Too Much Too Fast? Analyst Warns AI Data Center Scramble Threatens Power Projects, Electricity Customers

The power demands of artificial intelligence and the data centers that make it possible are bumping up against the “speed limit” of how fast the power grid can develop, which puts electric power projects, electricity markets and ratepayers all at risk.

According to a new report by Wood Mackenzie, data center companies now face a difficult choice: They can wait up to 10 years for transmission and generation to catch up with a massive build-out, or they can pay for data centers to generate their own power with microgrids and associated technology — a riskier prospect for hyperscalers.

These microgrids would be of unprecedented scale and complexity, and would face technical, regulatory and economic hurdles larger than the industry seems to recognize, according to the report.

"The power sector is fixated on data centre flexibility, but that is not the endgame for grid operators or data center operators," said Ben Hertz-Shargel, Global Head of Grid Transformation and Large Loads, Wood Mackenzie. "Firm grid service is the goal, backed by new transmission superhighways. But there is a lack of awareness throughout the power sector about the technical and regulatory risk confronting collocation projects, and the business risk of conditional interconnections."

To give a sense of the scale at work here, the PJM Interconnection has 78 GW worth of of committed data center load against only 36 GW of generation capacity in its pipeline. There are more than 90 GW of collocated generation in US interconnection pipelines.

Wood Mackenzie’s accelerated case for data center load growth includes 16.4 GW of natural gas capacity additions per year through 2035 to meet projected demand, despite only 4 GW per year from 2023 to 2025.

"Load growth and affordability are in direct opposition in the deregulated markets," said Chris Seiple, Vice Chairman, Energy Transition and Power and Renewables, Wood Mackenzie. "If prices rise to the level necessary to incentivize new generation, it will raise prices for all customers, prompting a political outcry."

Power grid operators, for their part, are racing as fast as this typically slow-paced market can, with regional transmission build-outs planned by the mid-2030s, and some making structural changes to their price tiers. PJM, for example, is bifurcating its generation market — creating one elevated price tier for new resources contracted by large loads, and a lower tier for existing resources.

In February, PJM approved a $11.8 billion investment in its transmission superhighway, one of the largest expansions in the market’s history. MISO, a grid operator in the Midwest, is planning a transformative long-range transmission investment that could total over $20 billion, on top of an in-progress investment of $10 billion. 

ERCOT, the Texas grid operator, plans to invest US$33 billion in its transmission superhighway, while SPP in the Great Plains states will be investing $8.6 billion.

In Texas, where electricity is too cheap on a per-megawatt basis to attract new natural gas power plant generation, there is no comparable plan to incentivize new grid-connected generation, and the plan seems to be for competitive power markets to deliver the new supply. 

ERCOT, has also been working with stakeholders to update its voltage and frequency ride-through requirements — rules designed to prevent data centres from reverting to backup power prematurely during minor grid disturbances. The risk is not hypothetical. In 2024, 60 data centers in Virginia dropped off the grid simultaneously following a minor disturbance, nearly causing a grid collapse.

"PJM's and SPP's rules are understood to give the regional grid priority rights over collocated generation," said Hertz-Shargel. "During shortages, data centers would be forced to reduce demand to their firm service level, even as their onsite generation was instructed to supply the grid. For some companies, this model is unworkable."

Further regulatory intervention in deregulated markets is likely as affordability pressures mount, with potential knock-on consequences for existing asset valuations. The report warns that developers and investors should be prepared for market rule changes driven not by grid planning, but by political pressure.

"Even for developers that see collocation as a viable bridging solution to grid power, the costs and technical challenges are formidable," said Hertz-Shargel. "Technology providers are only beginning to come to terms with this challenge, the mitigation of which is site-specific, making solutions hard to scale."

Key technical challenges include:

• The near-instantaneous changes in AI power demand can damage reciprocating engines and gas turbines
• Lithium-ion batteries can be used as a shock-absorber to prevent this, but risk running through their useful lifespan rapidly
• The battery response time must be extremely short, relying on technology that has not been widely commercialised
• The irregular manner in which AI cooling and GPU loads consume power introduces power harmonics, which if unfiltered cause equipment to overheat and degrade
• These loads can also cause sub-synchronous oscillations, posing fundamental stability risk to local generators and to distant ones on the transmission system

Data centres in late-stage development are unlikely to be able to comply with the proposed changes, however, potentially requiring a site redesign and the purchase of new uninterruptible power supplies and other electrical equipment. With rules still being written and late-stage changes already threatening project viability, regulatory risk remains one of the direst — and most underappreciated — challenges facing the industry.

Underlying all of these challenges is also a fundamental question about cost allocation. Grid operators are planning close to $100 billion in transmission investments in part to support data center load growth, including PJM ($11.8 billion), MISO ($30 billion+), ERCOT ($33 billion) and SPP ($8.6 billion). Without modification of traditional cost allocation rules, a significant portion of these regional investments could be spread across all existing ratepayers, not assigned specifically to the data centers driving their urgency.