Energy

Report: Only 3% of data centers will shift power demand by 2030

The energy think tank Ember estimates that only around 3% of global data center capacity will be able to shift their electricity use by 2030, according to new research out today.

The vast majority, or 87%, of data centers will still run cloud services and corporate IT systems in addition to AI inference, so their power consumption can’t easily be paused or moved, Ember found. Data centers dedicated to AI training, which are a better fit for load shifting, will account for just 13% of the market. 

Limited workload shifting doesn’t necessarily undermine the case for flexible data centers, but suggests that scaling them will likely depend less on rescheduling computing and more on alternative power supplies. In other words, the findings are a potential boost for onsite power solutions like gas turbines and batteries.

The Federal Energy Regulatory Commission in June directed grid operators to justify or revise their rules for connecting large loads, following orders from the Department of Energy to accelerate those connections. Making data centers more flexible — by shifting computing to other times or locations, reducing electricity use at grid operators’ request, or switching to on-site power when the grid is constrained — is seen as a way to secure faster grid hookups and potentially shield other customers from the cost of new grid infrastructure.

The load shifting impact

Ember analyzed nine major data center markets globally including Texas, California, and New York. They found that in the U.S. and most other countries, data center load shifting would reduce peak demand by between 0.1% and 0.6% by 2030. In the U.S., Texas is poised to contribute the most to peak demand reduction through load shifting, with up to 0.4%.

Meanwhile, in places where data centers will account for a higher proportion of peak demand, like Ireland and Malaysia, the potential reduction from load shifting is higher, at around 1.5%.

The biggest potential for grid relief, according to the report, is a combination of load shifting and onsite generation. Onsite generation, if used to help data centers decrease their draw from the grid, offers more relief at peak, while load shifting can contribute towards getting more out of the existing grid, avoiding new infrastructure that would otherwise be rarely used.

To illustrate the benefit, Ember analyzed two AI data centers in Texas, Stargate Abilene and Comanche Circle, which together account for about 5.6 gigawatts of new load online by 2030. At the typical U.S. data center utilization rate of 48%, they would draw an average of 2.7 GW. Both facilities plan to leverage an ERCOT grid connection as well as onsite generation.

Load shifting at the two data centers could relieve .55 GW during a grid stress event. That’s compared to shifting entirely to onsite generation, which could provide the full 2.7 GW of relief — 2% of ERCOT’s predicted 2030 peak load.

The battery potential

Developers seeking fast, firm power are increasingly turning to behind-the-meter gas, even as the risks of equipment shortages, mechanical failures, pipeline constraints, and price shocks are becoming apparent.

In much of  the U.S., the terms of that trade off — faster connection in exchange for curtailment risk — are being written right now. ERCOT and SPP have already adopted non-firm interconnection pathways. PJM, whose own data center curtailment plan was rejected by FERC earlier this month, is in the midst of revamping its framework ahead of the November 16th filing deadline.

According to new data from research platform Currence, non-firm interconnection rules could create a massive market for batteries in the U.S., up to 285 GWh by 2035. That modeling assumes half of forecasted data center load leverages some form of flexibility, and uses batteries to cover curtailments.

The largest market potential by far is PJM, which accounts for more than half of Currence’s projected storage market by 2035. The region’s concentrated data center load and constrained grid make non-firm interconnection a potentially attractive route to power for developers — though it may well be states in the region that ultimately set rules for who gets curtailed first.

Image credit: Currence

Each new flexible data center to come online leaves less spare grid capacity, making curtailment events longer and more frequent moving forward, Currence said.

Flexible data centers may therefore outgrow the standard four-hour battery in the near term, depending on curtailment rules in each market. In ERCOT, for example, if the grid operator implements a uniform rule that spreads curtailments evenly across flexible sites, each data center would only need an estimated 1.3 hours of storage by 2030. But if ERCOT uses a “last-in-first-curtailed” approach, the newest facilities would need at least 5.3 hours of storage duration. If that same approach were used in PJM and SPP, duration needs could reach seven hours by 2035, Currence found. 

The expanding need comes at an interesting moment in the battery market because, according to Currence, long-duration storage is creeping towards cost parity in some scenarios.

For durations of up to eight hours, lithium-ion batteries are still the tech to beat, thanks to plummeting prices, an existing global manufacturing base, and high roundtrip efficiency. That makes them the most likely choice for the near-term battery buildout at flexible data centers. 

Beyond eight hours, some non-lithium storage technologies are becoming cost-competitive. Certain LDES technologies with high capacity accreditation could be a fit for front-of-meter “bring-your-own-capacity” deals, Currence noted. Those longer duration technologies are also being used as backup power, replacing diesel generators at fully grid-connected sites. They can  absorb load swings when AI training jobs start and stop, as well. Government procurement programs are currently the largest source of offtake for non-lithium LDES, with data centers coming in second.

The post Report: Only 3% of data centers will shift power demand by 2030 appeared first on Latitude Media.

via Latitude Media https://ift.tt/9ZMLsYr

Categories: Energy