https://vimeo.com/share/6b64233d-9b4f-4b14-a94a-ca7e72b1d50e
How AI Data Centers Could Be Driving Up California Electric Rates Before They’re Even Built
Buried in next week’s San José City Council agenda are two items that, at first glance, appear unrelated.
The first would authorize San José Clean Energy (SJCE) to enter into a 15-year agreement worth up to $124.5 million for a share of a geothermal energy project in Utah. The second is the community choice aggregator’s latest Integrated Resource Plan (IRP), a long-range blueprint showing how the City expects to meet customer electricity demand through 2045.
Read separately, they are routine energy policy items. Read together, they tell a much bigger story about artificial intelligence, data centers, and the growing financial risks embedded in California’s electric rates.
The Utah agreement is particularly revealing because San José is not purchasing power directly from the geothermal developer, Fervo Energy. Instead, the City is buying a percentage share of Shell Energy North America’s existing power purchase agreement. Because Fervo is no longer directly selling capacity from this phase of the project to new buyers, stepping into Shell’s contract is the only practical way for San José to secure the resource. The electricity generated in Utah isn’t destined for Silicon Valley in any direct physical sense; rather, the purchase is an expensive rush to meet California’s increasingly complex clean energy, resource adequacy, and regulatory compliance requirements.
That alone raises an important question: Why are utilities and community choice energy providers across California being forced to scramble for such scarce, premium-priced resources?
Grid Reliability Mandates and the Race for Clean Energy
The short-term answer is a series of rolling mandates from the California Public Utilities Commission (CPUC) designed to prevent grid failure. Driven in part by the looming retirement of traditional baseload facilities like the Diablo Canyon Power Plant, regulators have ordered load-serving entities to lock down “firm clean resources”—like geothermal power—that can generate electricity 24 hours a day, rain or shine.
But the long-term, multi-billion-dollar answer can be found in the Integrated Resource Plan.
According to San José Clean Energy, future electricity demand is projected to increase dramatically, driven by electric vehicles, building electrification, and “large-load facilities”. That last category heavily features industrial data centers, which have become the central focus of the energy industry as artificial intelligence drives an exponential demand for computing power.
Anyone following the AI sector has seen the headlines: technology companies are announcing massive investments in data centers, utilities are scrambling to project the load, and regulators are attempting to ensure sufficient generation is built before the demand arrives.
The challenge is that nobody knows how many of these projects will actually get built, how quickly they will come online, or how much electricity they will ultimately consume.
The Financial Risk of Aggressive Electricity Demand Forecasts
What makes San José’s report noteworthy is that the City openly acknowledges this fog of uncertainty. Rather than blindly relying on the CPUC’s assigned forecast—which mandates an aggressive growth model based on older state data—San José Clean Energy modeled its own alternative scenarios. Staff looked at data center permitting activity, customer commitments, self-generation potential, supply chain constraints and other factors to develop what they believe is a more realistic forecast.
The numbers resulting from this analysis are striking.
The report concludes that the Net Present Value cost of the portfolio required under the CPUC’s aggressive forecast is approximately 70 percent higher than San José’s Alternative Portfolio. Even more telling, it is 135 percent higher than the City’s Internal Portfolio—a conservative, lower-bound scenario that models minimal, flat load growth to chart a path to carbon neutrality by 2030 absent massive data center expansion.
Just as notable is the explicit warning issued by city staff: if actual load growth through 2035 turns out to be closer to local, conservative expectations rather than the state’s aggressive projections, San José will be heavily overcommitted. Ratepayers could be exposed to massive market recovery risks, forced to sell off excess power contracts at a loss for energy that was never actually needed.
In plain English, California’s energy planners are demanding that communities sign decades-long, multi-million-dollar procurement contracts today based on macroeconomic AI forecasts that may or may not prove accurate.
Protecting Ratepayers Amidst Tech Infrastructure Spikes
That should matter to every ratepayer. For many Californians, electricity bills are already becoming one of the fastest-growing household expenses.
California has already experienced painful increases in electricity costs over the last decade. Wildfire mitigation investments, transmission upgrades, renewable energy mandates, and grid reliability requirements have all combined to apply continuous upward pressure on monthly bills. Now, a new compounding factor is emerging: the financial requirement to over-prepare for a future that assumes rapid, unprecedented electricity consumption from technologies that have not yet fully arrived.
To be clear, this is not an argument against artificial intelligence or data centers. Silicon Valley has always thrived on innovation, and AI will undoubtedly be one of the primary economic engines of the next few decades.
The question is how much structural financial risk should be placed squarely on the backs of existing ratepayers while that future remains deeply uncertain?
If regulators underestimate demand, California faces reliability challenges and power shortages. But if they overestimate demand, community choices and utilities will lock customers into decades of expensive over-procurement commitments. Neither outcome is desirable.
That is why the cautious language in San José’s Integrated Resource Plan deserves statewide attention. City staff are not arguing against clean energy procurement, nor are they opposing reliability mandates. Instead, they are explicitly making the case for a phased, flexible approach. Their warning is clear: aggressive, top-down state forecasts risk triggering binding procurement obligations long before the underlying commercial demand actually appears, threatening rate affordability.
The proposed Utah geothermal agreement is a perfect example of this tension. It may well prove to be a necessary investment to fulfill immediate state reliability mandates. But it also represents the leading edge of a much larger, highly speculative trend: locking in long-term commitments based on the assumption that the AI boom will expand the grid indefinitely.
Data centers may eventually require all of this new generation and more. But if there is one historical lesson the technology industry has taught us, it is that long-range demand forecasts are frequently wrong.
As California’s energy policies become more tightly bound to the infrastructure demands of artificial intelligence, policymakers must remember that every aggressive forecast carries a steep downside risk. And the people who ultimately pay for that risk are not the data center developers, the state regulators, or the utilities.
They are the ratepayers.