Et si votre facture d'électricité était la prochaine victime de la guerre ?

Key Takeaways

• The 9.5% Shock: US residential electricity prices jumped 9.5% year-over-year in January 2026, with some states seeing increases above 20%.
• The Gas Multiplier: Natural gas generates 40% of US electricity. When Henry Hub averaged $7.72 per million British thermal units (MMBtu) in January 2026, that cost flowed directly into power bills.
• The AI Capacity Tax: The PJM grid’s 2026-2027 capacity auction cleared at $329.17 per megawatt-day (MW-day), adding an estimated $9.33 billion in costs driven largely by data center demand.
• The Compounding Effect: The oil shock and the Artificial Intelligence (AI) buildout are hitting the same grid, the same ratepayer, and the same monthly bill from two independent directions simultaneously.

The Bill Arrives Twice

Your electricity bill does not care about geopolitics. It does not care about Artificial Intelligence. It cares about kilowatt-hours consumed and the rate your utility charges per unit. Both of those numbers are moving in the wrong direction at the same time, for entirely different reasons, and neither force has any incentive to stop.

Here is the situation as of March 2026: Residential electricity prices in the United States rose 9.5% year-over-year in January 2026, according to the Energy Information Administration (EIA). That is more than triple the rate of general consumer price inflation. In the District of Columbia, the increase was 30.3%. In Pennsylvania, 21.7%. In Maryland, 20.9%. These are not marginal increases. For a household spending $200 per month on electricity, a 20% increase is $40 more, every month, indefinitely.

The instinct is to look for a single villain. There isn’t one. S&P Global’s March 2026 CERAWeek report, titled “Electricity Affordability at a Crossroads,” explicitly warned that “the often-misunderstood drivers of recent electricity price increases, in most cases, do not relate to data centers.” The drivers vary by region. But the report also acknowledged a parallel reality: AI data center demand is “testing transmission and distribution infrastructure” in ways that will reshape costs for years.

Both statements are true simultaneously. The crisis is not that one force is crushing the consumer. The crisis is that two forces are converging on the same infrastructure at the same moment, and the consumer has no escape route from either.

Channel One: The Fuel Cost Squeeze

The first channel of the double squeeze is old-fashioned commodity economics. Natural gas is the single largest fuel source for US electricity generation, accounting for approximately 40% of total output in 2025. When the price of natural gas rises, electricity prices follow with near-mechanical certainty. The relationship is not subtle. It is arithmetic.

The January Spike

In January 2026, the Henry Hub natural gas spot price averaged $7.72 per MMBtu — more than double the EIA’s projected 2026 annual average. By February, the monthly average had fallen to $3.62 per MMBtu, and the EIA’s Short-Term Energy Outlook (STEO) projects a 2026 average of approximately $3.80 per MMBtu. The January spike was driven by a severe winter storm in the Southeast that pushed electricity demand close to all-time peaks across the Southern Company service territory.

But the January spike is not the real threat. The real threat is that the floor has shifted.

The War Premium

Brent crude averaged $69 per barrel across 2025. Then the retaliatory spiral between the United States and Iran detonated. The US strike on Iran’s Kharg Island oil terminal on March 13, 2026, followed by Iran’s retaliatory offensive across four countries, pushed Brent crude past $115 per barrel by March 19, 2026. The International Energy Agency’s (IEA) emergency release of 400 million barrels from strategic reserves failed to stop the climb.

Why does a crude oil price affect your electricity bill? Because oil and natural gas markets are physically and financially linked. Liquefied Natural Gas (LNG) flows through the same Strait of Hormuz chokepoint that crude oil does. Approximately 20% of global LNG trade passes through the strait. When that flow is disrupted, European and Asian buyers bid up LNG prices on the global spot market. That pulls US LNG export volumes higher, tightening domestic supply. The EIA itself noted that “reduced liquefied natural gas flows through the Strait of Hormuz have increased prices in Europe and Asia.”

The transmission mechanism works like this:

$\Delta P_{\text{electricity}} \approx \text{Heat Rate} \times \Delta P_{\text{gas}}$

A typical natural gas combined-cycle power plant has a heat rate of roughly 7,000 BTU per kilowatt-hour (kWh). If the Henry Hub price rises by $1.00 per MMBtu, the marginal cost of gas-fired electricity rises by approximately $7 per megawatt-hour (MWh). A sustained $2.00 per MMBtu increase over the EIA’s $3.80 baseline would add approximately $14 per MWh to the cost of gas-fired generation. Across the 40% of the electricity mix that runs on gas, that flows directly to ratepayers.

This is not a hypothetical. The EIA’s own analysis shows that in a high-demand scenario, the average 2027 wholesale electricity price across major hubs would rise $2.10 per MWh above the baseline forecast of $48 per MWh. In the isolated ERCOT (Electric Reliability Council of Texas) grid covering most of Texas, the increase would be $37 per MWh, a 79% jump.

Channel Two: The Infrastructure Tax

The second channel of the squeeze is structural. It does not spike and recede like commodity prices. It ratchets upward and stays.

The Capacity Auction

Every few years, the PJM Interconnection (the grid operator covering 13 states from Illinois to Virginia and Washington, D.C.) holds a capacity auction. Utilities and power generators bid to guarantee they will have enough generation available to meet peak demand several years out. The cost of those guarantees is passed directly to ratepayers.

In July 2024, the PJM capacity auction for the 2025-2026 delivery year cleared at $269.92 per MW-day. The auction for the 2026-2027 delivery year cleared at $329.17 per MW-day. S&P Global estimates that data center growth added approximately $9.33 billion in consumer costs during the 2025-2026 auction cycle.

This cost does not appear on your bill as a line item labeled “AI surcharge.” It is embedded in the base rate. It arrives as a general rate increase that your utility attributes to “infrastructure investments” or “reliability improvements.”

The 15-Gigawatt Gap

The reason the auction prices are climbing is supply-demand physics. PJM forecasts that peak electricity demand will grow by nearly 29 gigawatts (GW) by 2030, driven significantly by data center expansion and broader electrification. Planned firm capacity additions total only approximately 14 GW.

That leaves a projected shortfall of 15 GW. To put that in perspective, 15 GW is roughly the output of 15 nuclear reactors, or about 30 large natural gas plants. That gap does not close by itself. It closes by building expensive new generation and transmission infrastructure, and the cost of building it gets socialized across every ratepayer in the PJM footprint.

The EIA projects that US electricity load will increase by 1.9% in 2026 and 2.5% in 2027, with the highest growth concentrated in PJM and ERCOT, precisely the regions where data center buildout is most aggressive.

The Load Profile Problem

The deeper issue is not just that data centers consume a lot of electricity. It is how they consume it. A typical residential customer uses power unevenly: air conditioning peaks in summer afternoons, heating peaks in winter mornings. This variability allows grid operators to schedule generation efficiently.

Data centers consume power at nearly 100% utilization, 24 hours a day, 365 days a year. They do not modulate. They do not participate in demand response programs. If an Artificial Intelligence (AI) training cluster draws 100 MW, that 100 MW runs every second of every day for months. This “flat load” profile means data centers contribute zero flexibility to the grid while consuming maximum capacity. In grid engineering terms, they present constant demand, zero elasticity, and unlimited willingness to pay.

The EIA confirmed that if demand grows faster than expected, “natural gas is the primary source of available power generation to meet incremental demand,” with an assumed increase of about $0.50 per MMBtu in the cost of natural gas delivered to power generators. That increase, too, flows to ratepayers.

Where the Two Channels Converge

The critical insight is that these two forces are not just additive. They are compounding on the same infrastructure.

The Gas-on-Gas Collision

When the Strait of Hormuz blockade tightens global LNG supply, it pulls US natural gas prices higher. Simultaneously, when data center demand accelerates faster than new generation comes online, the grid leans harder on existing gas-fired plants, pushing utilization rates up and further tightening domestic gas supply.

The EIA’s high-demand scenario projects natural gas generation increasing by 7.3% between 2025 and 2027, compared to 1.7% in the baseline. That 123 billion additional kilowatt-hours of gas-fired generation requires burning more gas at the exact moment geopolitical forces are restricting the global gas supply.

The math is simple but brutal. The consumer faces:

$\text{Total Bill Impact} = \underbrace{\Delta P_{\text{gas}} \times \text{Heat Rate}}_{\text{Oil Shock Channel}} + \underbrace{\Delta \text{Capacity Cost}}_{\text{AI Infrastructure Channel}}$

Neither term in that equation is under the consumer’s control. The gas price variable is set by a war in the Middle East. The capacity cost variable is set by the capital expenditure decisions of Microsoft, Google, and Amazon.

The Regional Guillotine

The convergence is not uniform. Some regions are caught in the crosshairs of both channels simultaneously:

Maryland, Virginia, and D.C. (“Data Center Alley”) are in the PJM footprint, absorbing both the capacity auction increases and the gas price transmission. Maryland saw a 20.9% increase in average revenue per kWh in January 2026. The District of Columbia saw 30.3%. These regions also host the densest concentration of data centers in the world, meaning the infrastructure channel hits them hardest.

Texas (ERCOT) is the most exposed region to the fuel price channel because its isolated grid cannot import power from neighboring states during price spikes. The EIA’s high-demand scenario shows ERCOT’s wholesale price surging $37 per MWh, a 79% increase, in 2027. Texas is also experiencing aggressive data center buildout, with ERCOT projecting the fastest load growth of any region.

New England and New York have the highest existing electricity prices in the contiguous US. Massachusetts averaged 27.61 cents per kWh and Rhode Island 27.23 cents per kWh in January 2026. The EIA projects an additional $3.00 per MWh (5%) wholesale price increase in the high-demand scenario for these regions. These states are heavily reliant on natural gas for electricity generation and imported LNG, making them acutely sensitive to the Hormuz-driven global gas price increases.

The Steelman: Why This Might Not Be a Crisis

The strongest counter to this thesis comes from the very report that flagged the problem. S&P Global’s “Electricity Affordability at a Crossroads” analysis explicitly states that the largest retail electricity price increases over the past five years “in most cases, do not relate to data centers.” The drivers are “varied considerably” by region: California’s increases stem from wildfire mitigation costs, New England’s from pipeline constraints, and the mid-Atlantic’s from a mix of factors.

This is a legitimate corrective to simplistic “blame Big Tech” narratives. The rate hikes recorded through January 2026 are largely driven by infrastructure aging, extreme weather events, and the legacy costs of decades of deferred maintenance. Data centers are not yet the primary line item on most utility bills.

But the steelman has a critical expiration date. The PJM capacity auction data shows where the system is heading. A $9.33 billion cost addition from data center demand growth is not a rounding error. A 15 GW capacity shortfall is not a footnote. The question is not whether the AI infrastructure channel will become a primary driver of electricity costs. The question is when. S&P Global’s own declaration that “the era of linear energy transition has ended” suggests the structural pressures are accelerating, not moderating.

The Federal Reserve Cannot Fix This

The macroeconomic trap is that neither channel of the squeeze responds to monetary policy.

The Federal Reserve’s primary tool for combating inflation is interest rate manipulation. Raising rates reduces demand for credit-sensitive goods like cars and houses. But electricity is not credit-sensitive. It is survival. Households do not stop using air conditioning because the federal funds rate rises 25 basis points.

And the oil shock channel operates entirely outside domestic monetary control. The Fed cannot reopen the Strait of Hormuz. It cannot rebuild Kharg Island. It cannot compel Iran to stop retaliating. Those factors are set by geopolitics, not economics.

The result is a textbook case of the vise that nobody can open: inflationary pressure from commodity supply shocks that the central bank’s toolkit cannot address. Electricity bills rise. Grocery bills rise because refrigeration and food processing are electricity-intensive. Transportation costs rise because gasoline comes from the same crude oil. The consumer is squeezed from multiple directions simultaneously, and monetary policy is structurally powerless to provide relief.

What Comes Next

The Summer Test

The convergence point is July through September 2026. Summer electricity demand peaks as air conditioning load surges. If the Strait of Hormuz remains contested through Q2, gas prices will be elevated entering the hottest months. Data center load does not decrease in summer. The two channels will hit their maximum simultaneous impact during the exact months Americans use the most electricity.

ERCOT is the canary. Its isolated grid and aggressive data center expansion make it the most vulnerable to a compound spike. The EIA already models a 79% wholesale price increase in ERCOT under high-demand conditions. Layer a sustained gas price premium from the Hormuz crisis on top of that, and the possibility of $200-plus MWh wholesale power in Texas during August peak hours is structurally plausible.

The Political Trigger

Electricity bills are uniquely visible to voters. Unlike inflation in healthcare or insurance, which is abstracted through premiums and deductibles, the electricity bill arrives every month with a specific dollar amount. A 20% increase is impossible to ignore and impossible to spin.

If the double squeeze persists through summer, expect state public utility commissions to face intense political pressure. The question of who pays for the grid upgrades required by data centers will shift from a regulatory proceeding to a political campaign issue. The answer will determine whether the cost of the AI buildout continues to be socialized across all ratepayers, or whether data center operators are required to bear a larger share of the infrastructure they consume.

The Structural Ratchet

Here is the uncomfortable truth about the double squeeze: one channel is temporary, and one is permanent.

The oil shock will eventually end. Wars conclude, shipping lanes reopen, and commodity prices normalize. The fuel cost channel, however painful, is cyclical.

The infrastructure channel is not. Once a capacity auction clears at $329.17 per MW-day, those costs are locked in for years. Once a 15 GW shortfall drives billions in new generation construction, those capital costs are amortized across ratepayers for decades. The AI buildout does not slow down when the war ends. It accelerates.

The consumer who survives the summer of 2026 will find that the war premium on their gas bill eventually fades. But the AI premium on their capacity charge will still be there. And it will be larger the following year.

The double squeeze is two crises arriving in the same envelope. One is a fire. The other is the foundation settling. The fire gets the headlines. The settling does the permanent damage.

---