Power Demand Pushing Tech Giants Toward Nuclear?

Global data center electricity consumption is projected to rise from about 415 TWh in 2024 to about 945 TWh in 2030, more than doubling in six yearsCITE:E7.

How Are Tech Giants Partnering With Nuclear Operators? What Deals Have They Signed?

Microsoft, Google, and Amazon have each signed a distinct nuclear deal, spanning both a restarted legacy plant and multiple SMR venturesCITE:E1CITE:E2CITE:E3.

What Are SMRs? How Do They Differ From Traditional Nuclear Plants, and What Is the Global Development Status?

Small modular reactors (SMRs) are generally defined as nuclear units with a capacity of 300 MWe or below, built around factory-based modular manufacturing inten…

When Will Tech Giants' SMRs Come Online? What Is the Scale and Timeline?

Google and Amazon have set the clearest SMR timelines so far, both centered on the early-to-mid 2030sCITE:E2CITE:E3.

📎 Sources

  1. constellationenergy.com
  2. kairospower.com
  3. x-energy.com
  4. sustainability.atmeta.com
  5. world-nuclear.org
  6. utilitydive.com
  7. iea.org

Related data

Author's TakeNathan

The timelines laid out here separate the tech giants into two tiers: Microsoft's 835 MW restart is the only deal with a 2028 delivery date, because it revives an existing licensed reactor rather than building new SMR technology. Google, Amazon, and Meta are all betting on SMRs instead, and their own numbers show why that bet is slower — first reactors in 2030, full build-outs stretching to 2035 and 2039, against a demand curve the IEA already sees nearly doubling by 2030. NuScale's cancellation is the relevant precedent to watch, not because it predicts Kairos or X-energy will fail the same way, but because it shows a design that had already cleared NRC approval still collapsed once its price moved from $58 to $89 per megawatt-hour. The single indicator worth tracking over the next few years is whether Kairos's first reactor actually reaches operation in 2030 and whether X-energy's initial 320 MW Energy Northwest phase holds its cost line — those two milestones will show whether SMR economics have actually improved since NuScale, or whether the same cost-and-subscriber pattern repeats.

N
NathanTechnology Editor · Technical Lead

Related

FEATURE

AI Inference Prices Are Falling About 10x a Year — What the Token-Cost Data Actually Shows

Equivalent-capability AI inference prices have fallen roughly 10x per year across multiple benchmarks, with OpenAI's GPT-4o mini priced about 99% below its 2022 predecessor and DeepSeek-R1 undercutting OpenAI's o1 by roughly 27x. Yet Google now processes about 50 times more tokens monthly than a year earlier, and Gartner projects 2025 global generative-AI spending at $644 billion, up 76.4% — proof that cheaper tokens are fueling more total usage and spend, not less.

Nathan ·
FEATURE

China's AI Playbook Under Export Controls: Ahead on Models, a Generation Behind on Chips

China's AI industry shows a split scorecard: on models, DeepSeek-V3's open-weight, MIT-licensed mixture-of-experts architecture and DeepSeek-R1's benchmark parity with OpenAI's o1 preceded a $593 billion single-day NVIDIA sell-off, while the widely cited $5.576 million cost figure covers only the final training run. On hardware, Huawei's Ascend 910C still delivers about 60% of NVIDIA H100's per-chip performance, and HBM supply constraints cap actual 2025 shipments far below estimated capacity, even as US export controls kept tightening through January 2026.

Nathan ·
FEATURE

Private Credit's Climb to $2.5 Trillion: Why AI Data Centers Are Turning to Non-Bank Lenders

Private credit has grown into a global asset class exceeding $2.5 trillion, driven by banks retreating from riskier lending and investor returns climbing toward 12%. A handful of firms — Apollo, Blackstone, Ares, KKR, and Blue Owl — now dominate the market, and Blue Owl has extended it into AI infrastructure with a $27 billion data center financing for Meta, even as the IMF and the Federal Reserve disagree on how much systemic risk this concentration carries.

林紀旭 James Lin ·