Microsoft SMR Strategy, $16 B Constellation Deal, $130 B in Canceled Projects, and 223 Project Terminations (2025 to 2026)

The expansion of artificial intelligence is creating a system-level energy crisis, pitting the digital world’s demand for power against the physical world’s ability to supply it. Hyperscalers like Microsoft are making multi-billion-dollar commitments to nuclear energy to secure the 24/7 carbon-free power required for their AI data centers. Yet, this strategic push is colliding with a harsh reality of grid constraints, regulatory hurdles, and a cascade of project cancellations across the energy and data center sectors. This creates an execution paradox: the very industry driving the need for new power is also a factor in the disruption of the energy project pipeline, creating a significant capacity gap that threatens future growth. The immediate, viable path is financing existing nuclear assets, while the long-term vision for new reactors remains unproven against a backdrop of widespread project failure.

AI Power Demand: Microsoft’s Nuclear Deals and $130 B in Data Center Cancellations

The escalating power requirements for AI are forcing technology companies into the energy sector, where they are underwriting nuclear power, while the secondary effects of their growth are simultaneously causing widespread data center project delays. This dynamic illustrates a fundamental conflict between digital ambition and physical infrastructure limits. The period from 2025 to 2026 marks a sharp inflection point where the strategy to secure power through direct offtake agreements ran into the physical constraints of the grid and local opposition.

• Between 2021 and 2024, hyperscaler energy strategy focused primarily on renewable PPAs for wind and solar to meet sustainability goals. The scale of AI-driven demand was not yet a primary driver of utility-scale infrastructure planning.
• From 2025 to today, the narrative has shifted to securing firm, 24/7 power to manage the exponential growth in data center electricity consumption, which is projected to reach as high as 17% of the U.S. total by 2030. This has led to landmark nuclear deals, such as Microsoft’s $16 billion PPA with Constellation Energy for the entire 835 MW output of the restarted Three Mile Island Unit 1.
• This strategic push is directly contradicted by market reality. In the first three months of 2026 alone, over $130 billion in AI data center projects were blocked or delayed due to power shortages and local opposition. This represents a 7 GW capacity crisis, as nearly half of U.S. AI data centers planned for 2026 have been stalled.

Over $82 B in Canceled Projects: The Impact of Policy and Market Volatility

Beginning in 2025, a sharp reversal in federal policy combined with market volatility triggered a wave of cancellations across the clean energy sector, erasing tens of billions of dollars in planned investments and over 100, 000 jobs. These terminations were not isolated incidents but a systemic response to a deteriorating investment environment, directly undermining the development of the new generation capacity required by the technology sector.

• Policy changes were the primary driver, directly stalling or canceling 223 projects representing $82.9 billion in investment. This includes the Department of Energy’s termination of financial awards for these projects, claiming a savings of $7.56 billion but creating a significant project financing gap.
• The offshore wind sector was particularly affected, with the Department of the Interior suspending leases for five major projects, including Ørsted’s Revolution Wind. This was followed by the administration paying nearly $4 billion to energy firms specifically to cancel planned offshore wind developments.
• Major corporations independently canceled flagship projects due to economic reviews and shifting priorities. Shell abandoned its 820, 000 metric-ton-per-year advanced biofuels project in Rotterdam, and Air Products canceled a major hydrogen and ammonia complex in Louisiana, citing strategic re-evaluation.
• This pattern of cancellation was established in 2025, when over $35 billion in U.S. clean energy and EV projects were terminated, resulting in the loss of more than 38, 000 jobs and setting the stage for the broader infrastructure failure seen in 2026.

Table: Canceled Clean Energy and Data Center Projects (2025 – 2026)

Microsoft 20-Year PPA: Hyperscaler Alliances for Nuclear Power (2025 to 2026)

To counteract the volatility of renewable energy and secure baseload power, technology companies have become primary financiers for the nuclear industry, signing long-term PPAs that provide the financial certainty needed to operate and potentially expand the nuclear fleet. These partnerships, led by Microsoft, Meta, and Google, represent a strategic shift from simply buying clean energy to directly enabling its generation.

• The defining partnership is Microsoft’s 20-year, $16 billion agreement with Constellation Energy to offtake the full 835 MW capacity from the restarted Three Mile Island Unit 1. This deal provided the financial backing for the plant’s restart, which was reported to be ahead of schedule in June 2025.
• Meta followed this trend in January 2026 with landmark agreements with Vistra, Terra Power, and Oklo. The Vistra deal supports 2, 176 MW of operating generation and 433 MW from power uprates, the largest such agreement in history.
• Google has also entered the nuclear space, partnering with Kairos Power to deploy up to 500 MW of new advanced nuclear projects. The company also announced plans for three 600 MW nuclear projects to power its data centers.
• These alliances underscore a new reality: the AI power demand is so immense that technology companies are now functioning as de facto energy financiers, using their balance sheets to secure multi-decade power supplies.

Table: Hyperscaler Nuclear Power Partnerships (2025 – 2026)

Georgia Power Demand: Regional Grid Strain and Bipartisan Backlash

Geographic analysis reveals that the AI energy crunch is not a national monolith but a series of acute regional crises, with states like Georgia, Virginia, and Arizona becoming flashpoints. In these areas, the rapid, concentrated growth of data centers has overwhelmed local utility planning, sparking legislative backlash and creating direct competition for land and power with other industries.

• In Georgia, the surge in data center development has provoked public outrage and a wave of bipartisan bills in January 2026. Legislators are seeking to increase transparency and potentially remove tax incentives for the industry, citing concerns over the strain on energy and water resources.
• This follows a period from 2021-2024 where states actively courted data centers with incentives. The shift in 2025-2026 reflects the unforeseen consequences on utility rates and grid stability, turning a perceived economic benefit into a political liability.
• In the UK, this conflict became explicit when BP pulled out of its H 2 Teesside hydrogen project after landowners sought to build a data center on the proposed site instead, illustrating how data centers are now directly displacing other clean energy projects.
• The PJM Interconnection, which serves the Mid-Atlantic and parts of the Midwest, has become another key region. Meta’s deal with Vistra to support nuclear plants in PJM territory is a direct response to the grid operator’s warnings about soaring demand and potential capacity shortfalls.

Technology Maturity: Existing Nuclear Fleet vs. Speculative New-Builds

The strategic approach to nuclear power is bifurcated, reflecting a stark difference in technology and execution maturity. Supporting existing nuclear plants through life extensions and restarts is a proven, near-term strategy for securing large-scale, carbon-free power. In contrast, the vision of powering data centers with new Small Modular Reactors (SMRs) or other advanced designs remains speculative and challenged by historical execution failures and unfavorable economics.

• The success of the Three Mile Island restart, underwritten by Microsoft, validates the financial model for preserving and enhancing the existing nuclear fleet. This is a mature, low-risk pathway to adding hundreds of megawatts of firm power to the grid.
• Conversely, the history of new-build nuclear projects is defined by massive cost and schedule overruns. A previous “nuclear renaissance” saw over 40 canceled capital projects valued at approximately $250 billion. The Flamanville EPR in France, which started up 12 years late at a sixfold cost increase, serves as a recent cautionary tale.

– The Levelized Cost of Energy (LCOE) for new nuclear ranges from $175/MWh to $255/MWh. This is substantially higher than utility-scale solar at $30-$40/MWh or onshore wind at $39/MWh, making new nuclear economically challenging without significant subsidies or strategic imperatives like the AI power demand.

• While the Inflation Reduction Act and other policies provide support, the sector still faces major bottlenecks in the supply chain for components like large power transformers and a need for streamlined regulations for first-of-a-kind SMR projects.

SWOT Analysis: Microsoft’s Nuclear Strategy and the AI Energy Demand

The push by Microsoft and other hyperscalers into nuclear power is born of necessity, but it is a strategy defined by high-stakes opportunities and significant external threats. The analysis reveals that while the strengths and opportunities are compelling, the weaknesses and threats related to project execution and market stability are substantial and could derail the long-term vision.

Table: SWOT Analysis for Hyperscaler Nuclear Power Strategy

Scenario Modelling: Microsoft’s Near-Term Focus on Existing Assets

The most critical strategic action for hyperscalers in the next 18-24 months will be to secure capacity from existing clean power assets rather than relying on unproven, long-lead-time new-build projects. The market is signaling a clear preference for derisked, operational assets to close the immediate power gap. If hyperscalers cannot secure enough firm power from the existing grid, watch for a significant slowdown in AI infrastructure deployment or a pivot to natural gas, which would compromise corporate carbon-neutrality goals.

• If this happens: Watch for a surge in new Power Purchase Agreements targeting the uprating and life extension of existing nuclear plants, similar to the Microsoft/TMI and Meta/Vistra deals.
• And watch this: Monitor interconnection queues and policy decisions in key data center regions like Georgia and Virginia. Further legislative backlash or moratoriums on new data centers are leading indicators of a deepening power constraint.
• These could be happening: Hyperscalers may increase investment in advanced geothermal and long-duration energy storage as alternative firm power sources. They could also begin acquiring smaller utilities or developing microgrids to gain direct control over their power supply. The focus will be on executable projects that can deliver power within a 3-5 year timeframe, not the 10+ years required for traditional nuclear.

The questions your competitors are already asking

This report covers one angle of the AI energy crisis. The questions that matter most depend on your work.

• data center construction moratoriums by state
• US nuclear power plants eligible for life extension
• tech companies buying power plants
• small modular reactor projects with utility partners

This report does not answer these. Enki Brief Pro does.

Your question, your angle, your framework. SWOT, PESTL, scenario modelling. The same niche depth, built around the decision your work actually depends on.

Run your first brief in Enki Brief Pro