Grid congestion and energy curtailment have become outsized problems for the state of California, spiking utility rates and defeating the purpose of renewable energy by using local methane gas. A new study from Next 10 says the state’s expansive solar industry could turn that weakness into a strength.

Titled ‘Curtail to Compute: Siting Datacenters to Leverage California’s Stranded Renewable Energy,’ the new report by the San Francisco-based nonprofit details the benefits of siting data centers in rural locations, closer to solar generation sources. The recent rise in energy curtailment has wasted enough energy to power 500,000 Californian homes for a year, with 70% of that wasted electricity tracing back to grid congestion, rather than oversupply.

Typically, the solution to this would be to build more transmission infrastructure, which Next 10 says is both time-consuming and expensive for California residents. F. Noel Perry, the company’s founder, says a suitable alternative is staring the Golden State right in the face.

“For data center owners seeking a cost-effective and clean power supply, California’s strong renewable energy portfolio makes the state an attractive location,” he says. “And existing solar curtailment provides the opportunity for large energy users to step in and address grid congestion while avoiding the need for extensive infrastructure buildout that would otherwise fall on ratepayers. This report shows that cleaner data centers are possible, and California can provide a model.”

As an example, the Pacific Gas & Electric Company’s (PG&E) “Fresno Zone” is the most congested area of the state. Without any intervention, the zone’s ‘Path 15’ transmission line is likely to be congested for up to 84% of the year, according to Next 10. Siting data centers along rural areas would not only solve that problem, the nonprofit says, but save the state of California and its ratepayers potentially billions.

“CAISO estimates that it would cost $700 million to $1.1 billion to build the needed high-voltage transmission lines serving the Bay Area alone,” Next 10 says. “Siting a data center along a rural transmission corridor where solar energy is abundant and otherwise wasted not only allows that power to be used productively, it also avoids the need for expensive infrastructure buildout.”

Comparing operational feasibility

Benjamin Lee, a professor of electrical and systems engineering and computer and information science at the University of Pennsylvania and the study’s lead researcher, says that data centers need cost-effective, low-carbon solutions in order to be environmentally feasible.

As part of the report, Lee and Next 10 analyzed three different options for Californian data center construction. The results, Lee says, show an alternative path forward, straying from expensive, time-consuming infrastructure upgrades.

“Meanwhile, we need to make sure the huge amounts of solar energy the state generates are put to good use, especially in rural areas of the state,” Lee says. “Our models show data centers benefit from added load flexibility when using curtailed energy, and on-site battery storage adds to the efficiency. This approach has the added benefit of reducing the environmental impact of AI, which frequently relies on more polluting energy sources.”

Specifically, the report’s “Urban Baseline” scenario, a hypothetical 20-MW data center in Santa Clara, would rack up just less than $280 million in capital expenditure, plus about $29 million in operational expenditure.

An opposite “Rural Curtailment-Only,” positioned specifically to get the most out of curtailed solar energy, cuts that cap-ex number to $183.5 million, slashing the rent alone to the tune of 94%. Annual operational expenses also drop in this scenario, dropping over three million to $25.9 million per year.

There is, of course, a third option in all of this: a rural, curtailment-focused site equipped with a battery. Aptly named “Rural + Battery Storage,” this scenario is more expensive than the rural option, but could be more reliable in return thanks to a 10 MW battery attachment. The battery option is still cheaper than the urban option at $195.5 million in capital expenditure, and $26.7 million in annual operating expenses.

Working toward an opportunity

Artificial intelligence is not going anywhere, with computer chip manufacturer Nvidia becoming the richest company in the world in 2024. By extension, the truly massive data centers used to power models like ChatGPT and Claude are also likely to stick around.

Next 10 research director Stephanie Leonard says the Golden State’s solar industry has spent enough time missing this opportunity. The best time to start strategizing data center siting was 2024, when the sector truly took off; the second best time is as soon as possible.

“Curtailed solar energy represents a missed opportunity for California to put carbon-free electricity to use,” Leonard says. “But by siting a data center close to the power line congestion, we can move energy demand closer to the source instead of spending more money and resources to move the energy further away from its generation point.”

Of course, there are sources of uncertainty for enormous course changes like this one, with the report citing shifting market dynamics and a possible lack of investment as chief reasons to pump the proverbial brakes. Curtailment itself is still an issue, creating “create both challenge and opportunity for sustainable computing” in equal measure.

Still, Lee, Next 10, and the report find that shifting the data center siting and construction strategy to include solar curtailing and storage options presents “significant economic advantages” in the state of California.

“Building such datacenters will require significant capacity for on-site energy storage to support additional computation when curtailed energy is scarce or permit energy trading in the wholesale market,” the report concludes. “Looking forward, we acknowledge significant uncertainty around revenue models for computation and challenges in finding and serving delay-tolerant workloads.”

Next 10 and the University of Pennsylvania will host a webinar on March 24 at 2 p.m. Eastern, discussing key findings and what it could mean going forward. Registration is open now.

Tags: artificial intelligence, data centers, grid management, Next 10, nonprofit, sustainability, utility-scale