This is a different kind of Powerstack. We usually bring you the news in power and data centers with spicy analyst takes. This edition focuses on one single piece of Currence research and how we built it.
The big question
For the past few months, one question came up on client calls and at conferences and in our analysts’s stress dreams. If data centers start accepting curtailment to connect faster, how much storage will they need?
ERCOT and SPP adopted non-firm interconnection this year. FERC rejected PJM’s first design in October. And now grid operators are set to file their large load proposals in mid-November. Those proceedings will set the rules for the largest data center markets in the country.
We looked for a credible number and couldn’t find one. So our storage analyst Lukas Karapin-Springorum built a model to answer it.
All the goodies are in this report. You can download the report here.
This is Currence’s weekly newsletter on the moves and motives shaping the load growth era. Not a subscriber yet?
How to think about the problem
A flexible data center agrees to cut load when the grid is stressed. A battery behind the meter keeps the servers running through each cut. So the storage question breaks into three smaller ones.
How much load goes flexible?
How long does each curtailment last?
Which technology covers that duration at the lowest cost?
How we built the model
Lukas started with the data center load forecasts from ERCOT, SPP, PJM, and MISO. He then modeled when each grid runs short of spare capacity and how long each shortfall lasts.
Then he connected it to the data our clients use every day:
Cost benchmarks and readiness scores for every major storage technology
1,250+ tracked data center projects, each with a credibility score
30,000+ US power projects linked to those data centers
Our speed-to-power model, which prices what an early grid connection is worth to a hyperscale campus
What we found
If half of forecasted data centers go flexible, we could need 285GWh of storage across ERCOT, SPP, PJM, and MISO by 2035. Demand starts small, at 17GWh in 2028, then grows quickly as more load connects.
In our model of a hyperscale campus, getting power one year early adds $6.3m per MW in revenue. That makes a battery a cheap-ish way to buy speed.
Battery size is the bigger open question. Under one curtailment rule, today’s standard 4-hour battery is enough. But flexible data centers will outgrow that by 2030 in every market we modeled.
The full report breaks down storage demand by market and year, shows the battery duration each market needs under different curtailment scenarios, and compares the cost of capacity across lithium-ion and non-lithium technologies.
The case for non-lithium
Longer batteries change which technologies can compete. Lithium-ion is still the default for short durations. At eight hours and beyond, several non-lithium technologies now match it on cost. In our cost model, Energy Dome, XL Batteries, and Hydrostor all offer cheaper capacity than an 8-hour lithium-ion battery.
Questions worth asking
If you develop power: Which market has the most flexible load and the least storage built for it today?
If you build storage: Will your product fit the duration buyers need in 2030?
If you develop data centers: Should you own the battery, contract it, or buy the developer?
If you invest: Which existing assets gain value when flexible data centers arrive?
This report covers the market by region, the duration curves under both rules, and which technologies are ready to compete. The full model, with the data behind every chart, is available to clients on the Currence platform, and through the API and the MCP.
Walk through the model with us
On November 5, Lukas will walk our clients through the full model. If you're interested in learning more talk to our team and we can share an exclusive invite to our client-only webinar.
Interested in diving deeper? Talk to our team and learn how the teams that finance and build the energy system use Currence to stay ahead in the energy and AI buildout, including Microsoft, Southern Company, Tokyo Gas, Jefferies, Galvanize, B Capital, and others.
