Big Tech AI nuclear commitments and data center power demand

Tech giants have pledged over 10 GW of new nuclear power. Corporate demand broadening: Walmart signed 176 MW PPA. South Korea considering new nuclear for AI/chip power. Grid bottleneck intensifies: $130B projects blocked, PJM capacity auction costs surge ($6.3B latest, $29.4B cumulative). New York paused large data-center permits (12 GW queue) via moratorium, a significant counter-narrative that could shift demand to nuclear-friendly states; however, Hochul is now pushing advanced nuclear (8.4 GW target). Texas also sees local opposition. A new counter-narrative: data center opposition is escalating with an 'Arc of Outrage'—$3T in proposed projects and 353 cancellations/delays ($205B). A podcast quantified the backlash: 38 states, 120 moratoriums, $130B in projects held up. New data point: Georgia data center boom sparks protests over rising electricity costs. J.P. Morgan projects $5-7 trillion AI buildout with power as #1 constraint, reinforcing nuclear need. BNEF now forecasts US data centers will consume 20% of electricity by 2035 (106 GW, up 36% from April), with crypto miners pivoting to AI adding 6-12 GW. Indiana launched Advanced Nuclear Ready Community program. Bank of America projects $10T nuclear market. AI data center demand is boosting utility stocks: Constellation's nuclear fleet driving 25% earnings growth, Entergy's Meta deal, and Blue Energy SMR investment. A key counter-narrative: $23-25B cost shift to ratepayers from AI data center deals is drawing scrutiny. Analysts see select utilities primed for AI-driven power boom: Constellation top pick, Entergy and NextEra also highlighted. The timing gap between AI demand and SMR deployment is critical: first AI-dedicated nuclear power will come from a restarted reactor (TMI) in 2027, not SMRs. Utility ETFs up 8% YTD on AI power demand, with XLU offering exposure to merchant nuclear operators. Constellation's uncontracted nuclear inventory smaller than supply gap, giving pricing power. A new risk: data center construction may be masking a broader construction recession; if moratoriums spread, the AI power demand thesis could weaken. Recent articles confirm Big Tech's land-grab near power plants for AI-driven nuclear deals, but also hint that most reactors won't come—a caution on execution risk. UK grid constraints and rising costs are a real bottleneck for AI data centers, with contract solutions being explored. Utilities are shifting to scenario-based planning due to data center load (NV Energy 4x demand by 2045, 80% from data centers). Cleantech deal boom includes nuclear startups X-Energy, Inertia, Blue Energy. Federal government announced 1 GW AI data center at Savannah River Site with 2 GW on-site power, initially gas bridging to advanced nuclear—a major policy-driven catalyst bypassing grid bottlenecks. New local opposition data points: City of El Paso intervened in a data center rate case; South Korea's Honam chip hub faces local opposition to power and water infrastructure, including nuclear reactors. Trump administration launched $200M DOE program tapping Oklo and X-Energy to power AI data centers, a major policy catalyst. China's nuclear surge (110 GWe by 2030) provides a benchmark for standardized build execution. A new data point: data centers expected to use 4x more electricity by 2035, reinforcing the demand thesis. Linn County, Iowa is considering a nuclear exception to its data center ban for the Duane Arnold restart with Google, a real-world test case for local opposition dynamics. A new legislative signal: the Ratepayer Protection Act advancing unanimously through committee could force data centers to pay for grid infrastructure, not ratepayers, directly impacting the AI-nuclear thesis. The 71% opposition Gallup poll and 430 local opposition groups confirm the 'Arc of Outrage' is real and escalating. The cost shift to ratepayers is a growing political friction point, with PJM capacity auction costs reaching $29.4B cumulative. The Ratepayer Protection Pledge is expanding rapidly, with 220+ partners including major tech companies and utilities (NextEra, Duke). Trump claims 80% of US power under check via voluntary pledges; Louisiana-Meta deal saved $2.6B. This reduces ratepayer opposition and accelerates nuclear deals. Additionally, Brookhaven National Lab partnered with AWS to launch GridSearch AI, aiming to accelerate power grid interconnection decisions 20-100x, directly addressing a key bottleneck for new nuclear and data center connections. A new counter-narrative article questions SMR viability for data centers, highlighting non-binding deals and cost challenges. A GIS Reports article adds geopolitical dimension: drone strikes on AWS facilities and sovereign AI blocs, reinforcing power as the key constraint.