The AI data center boom is now the dominant force on the US grid, and it reaches every commercial building through three channels: reliability, rates and local politics. Hyperscalers are contracting the scarcest asset in energy, firm power that exists today, with Google signing roughly 3,590 megawatts from Constellation one week after Amazon’s Calvert Cliffs deal. Regulators are writing new reliability rules for these loads, and their demand is pushing wholesale power costs up across regions. Meanwhile, local fights over data center siting are turning into a political risk for development. An owner controls none of this directly, which is exactly why it belongs in their risk planning.
Key Facts at a Glance
Google signed a 20-year agreement for about 890 megawatts of new nuclear capacity plus a 15-year, 2,700-megawatt supply deal with Constellation, roughly 3,590 megawatts in total across Illinois, New Jersey and Pennsylvania.
It followed Amazon’s Calvert Cliffs deal by a week, underscoring a race for firm, carbon-free power that already exists.
NERC set mandatory reliability rules for large loads at 50 megawatts and 100 kilovolts, after grid disturbances tied to sudden data center load losses.
PJM wholesale power costs are running roughly 50 percent higher, with the grid’s market monitor attributing about 9 percent of 2026 wholesale costs to data center load.
Local data center siting fights are intensifying, with a Memphis moratorium vote collapsing into protests and moratoriums spreading across states.
The Race for Existing Power
The defining move is hyperscalers contracting firm power directly. Google signed a deal with Constellation for roughly 890 megawatts of new nuclear capacity on a 20-year agreement, plus 2,700 megawatts of supply from existing plants, about 3,590 megawatts across Illinois, New Jersey and Pennsylvania. It came a week after Amazon’s deal to uprate Maryland’s Calvert Cliffs plant. The pattern is unmistakable: the scarcest asset in American energy is not a reactor that might arrive in the 2030s, it is firm, carbon-free power available now, and the largest buyers are contracting it for decades.
For a commercial owner, the read-through is about scarcity. When hyperscalers lock up the firm power that already exists, less of it is available to everyone else, and the price of what remains rises. The owner is not competing for a reactor, but is competing, indirectly, for the same megawatts and the same grid capacity.
How Power Reaches Your Building, Part 1: Reliability
The first channel is reliability. Data center loads are large, concentrated and can drop off the grid suddenly, and those sudden losses have caused disturbances serious enough that NERC set mandatory reliability rules for large loads at 50 megawatts and 100 kilovolts, a threshold federal regulators had declined to name. The rules are aimed at keeping a single large load from destabilizing the grid everyone else depends on. For owners, the signal is that grid reliability is now being actively managed around these loads, and that resilience, onsite backup and the ability to ride through disturbances, has risen in value.
How Power Reaches Your Building, Part 2: Rates
The second channel is cost. Data center demand is pushing wholesale power prices up, and the increase flows to commercial bills. In PJM, the largest US grid region, wholesale power costs are running roughly 50 percent higher, and PJM’s independent market monitor attributed about 9 percent of 2026 wholesale costs specifically to data center load. That is a direct, measurable addition to what commercial customers pay, before any capacity charges. The owners least exposed are those who have lowered their own peak demand and generate power onsite; the most exposed are those buying all their power from an increasingly expensive grid.
How Power Reaches Your Building, Part 3: Community and Politics
The third channel is environment. Data center siting has become a political flashpoint. A Memphis moratorium vote collapsed into protests, and moratoriums are spreading across states as communities weigh power demand, water use and rising bills. For developers and owners of power-intensive property, this adds an entitlement and reputational risk that did not exist a few years ago. A project’s access to power is no longer the only gating question; its social license is now part of the calculus.
What Owners Should Plan For
None of these forces reverse on their own, so the planning question is how to reduce exposure to each. On reliability, that means securing critical load with onsite backup so a grid disturbance does not become a tenant’s problem. On rates, it means lowering the building’s peak demand and evaluating onsite generation, since the cost of grid power is set to keep climbing regardless of any single owner’s actions. On siting, for those developing power-intensive property, it means treating community engagement and local approval as early-stage risks, not late-stage formalities. The data center boom is the backdrop every commercial owner now underwrites against. The response is not to fight it, but to reduce the building’s dependence on the grid it is straining.
Frequently Asked Questions
Why should a data center deal between Google and a utility affect my building?
Because it tightens the supply of firm power. When large buyers contract the generation and grid capacity that already exist, less remains for other customers, which pressures prices and reliability for everyone on the same grid. The effect is indirect but real, and it shows up on commercial bills.
Is the 9 percent PJM figure the whole data center impact?
No. The market monitor’s roughly 9 percent is the share of wholesale energy costs attributed to data center load in 2026. It does not include the separate capacity charges, which have risen sharply, or the longer-term grid investment these loads require. The total effect on a commercial bill is larger than that single figure.
What can a mid-market owner actually do about this?
Reduce dependence on the grid the data centers are straining. Lowering peak demand cuts exposure to rising capacity and energy charges, onsite backup protects against reliability disturbances, and onsite generation hedges against the climbing cost of grid power. None of it stops the boom, but all of it reduces how much the boom costs the building.
Sources
SiliconANGLE, Google goes nuclear with Constellation (890 MW new, 2,700 MW supply): https://siliconangle.com/2026/10/06/google-goes-nuclear-to-power-its-ai-data-centers-committing-to-purchase-890-megawatts-of-energy-from-constellation/
Microgrid Knowledge, NERC sets data center reliability rules at 50 MW and 100 kV: https://www.microgridknowledge.com/distributed-energy/demand-response/article/55394284/
Utility Dive, data center load made up 9% of PJM wholesale costs in 2026: https://www.utilitydive.com/news/data-center-load-pjm-wholesale-market/828917/
Action News 5, Memphis data center moratorium vote ends amid protests: https://www.actionnews5.com/2026/10/06/data-center-moratorium-vote-ends-abruptly-amid-protests-fistfight/
The reliability and rate impacts get most of the attention, but the planning challenge feels just as important. Infrastructure has to be planned well ahead of demand, yet the ultimate scale and timing of large-load growth remains uncertain.