June 1, 2026
By Vanguard with the work of Daniel Yergin, Fatih Birol, Michael Porter, Mariana Mazzucato, and Vaclav Smil.

Executive Thesis

Artificial intelligence is turning electricity into a strategic asset.

For much of the past decade, AI competition was framed around models, data, chips, talent, and capital. Those remain essential inputs. But in 2026, another constraint has moved to the center of the AI economy: power. The next phase of AI deployment will depend not only on who can build the most advanced models, but on who can secure reliable, affordable, scalable, and increasingly clean electricity.

This shift is changing the relationship between technology strategy, energy policy, industrial competitiveness, and national security. Data centers are becoming critical infrastructure. Grid capacity is becoming a competitive bottleneck. Clean energy supply chains are becoming geopolitical assets. Companies that can secure long-term power access may gain strategic advantage. Companies that cannot may face higher costs, deployment delays, carbon exposure, and dependence on constrained energy markets.

The energy transition is therefore no longer only a climate agenda. It is an economic competitiveness agenda.

The tension is clear. AI requires enormous compute capacity. Compute capacity requires data centers. Data centers require electricity, cooling, grid interconnection, land, transmission, backup power, and increasingly sophisticated energy management. At the same time, governments and corporations are under pressure to reduce emissions, improve energy security, and protect strategic supply chains from geopolitical disruption.

Leaders now face a more difficult question than whether AI should be scaled. They must ask: Can AI be scaled without undermining energy resilience, sustainability commitments, and national strategic autonomy?

The AI-Energy Demand Shock

The energy implications of AI are becoming measurable.

The International Energy Agency estimates that data centers consumed roughly 415 terawatt hours of electricity in 2024, approximately 1.5 percent of global electricity consumption, after growing at about 12 percent annually over the prior five years. The IEA expects demand from data centers to continue rising as AI workloads expand from training to inference, enterprise deployment, and consumer use.

In the United States, the trend is already visible in power forecasts. The U.S. Energy Information Administration expects U.S. electricity consumption to reach record highs in 2026 and 2027, with AI-related data centers and broader electrification contributing to demand growth. Commercial electricity consumption is projected to exceed residential consumption for the first time on record in 2026.

The strategic implication is straightforward: AI growth is becoming grid growth.

The problem is that power systems were not built for sudden, concentrated, high-load digital infrastructure expansion. Data centers often cluster near fiber routes, cloud regions, tax incentives, water access, large customer bases, and existing power infrastructure. Concentration can create local grid stress even when national electricity supply appears adequate. Research on AI data center siting suggests that the rapid buildout of compute capacity is likely to create regional power-system pressure, particularly in data center hubs where interconnection, transmission, and local capacity are already constrained.

This creates a new geography of AI advantage. The strongest locations may not be those with the most attractive tax incentives or lowest land costs. They may be those with available power, fast permitting, grid flexibility, clean energy access, cooling capacity, and political willingness to support digital infrastructure.

Power as the New Competitive Bottleneck

For companies, electricity is becoming a strategic input similar to chips or cloud capacity.

A firm that can secure long-term energy access can scale AI workloads with greater certainty. A firm that cannot may face delayed data center projects, higher cloud costs, carbon conflicts, or dependence on third-party providers with limited capacity. Moody’s 2026 outlook describes AI infrastructure as a critical bottleneck, with demand for compute driving data center construction, long-term capacity commitments, specialist chip shortages, grid constraints, and power needs that reshape access to AI infrastructure.

This matters because AI adoption is already uneven. Well-capitalized companies can sign large cloud contracts, invest in proprietary infrastructure, and secure power agreements. Smaller firms may depend on cloud providers whose pricing and capacity are shaped by these upstream constraints. If power becomes scarce, the AI adoption gap may widen.

Power also changes the economics of AI. Training frontier models is energy-intensive, but inference at scale may become even more important as AI tools are embedded into search, enterprise software, customer service, financial analysis, coding, design, and industrial systems. If inference demand becomes persistent and widespread, electricity cost becomes part of the marginal cost of intelligence.

This may push companies toward more efficient models, specialized chips, workload optimization, edge computing, and energy-aware AI architecture. The future of AI performance will not be measured only by benchmark accuracy. It will also be measured by cost per query, watts per task, carbon intensity, latency, and reliability.

The most competitive AI systems may be those that are not only powerful, but energy-efficient.

China’s Clean Energy Supply Chain Advantage

The energy transition creates a second strategic issue: supply chain concentration.

China dominates several clean energy supply chains that are central to electrification and decarbonization. In solar photovoltaics, the IEA has reported that China’s share across all major manufacturing stages of solar panels exceeds 80 percent, following large-scale investment in polysilicon, ingots, wafers, cells, and modules.

This dominance matters because the AI economy depends on the broader energy transition. Data centers need reliable power, and governments want that power to be increasingly clean. But if clean energy deployment depends heavily on Chinese-dominated supply chains, then AI infrastructure strategy becomes entangled with geopolitical exposure.

The same issue extends beyond solar. Batteries, critical minerals, grid components, rare earths, power electronics, and advanced manufacturing capacity all influence the pace and security of electrification. Reuters has reported ongoing efforts by the U.S. and allies to reduce dependency on China in critical minerals, including proposals around price supports, guaranteed purchases, subsidies, and tariffs. These debates reflect the difficulty of building non-Chinese supply chains that are both secure and cost-competitive.

The strategic dilemma is that companies and governments want clean energy quickly, cheaply, and securely. In practice, those goals can conflict. The lowest-cost supply chain may carry geopolitical risk. The most secure supply chain may be more expensive. The fastest deployment pathway may deepen dependency. The most resilient pathway may slow near-term decarbonization.

This is the central trade-off of the AI-energy transition.

Industrial Policy and National Security

AI, energy, and national security are increasingly inseparable.

Governments understand that AI capability depends on compute, and compute depends on physical infrastructure. Semiconductor policy, data center permitting, grid expansion, power generation, transmission, critical minerals, and clean energy manufacturing are now part of the same strategic conversation.

This is why industrial policy is returning to the center of economic strategy. The objective is not only to reduce emissions. It is to ensure that critical industries have access to power, technology, minerals, manufacturing capacity, and secure supply chains. AI leadership requires more than software excellence. It requires an industrial base capable of supporting digital infrastructure.

National security concerns arise in several areas.

First, dependence on foreign clean energy supply chains can create vulnerability. If geopolitical tensions disrupt solar, battery, mineral, or power equipment supply, AI infrastructure expansion may slow.

Second, concentration of data centers creates physical and cyber risk. These facilities support cloud platforms, financial systems, defense applications, enterprise operations, and AI services. Their reliability has strategic importance.

Third, energy constraints can become a limiting factor in technological competition. Countries that cannot expand power infrastructure quickly may lose ground in AI deployment.

Fourth, regulatory fragmentation can influence where companies build. Data sovereignty rules, energy permitting, carbon policy, and national security reviews may shape the geography of AI infrastructure.

The result is a new form of strategic competition: the race to build energy-secure AI economies.

Sustainability Versus Growth: A False Binary, but a Real Trade-Off

Companies often frame the issue as a choice between sustainability and growth. That framing is too simple.

In the long term, clean energy expansion can support both AI growth and climate goals. Renewable power, nuclear energy, storage, grid modernization, demand response, geothermal, hydro, and advanced energy management can all contribute to a more resilient and lower-carbon power system. Data centers can also become flexible loads, shifting workloads across regions or times of day to reduce pressure.

But in the short and medium term, trade-offs are real.

If AI data center demand rises faster than clean power deployment, fossil fuel generation may remain higher than expected. If data centers consume large amounts of available renewable power, other sectors may face slower decarbonization. If grid upgrades lag, new clean generation may not reach demand centers. If companies rely heavily on renewable energy certificates without adding new clean capacity, emissions reductions may be overstated.

Recent research on Europe’s AI-driven data center growth finds that AI demand could create significant additional electricity requirements by 2050 and may create intermediate emissions risks if policy and infrastructure do not adapt fast enough. The study also emphasizes that after 2030, firm power and system flexibility may matter more than the simple abundance of clean energy.

This point is critical. Clean energy availability is necessary, but not sufficient. AI infrastructure requires electricity that is not only clean, but reliable, dispatchable, connected, and available when needed.

The energy transition must therefore move from capacity targets to system performance.

Scenario One: Clean Power Acceleration

In the first scenario, governments and companies accelerate clean power deployment fast enough to meet AI demand without materially increasing emissions or grid stress.

This scenario requires faster permitting, expanded transmission, large-scale storage, advanced nuclear or life-extension of existing nuclear plants, demand-response markets, more efficient data centers, and closer coordination between utilities, hyperscalers, regulators, and local communities.

The reward is substantial. Countries that can provide clean, reliable power at scale may attract AI infrastructure, advanced manufacturing, cloud investment, and high-value digital industries. Companies operating in these markets may face lower carbon risk and greater long-term energy certainty.

The risk is execution. Clean power projects, transmission lines, interconnection queues, and permitting reforms take time. The AI buildout is moving faster than many energy systems can adapt.

The strategic response is early power procurement and infrastructure partnership. Firms should not treat electricity as a commodity purchased after site selection. Power strategy must come first.

Scenario Two: Fossil Backfill

In the second scenario, AI demand grows faster than clean energy and grid modernization. Utilities and governments respond by extending fossil generation, building new gas capacity, or delaying coal retirements to maintain reliability.

This scenario may support AI growth in the short term, but it creates emissions pressure, reputational risk, policy conflict, and exposure to fuel volatility. It may also weaken corporate climate commitments if companies cannot credibly match AI growth with additional clean energy.

The reward is speed and reliability. The risk is long-term carbon lock-in and public backlash.

For companies, the strategic response is transparency and mitigation. If fossil generation remains part of the reliability mix, firms should invest in efficiency, additional clean energy, storage, grid support, and credible emissions accounting. They should avoid claiming that AI growth is fully clean if the underlying system relies on fossil backup.

Scenario Three: Geopolitical Supply Shock

In the third scenario, clean energy deployment is disrupted by trade restrictions, mineral supply constraints, tariffs, sanctions, or conflict. Solar panels, batteries, transformers, rare earths, or other key inputs become more expensive or less available.

This scenario exposes the dependency problem. Countries attempting to build clean power quickly may discover that supply chain security is as important as project financing. Companies with data center expansion plans may face delays if energy infrastructure cannot be built on schedule.

The reward for firms that prepared is significant. Those with diversified suppliers, long-term procurement agreements, alternative technologies, and strong policy relationships can continue building while competitors stall.

The strategic response is supply chain diversification. Energy procurement strategy should include visibility into upstream equipment and minerals, not only power purchase agreements.

Scenario Four: Regional Energy Advantage

In the fourth scenario, AI infrastructure clusters in regions with abundant power, favorable regulation, strong grid systems, and clean energy potential. Some regions become AI-energy hubs. Others lose projects because they cannot provide reliable interconnection or acceptable cost.

This scenario is already emerging. Goldman Sachs has projected that U.S. data centers’ share of peak summer power demand could rise sharply from 2025 to 2027, underscoring the importance of regional power planning.

The reward is regional economic development. Data center clusters can attract investment, jobs, tax revenue, cloud ecosystems, and advanced services. The risk is local strain: higher power prices, land-use disputes, water concerns, grid congestion, and political resistance.

The strategic response is community-aligned infrastructure development. Firms should partner with utilities, local governments, and communities to ensure that AI infrastructure strengthens rather than burdens regional power systems.

Corporate Strategy: Securing the Energy Advantage

Executives should treat energy as a board-level AI constraint.

The first priority is energy mapping. Companies should identify how AI workloads affect electricity consumption, cloud costs, carbon exposure, and regional infrastructure dependency. Many firms understand their software stack better than their energy exposure. That gap will become a strategic weakness.

The second priority is power procurement. Long-term power purchase agreements, renewable energy partnerships, behind-the-meter generation, storage, nuclear contracts, and utility collaboration may become central to AI strategy.

The third priority is workload optimization. Not every AI workload requires the same location, latency, or model size. Companies should optimize workloads across regions, models, and compute environments based on cost, energy availability, carbon intensity, and performance.

The fourth priority is efficiency. Model compression, specialized chips, cooling innovation, server utilization, software optimization, and scheduling can reduce power intensity. Energy efficiency should become part of AI governance.

The fifth priority is supply chain visibility. Companies should understand the supply chains behind their clean energy strategy: solar modules, batteries, transformers, grid equipment, critical minerals, and backup systems.

The sixth priority is geopolitical risk management. AI infrastructure plans should include scenarios for tariffs, export controls, mineral constraints, grid delays, and policy changes.

The seventh priority is stakeholder credibility. Companies should communicate honestly about energy use, emissions, procurement, and infrastructure impact. Trust will matter as public scrutiny of AI’s energy footprint increases.

The Leadership Framework

A practical leadership framework for AI-energy competitiveness includes five questions.

1. Can We Power the Strategy?

AI ambition should be tested against power availability. If the energy plan is unclear, the AI plan is incomplete.

2. Can We Power It Reliably?

Intermittent clean energy alone may not meet the needs of critical AI infrastructure. Reliability, storage, firm generation, and grid stability must be included.

3. Can We Power It Cleanly?

Companies should assess whether AI growth is adding new clean capacity or merely reallocating existing clean power.

4. Can We Power It Securely?

Supply chain concentration, critical minerals, data center clustering, and geopolitical exposure should be treated as strategic risks.

5. Can We Power It Competitively?

Energy cost, latency, location, regulation, and infrastructure delays will influence AI economics and business competitiveness.

These questions move energy from the sustainability office to the strategy office.

The National Competitiveness Agenda

For governments, the AI-energy transition requires coordinated policy.

First, grid modernization must accelerate. Transmission, distribution, interconnection, storage, and digital grid management are now competitiveness priorities.

Second, permitting must become faster without eliminating environmental and community review. Delayed infrastructure is a hidden tax on AI leadership.

Third, clean energy supply chains must diversify. Domestic and allied capacity in solar, batteries, critical minerals, nuclear components, transformers, and grid equipment will matter.

Fourth, data center policy must become more sophisticated. Governments should encourage efficiency, transparency, grid support, and regional planning rather than treating every project as automatically beneficial.

Fifth, workforce development must align with energy and AI infrastructure. The economy will need electricians, engineers, grid planners, construction workers, cybersecurity experts, energy traders, technicians, and AI infrastructure specialists.

Sixth, international coordination is needed. Critical minerals, clean technology supply chains, export controls, and energy security cannot be managed by one country alone.

The countries that integrate AI strategy with energy strategy will be better positioned for the next phase of industrial competition.

The Strategic Trade-Off

The AI-energy transition creates a difficult but manageable trade-off.

Too little power, and AI competitiveness slows. Too much fossil backfill, and climate goals weaken. Too much dependence on China-dominated clean energy supply chains, and strategic vulnerability grows. Too much protectionism, and clean energy deployment becomes slower and more expensive. Too little regulation, and local communities may resist data center expansion. Too much regulation, and investment may move elsewhere.

The answer is not a single policy or corporate move. It is a portfolio approach: diversified energy sources, cleaner grids, efficient AI systems, resilient supply chains, transparent reporting, and regional planning.

Businesses that understand this early will have an advantage. They will secure capacity before bottlenecks worsen. They will build credibility before scrutiny intensifies. They will design AI systems that are economically and energetically efficient. They will reduce exposure to geopolitical shocks. They will align growth with sustainability rather than treating the two as permanently opposed.

The Next AI Advantage Is Energy Advantage

The AI economy will not be built in the cloud alone. It will be built on power systems.

As AI demand grows, electricity becomes more than an operating cost. It becomes a determinant of speed, scale, resilience, and national advantage. The companies and countries that can provide reliable, clean, secure, and affordable power will shape the next phase of technological competition.

The energy transition is therefore entering a new stage. It is no longer only about reducing carbon. It is about powering the industries that will define economic leadership.

For executives, the mandate is clear: energy strategy must become part of AI strategy. For policymakers, energy infrastructure must become part of competitiveness strategy. For investors, power availability must become part of the AI valuation model.

AI may be digital in output, but it is industrial in foundation.

The winners will be those who understand that the future of intelligence depends on the future of power.