AI Is No Longer Waiting for the Grid — The Grid Is Now Limiting AI

For years, the story was simple: AI needs more power, and the grid will find a way to deliver it. Hyperscalers planned billions in investment, transformer manufacturers ramped up production, and utilities scrambled to keep up.

That story just changed.

On July 14, 2026, New York Governor Kathy Hochul signed an executive order imposing a one-year moratorium on new large data centers that would consume 50 megawatts or more of electricity. New York became the first U.S. state to temporarily ban the construction of large AI data centers.

The reason? Not water. Not community opposition. The grid.

As Hochul put it: "These hyperscale AI data centers consume enormous amounts of power, truly threatening to outpace our grid's capacity, and they drive up costs for local ratepayers."

Half a world away, Australian Prime Minister Anthony Albanese announced on the same day a "legal obligation" for large data centers to underwrite their own new power supply and pay their full share of grid connection costs.

Two governments. Two continents. One message: AI can no longer assume the grid will accommodate it.

Key Takeaways

• ✓ New York has imposed a one-year moratorium on data centers over 50 MW, citing grid capacity and ratepayer costs. The state will use the pause to develop new permitting standards for large-scale power infrastructure projects.
• ✓ Australia is leGISlating a "legal obligation" for data centers to underwrite new renewable generation and pay full grid connection costs — shifting the financial burden of utility infrastructure expansion from ratepayers to developers.
• ✓ The global transformer shortage is a structural constraint on AI expansion.Lead times for large power transformers now average 128 weeks, and hyperscalers are competing with utilities for limited electrical equipment manufacturing capacity.
• ✓ Grid connection timelines (4–10 years) now exceed AI data center planning cycles (2–3 years). This mismatch is forcing governments to intervene in power distribution and transmission capacity planning.

New York: The First State-Level Moratorium

The executive order applies to new or expanded data centers capable of using at least 50 megawatts of electricity. During the one-year moratorium, the New York State Department of Environmental Conservation will not issue discretionary permits for facilities meeting that threshold.

During the suspension period, New York plans to analyze the impact of large data centers on the power grid, electricity rates, cooling water usage, and local communities. The state will also develop new permitting standards for future data center development.

Perhaps most significantly, the order establishes a framework requiring data centers to contribute to grid improvements. The state may require data centers to make upfront capital contributions to finance grid improvements, participate in demand response programs, support the procurement of new clean energy supply, and establish an insurance pool to cover potential grid impacts.

In other words: if you want to build big AI in New York, you will help pay for the grid you are straining.

Australia: From "Polite Request" to Legal Obligation

On the same day, Australian Prime Minister Anthony Albanese delivered a major speech at the University of Sydney, announcing a dramatic shift in policy.

"We will create a legal obligation for the next generation of large-scale data centres to underwrite new power supply," Albanese said.

The proposed standards for large data centres include:

• A legal obligation to underwrite their own new power supply
• A requirement to pay their full share of grid connection costs so energy bills are not passed on to households or businesses
• A mandate to reduce power when needed to strengthen the grid
• A requirement to be as water efficient as possible

One analyst described this as moving from a "polite request" to a binding obligation. The government is effectively telling hyperscalers: if you want to consume grid capacity, you must also create it.

Why This Is Happening: The Numbers Behind the Crisis

AI's Power Appetite Is Outpacing the Grid

Global data center electricity demand is projected to more than double from approximately 415 TWh in 2024 to 945 TWh by 2030. In 2025, global data centers consumed 447 TWh. In 2026, that figure jumped to 565 TWh—a single-year increase of 118 TWh.

AI-optimized servers alone saw their electricity consumption explode from 95 TWh in 2025 to 175 TWh in 2026—an 84% increase in just one year.

Grid Connection Takes Years—Longer Than AI Projects

A critical mismatch is emerging: AI data centers typically plan for a 2-to-3-year construction cycle, but in many regions, connecting a new facility to the grid can take 4 to 10 years.

The transformer Bottleneck

transformer lead times have been rising steadily for over five years. Large power transformers now average 128 weeks. Generator step-up units average 144 weeks. In some markets, delivery timelines for power equipment have pushed past four to five years.

Hyperscalers have planned over $650 billion in AI spending in 2026 alone, yet close to half of all planned U.S. hub builds this year are projected to be delayed or cancelled. The bottleneck is not capital or talent. It is grid infrastructure and the electrical equipment that supports it—transformers, switchgear, substations, and distribution networks.

Why This Matters Beyond AI

Although the immediate trigger is AI infrastructure, the implications extend much further.

The same substations, transformers, and transmission assets required for hyperscale data centers are also needed for:

• Renewable energy integration– Solar and wind farms require significant transformer capacity to connect to the grid
• Utility grid modernization – Aging electrical infrastructure across developed economies needs replacement
• EV charging infrastructure – Public charging networks depend on distribution transformers and switchgear
• Battery energy storage systems (BESS) – Storage projects require step-up transformers for grid connection
• Industrial electrification – Factories shifting from fossil fuels to electricity add load to distribution networks

As more industries compete for the same electrical infrastructure, transformer procurement timelines are likely to remain under pressure for the foreseeable future. The grid capacity challenge is not an AI problem—it is an everyone problem.

What This Means for the Industry

The AI-Grid Relationship Has Entered a New Phase

For the past three years, the narrative has been: AI demand is driving transformer demand. That narrative is still true, but it is no longer the full story.

Phase 1: AI drives demand → grid scrambles to catch up.

Phase 2: Grid capacity limits AI → governments intervene.

We are now firmly in Phase 2.

What This Means for Utilities and EPC Contractors

If you are a utility planner, the message is clear: grid capacity is now a binding constraint on economic development. AI companies will not wait forever. Regions that can offer reliable, timely grid connections will attract investment. Regions that cannot will lose it.

If you are an EPC contractor, the message is equally clear: transformer and switchgear procurement must move earlier in the project lifecycle.

What This Means for Transformer and Switchgear Manufacturers

For manufacturers like Derui Electric, this is a moment of both challenge and opportunity. Utilities and hyperscalers alike are now prioritizing grid equipment with urgency they have never shown before. Governments are beginning to require data centers to contribute to grid expansion—meaning more investment in transformers, switchgear, and substations.

What Derui Electric Is Watching

• The pace of grid modernization investment. Government interventions in New York and Australia are signals that grid capacity is becoming a political issue. That typically accelerates utility infrastructure spending.
• The evolution of permitting standards. New York's one-year study period will produce new requirements for data center construction. Those requirements will likely include higher standards for grid integration and power equipment resilience.
• The response of hyperscalers. Will AI companies build their own power generation, as Australia is now requiring? Will they invest directly in grid infrastructure? Their response will shape the next phase of the market.

Conclusion: The Grid Is Now a Strategic Constraint

For years, the transformer industry has benefited from AI's insatiable demand for power. That demand is not going away. But the constraints on meeting it are becoming more visible—and more binding.

New York's moratorium and Australia's new legal obligations are not isolated events. They are early warnings of a broader trend: governments are stepping in to manage the tension between AI ambition and grid reality.

For procurement professionals, the implication is clear: grid equipment is no longer a downstream consideration. It is a strategic priority.

Sources: New York Governor's Office, AM New York, Good Morning America, RenewEconomy, ARN, The Register, IEA Electricity 2026, Xinhua Finance, OFweek, Sina Finance.

Frequently Asked Questions (FAQ)