Forgive the irony, but I’m using AI to analyze today’s costs of gas turbines vs. solar+battery to run AI data centers
I read an article in today’s Politico Surge newsletter reporting on the ever-higher prices for gas turbines, and in this case, one proposed to power a data center in Kentucky.
I asked Gemini the following query:
How much solar kilowatt production could the following expenditure described in the Politico Surge story produce? How much battery storage? “Consider a seemingly mundane regulatory filing in Kentucky last week, where a local utility submitted an application to build a new combined-cycle natural gas plant needed to power a data center. Its price tag: a cool $2.81 billion, according to regulatory filings.”

Here is the result:
At $2.81 billion—the price tag for Kentucky Power’s proposed 760–769 MW combined-cycle natural gas plant to serve TeraWulf’s data center—the same capital expenditure could buy roughly 1.5 to 1.9 million kW (1.5–1.9 GW AC) of standalone utility-scale solar, 8.4 to 18.1 million kWh (8.4–18.1 GWh) of standalone battery storage, or a combined 1.55 million kW DC of solar paired with 3.73 million kWh of storage. [TECS Engineering]
What $2.81 Billion Buys (U.S. Utility-Scale Benchmarks)
|
Option |
U.S. Cost Benchmark |
What $2.81 Billion Buys |
|
Natural Gas (Kentucky Filing) |
~$3,650–$3,700/kW |
760,000–769,000 kW (760–769 MW) |
|
Solar Only (AC) |
$1,463–$1,865/kW AC (EIA / Southern U.S. avg.) |
1.51 – 1.92 million kW AC (1.51–1.92 GW AC) |
|
Solar Only (DC) |
$1,120/kW DC (DOE 2025 Q1 benchmark) |
2.51 million kW DC (2.51 GW DC) |
|
Battery Storage Only (4-hr) |
$155–$334/kWh (Lazard low-end to NREL U.S. avg.) |
8.4 – 18.1 million kWh (8.4–18.1 GWh) |
|
Solar + Storage Hybrid |
$1,810/kW DC (DOE 100 MW DC + 240 MWh benchmark) |
1.55 million kW DC solar + 3.73 million kWh storage |
Breakdown of the Math
-
- Utility-Scale Solar: Using the U.S. Department of Energy’s (DOE) benchmark of $1.12/W DC ($1,120/kW DC), $2.81 billion buys 2,508,929 kW DC. Converting to AC grid output—where large U.S. plants (>100 MW) average $1,496/kW AC and Southern U.S. projects average $1,463/kW AC—yields 1.87 to 1.92 million kW AC. At the broader U.S. all-sizes average of $1,865/kW AC, it buys 1.51 million kW AC. [SolarInfoPath]
- Utility-Scale Battery Storage: Based on NREL’s U.S. 4-hour lithium-ion benchmark (~$334/kWh all-in, combining energy and power hardware costs), $2.81 billion buys 8.41 million kWh (8.41 GWh) of storage, capable of discharging 2.1 GW for 4 hours. Using Lazard’s lower-end U.S. installed cost of $155/kWh, that rises to 18.13 million kWh (18.13 GWh). (Note: Outside the U.S., where Chinese-manufactured equipment faces lower tariffs, all-in utility storage has dropped to ~$125/kWh, which would buy ~22.5 GWh.)
- Combined Solar + Storage: Using the DOE’s standard utility hybrid benchmark ($1.81/W DC for a 100 MW DC solar array paired with 240 MWh of batteries, or $181 million per unit), $2.81 billion funds roughly 15.5 units—delivering 1.55 GW DC (~1.2 GW AC) of solar plus 3.73 GWh of battery storage.
The Key Trade-Off: Nameplate vs. 24/7 Output
While $2.81 billion buys significantly more nameplate kilowatts of solar and batteries than gas, the two options behave differently for a 24/7 data center:
-
- Fuel vs. Capital: The gas plant’s $2.81 billion is only the upfront construction cost; it will also burn tens of millions of dollars in natural gas annually, whereas solar has zero fuel cost.
- Reliability: A 760 MW combined-cycle gas plant can run around the clock at an 80–90% capacity factor. In Kentucky’s climate, a 1.2 GW AC solar + 3.73 GWh battery system would generate far less total annual electricity and would still require grid backup or additional firm power to survive multi-day winter cloudy streaks.
Do You Know the Answer?
As for the conclusion that the solar+battery setup described by Gemini “would still require grid backup or additional firm power to survive multi-day winter cloudy streaks,” there are other considerations. First, grid connection is a viable solution and one likely desired for the gas turbine situation too, since gas turbines also have downtime (or, alternatively, their own added battery backup investment). Second, different configurations for BESS with the solar could further insulate the facility downtime from diminished solar output. Third, better general management of grid capacity (demand response, VPPs, digital transformers, line sensors, AI-aided balancing, etc.) would offer further capacity support for ongoing sufficient power (this applies to both situations).
What’s your bet? Solar+Batteries are going to be more or less expensive going forward? Are natural gas prices going to be more or less expensive? Are gas turbine costs are going to drop or keep climbing?
Yeah, you know the answers.