SMRs for Data Centers, Microsoft’s 10.5 GW Brookfield Plan and $80 B Cameco Nuclear Deal (2024-2026)

The data center industry is undergoing a fundamental restructuring, moving from a passive consumer of grid power to an active producer of its own electricity. This shift is driven by the explosive power demand of Artificial Intelligence, which has created a systemic constraint where the public grid cannot provide power fast enough for hyperscale development. As a result, operators are building a parallel, private energy system, deploying on-site generation to bypass multi-year grid interconnection queues and secure the gigawatts of power necessary for AI infrastructure.

Grid Constraint Risks, Hyperscalers Build Parallel Power Systems

The core risk to data center growth has shifted from capital access to power availability, forcing operators to adopt a “bring your own power” model to control project timelines.

• Prior to 2024, data center power demand growth was predictable, allowing development to align with utility capacity expansion. The AI-driven demand shock has broken this model, with new AI server racks requiring 50-100 k W each, up from less than 10 k W a few years ago.
• The primary constraint is not a lack of total generation capacity, but the inability to connect to it. Interconnection queues in the U.S. have swelled, with typical wait times extending from five to seven years, a timeline that is untenable for the fast-paced AI hardware deployment cycle.
• In 2026, this has become the default strategy for large-scale deployments, where “time-to-power” is now the most critical competitive metric. Nearly 190 GW of new hyperscale capacity has been announced across 777 projects, all predicated on securing power outside of traditional utility timelines.
• This has led to the emergence of a parallel energy system, with analysts estimating a $150 billion private grid is being constructed by data center operators to meet their own demand. This includes a shift by former crypto miners like Core Scientific and Hut 8, who are converting their power-rich sites for AI workloads.

UPS and Generators Lead Hyperscale On-Site Power Market
UPS systems command the largest share (34.1%) of the $258.37 billion hyperscale data center electric infrastructure market by 2026, followed by generators at 21.2%. This signals a strong foundational reliance on robust power backup and primary generation for on-site resilience.

Uninterrupted Power Drives Over Half of Infrastructure Investment
The combined 55.3% share of UPS and generators underscores hyperscalers” critical need for uninterrupted power. This priority drives significant investment, making power resilience a non-negotiable factor in data center design and operation, especially for high-density computing loads.

Hyperscalers Pivot to Onsite Power Generation by 2030
Facing 4-7 year grid connection delays and contributing 55% of U.S. electricity demand growth, 33% of hyperscalers plan for 100% onsite power generation by 2030. This strategic shift addresses power constraints and bolsters resilience for the 97% global data center occupancy rate, with over $600 billion invested in 2026 alone.

(Source: Data Center Power 2026: On-Site Generation is Mandatory)

$150 B in Private Grids, Hyperscaler On-Site Generation Investments

Hyperscalers and data center developers are committing tens of billions in capital to vertically integrate power generation, moving energy infrastructure from an operational expense to a core asset.

• The scale of investment reflects the magnitude of the demand, with data centers projected to consume up to 9% of all U.S. electricity by 2030, according to the Electric Power Research Institute (EPRI). Some high-growth scenarios place this figure as high as 17%.
• In January 2026, American Intelligence & Power announced a strategic alliance with Caterpillar to deploy 2 GW of dedicated power, primarily from natural gas turbines, for hyperscale AI infrastructure.
• Also in January 2026, Vantage Data Centers and Liberty Energy formed a partnership to develop and operate 1 GW of on-site power solutions, highlighting the trend of data center operators directly partnering with energy infrastructure firms.
• This investment is a direct response to grid failures. In the U.S., nearly 7 GW of planned data center capacity for 2026 faced delays or cancellations specifically due to the lack of available grid power.

Table: On-Site Power Generation Investments & Cancellations

Hyperscalers Accelerate On-Site Power Generation Due to Grid Constraints
Facing 4-7 year grid connection delays and driving 55% of U.S. electricity demand growth, 33% of hyperscalers plan 100% on-site power generation by 2030. This aggressive shift is a direct response to current power constraints and the need for energy independence.

Market Decentralization Driven by Power Availability and Frontier Market Growth
The urgent pivot to on-site power is decentralizing the data center landscape, with 64% of new capacity shifting to ‘frontier markets.’ This redirection is evidenced by Virginia’s market share declining by 35% by 2028, while Texas projects a 142% increase, fundamentally reshaping infrastructure investment.

Onsite Power Expectations for Data Centers Skyrocket, 33x Growth by 2030
Developers’ expectations for 100% onsite power in data centers by 2030 have surged 33x from 1% (April ’24) to 33% (Nov “25). Similarly, projections for 2035 have grown 3.7x to 44%. This dramatic acceleration signals a major industry shift towards energy self-sufficiency.

(Source: AI Data Center Power: Grid Limits Reshape Energy in 2026)

Hyperscaler Energy Partnerships, 777 New Projects in Development

To execute the on-site power strategy, hyperscalers are forming deep alliances with energy producers, equipment manufacturers, and infrastructure funds, creating a new ecosystem for power development.

• The strategy extends beyond simple power purchase agreements to include joint ventures and direct investments. Microsoft is pursuing a landmark 10.5 GW renewable energy plan with Brookfield Asset Management, which includes exploring nuclear power options.
• These partnerships aim to secure the full energy supply chain. For its long-term nuclear ambitions, Brookfield has engaged in an $80 B deal involving Cameco and Westinghouse to secure fuel and reactor technology.
• Both Google and Microsoft are driving a massive capital expenditure cycle, deploying fuel cells and other on-site technologies to power their AI expansion, with Google explicitly adopting a “power-first” model for site selection.
• New joint ventures like the $2 B Clean Core AI project in Texas are emerging to build dedicated power and data center campuses from the ground up, designed specifically for AI workloads.

US Power Demand, Data Centers to Consume 9% of Electricity by 2030

The concentration of data center development in the United States is creating regional power crises, forcing states and regulators to reassess energy policy and economic incentives.

• The U.S. is the epicenter of this trend, with EPRI forecasting that data centers could add approximately 300 TWh of new electricity demand by 2030, equivalent to the current residential consumption of several states combined.
• This intense regional demand has led to pushback. In May 2026, the governor of Ohio announced a pause on new data center tax exemptions, citing the immense strain these facilities place on the state’s electrical grid.
• Federal regulators are also responding. In June 2026, the Federal Energy Regulatory Commission (FERC) launched a targeted initiative to accelerate the grid integration process for large loads like data centers, acknowledging the existing system is a bottleneck.
• In a bid to preempt further regulatory backlash and ratepayer revolts, major hyperscalers signed a White House “Ratepayer Protection Pledge” in March 2026, committing to fund grid upgrades necessary to support their facilities.

On-Site Generation Technology, Natural Gas Bridge to SMR Future

Hyperscalers are pursuing a two-tiered technology strategy: deploying commercially mature technologies for immediate needs while investing in next-generation nuclear for long-term, carbon-free baseload power.

• The immediate solution for bypassing grid queues is on-site natural gas generation. Gas turbines are a mature technology with a reliable supply chain, allowing for rapid deployment to meet urgent “time-to-power” deadlines.
• Fuel cells represent a cleaner, though more expensive, bridge technology. They offer high efficiency and low on-site emissions, making them suitable for deployment in denser areas, but their adoption at the multi-hundred-megawatt scale is still emerging.
• Small Modular Reactors (SMRs) are the strategic long-term objective. Though not yet commercially deployed, hyperscalers view SMRs as the only technology capable of providing the clean, 24/7, gigawatt-scale power required for future AI campuses.
• To manage the carbon footprint of their bridge strategies, companies like Microsoft are signing large-scale carbon removal agreements, such as its deal with Liferaft for 1 million tonnes of biomass-based carbon removal, while exploring battery solutions from companies like EPC Power to improve efficiency.

2027 Outlook, Hyperscalers Face Regulatory and Supply Chain Hurdles

As hyperscalers evolve into major energy players, their primary challenges will shift from technology to supply chain execution and navigating a complex regulatory environment.

• If this happens: The aggressive build-out of on-site natural gas generation continues at its current pace.
• Watch this: Increased regulatory scrutiny from the EPA and state-level environmental agencies over emissions from data center power plants, potentially leading to permitting delays and new emissions standards.
• These could be happening: Supply chain bottlenecks emerge for critical power equipment, including gas turbines, high-voltage transformers, and switchgear, as multiple multi-gigawatt projects compete for a finite manufacturing capacity. Watch for further vertical integration as operators attempt to secure their supply chains.

The questions your competitors are already asking

This report covers one angle of the energy strategy for AI data centers. The questions that matter most depend on your work.

• Data center projects delayed by power shortages
• Cost to build private power for data centers
• Regulatory approvals for small nuclear reactors
• Microsoft Brookfield nuclear power plan details

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