No, AI Data Centers Are Not Raising Your Electric Bill — Here’s the Proof

Headlines across the country blame AI data centers for rising electric bills. A new analysis by SemiAnalysis — one of the most respected independent research firms covering AI infrastructure — reveals the truth: the problem is not data centers. It is a broken auction system in 13 eastern states. One that does not exist in Texas, California, or Imperial Valley.

The report, published March 2026, analyzes the two largest energy markets in the United States — PJM Interconnection (covering 13 eastern states and D.C.) and ERCOT (the Electric Reliability Council of Texas). Both regions are experiencing massive data center expansion. Only one is seeing electricity bills spike. The difference is not the data centers. It is the market design.

The PJM Problem: A 9.3x Price Spike

PJM Interconnection is the grid operator for 65 million people across 13 eastern states including New Jersey, Pennsylvania, Virginia, and Ohio. It operates a capacity market — a system where customers pay power plants to be “available” on standby, even if they only run a few hours per year. This cost is set by an annual auction called the Base Residual Auction (BRA).

In the 2025/26 auction, the capacity clearing price jumped from $29 per megawatt-day to $270 per megawatt-day — a 9.3x increase. The 2026/27 and 2027/28 auctions hit the federally imposed price cap at $329 and $333 per megawatt-day.

The total cost of the 2025/26 capacity auction: approximately $16 billion, spread across every electricity customer in the PJM region. That is $16 billion for power plants to sit on standby.

Why Data Centers Are Not the Culprit

Many immediately blamed surging data center demand. PJM’s own Internal Market Monitor (IMM) ran simulations showing that removing all data centers from the forecast would reduce capacity payments by $9.33 billion — a 64% reduction. Keeping only already-energized data centers cut costs by $7.74 billion.

This looks damning for data centers. But there is a critical detail the headlines miss: the BRA is a simulation, not a real market.

The capacity price is set by PJM’s internal demand and supply forecasts — the Variable Resource Requirement (VRR) curve. This curve is built on PJM’s models, not on what traders, utilities, or market participants actually believe. The forecasts have a documented history of massive errors:

• 2024: PJM’s data center load forecast was cut by 800 MW versus the 2023 forecast.
• 2025: The forecast was cut again by 1.1 GW versus what had been forecasted just one year prior.
• Supply side: PJM methodology changes made 14 GW of natural gas capacity “disappear” on paper — not because plants closed, but because of an accounting rule change.

Texas Proves the System Works

ERCOT manages the Texas grid, which is experiencing the same explosive data center growth as PJM. OpenAI, Google DeepMind, Anthropic, and xAI are all building massive facilities in Texas. ERCOT’s 2025 Long-Term Load Forecast projected 77.9 GW of potential data center load by 2030.

The difference? ERCOT uses an energy-only market — no capacity auction. Prices are set by real-time supply and demand. When electricity gets scarce, prices spike automatically, incentivizing power plants to build and operate. No central planner decides how much capacity the grid needs. No simulation sets costs a year in advance.

The result: stable electricity prices despite massive data center growth. Texas power futures have increased only 11-17% over the past year — meaningful, but nothing resembling PJM’s 9.3x capacity explosion.

Factor	PJM (Eastern U.S.)	ERCOT (Texas)
Market type	Capacity auction (BRA)	Energy-only
Data center growth	Rapid	Rapid
Household bill impact	+$25-30/month	Stable
Capacity price	$270/MW-day (9.3x increase)	N/A (no capacity market)
Winter Storm Fern (Jan 2026)	21 GW lost (15% fleet failure)	Grid held, no emergency
Regulatory reform speed	Years (13 states + FERC)	Months (single state)

The Winter Storm Fern comparison is especially telling. In January 2026, PJM — which paid record capacity prices specifically to guarantee reliability — lost 21 gigawatts of generation, 15% of its cleared fleet, to frozen equipment and fuel delivery failures. The Department of Energy had to issue emergency orders under Section 202(c) of the Federal Power Act. Real-time prices in PJM’s Virginia datacenter corridor spiked to $1,800/MWh.

ERCOT, paying zero for capacity insurance, held its grid through the same storm. No emergency procedures were triggered. Demand ran below forecasts. The post-Uri winterization reforms worked.

Where Does Imperial Valley Fit?

Imperial Valley residents receive electricity from the Imperial Irrigation District (IID) — a publicly owned utility that is not part of PJM, not part of CAISO (the California grid operator), and not subject to any capacity auction mechanism.

This matters enormously. IID:

• Sets its own rates — IID’s Board of Directors, elected by local ratepayers, controls electricity pricing. No federal auction drives costs.
• Owns its own generation — IID operates geothermal, solar, and other power assets. It is not dependent on capacity markets for reliability.
• Has no capacity auction — The mechanism that caused PJM’s 9.3x price spike simply does not exist in IID’s territory.

The IVDC does not just avoid burdening IID’s grid — it actively strengthens it:

• Developer-funded $100M+ 330MW substation — Built entirely at the developer’s expense, not ratepayer cost. This infrastructure becomes a permanent asset for IID’s grid.
• 862 MWh Tesla Megapack battery system — The largest battery deployment at any American data center. It stabilizes IID’s grid during peak demand and renewable intermittency.
• $28.75 million per year in property tax revenue — This directly reduces the tax burden on every IID ratepayer and homeowner in Imperial County.

Data Centers as Grid Resources

The SemiAnalysis report highlights a fact that most critics ignore: data centers can function as grid resources, not just grid loads.

During Winter Storm Fern, the Department of Energy identified 35 gigawatts of backup generation capacity at data centers and industrial sites across the country. DOE issued emergency orders authorizing grid operators to access this capacity. While ERCOT did not need it, the capacity was there — a massive reliability buffer that neither market had priced into forward planning.

The IVDC’s 862 MWh Tesla Megapack battery system already serves this function for IID. When renewable generation fluctuates or demand spikes, the battery system provides instant grid stabilization — benefiting every IID customer, not just the data center.

The narrative that data centers are parasites on the electric grid has it exactly backwards. In Imperial Valley, the data center is building the grid.

The Bottom Line

The next time someone claims data centers will raise your electric bill, ask them one question: which market design?

In PJM’s territory, a flawed capacity auction — driven by simulation-based forecasts with a documented history of errors — is adding $25-30 per month to household bills. In Texas, the same data center growth has produced stable prices and a grid that survived a winter storm better than PJM’s $16 billion insurance policy. In Imperial Valley, the IVDC is investing hundreds of millions into the grid, deploying the largest battery system at any American data center, and generating $28.75 million per year in tax revenue.

Data centers are not the problem. Bad policy is.

Original Article: https://www.ourimperialvalley.com/ai-datacenters-not-raising-electric-bills/