Utility-scale battery capacity tripled in two years. California and Texas are the giants.
By Wolf Richter for WOLF STREET.
Utility-scale battery systems have been booming, especially in Texas and California, the two states with by far the largest installed capacity. In the US overall, installed battery capacity shot up from less than 1 gigawatt (GW) in 2020 to 27 GW in 2024.
These batteries serve a variety of uses, but the top use by battery operators is price arbitrage: Buy electricity when demand and prices are low, such as at night, and sell electricity when prices are highest such as during peak demand times in the day, and profit from the spread. Wolf Street has long discussed the boom in battery installations, and their use for price arbitrage. Here are some numbers from the EIA about the use of price arbitrage.
Of the 27 GW of installed utility-scale battery capacity in the US (gray columns), 66% had price arbitrage among its uses (red columns), and 41% of the capacity was primarily used for price arbitrage (brown columns, chart via the EIA).
The California Independent System Operator (CAISO) reported 11.7 gigawatts (GW) of installed utility-scale battery capacity in 2024, more than double the capacity in 2022, and up from near zero in 2020.
In 2024, 43% of total installed utility-scale capacity was primarily used for price arbitrage (left chart below, dark brown columns, chart via the EIA).
Of the US states, California is #1 in solar power generation and #6 in wind power generation.
The Electric Reliability Council of Texas (ERCOT) reported 8.1 GW of installed utility-scale battery capacity in 2024, up from near zero in 2020.
In 2024, 50% of the capacity was primarily used for arbitrage (right chart, dark blue columns).
Texas is #1 in wind power generation and #2 in solar power generation.
California and Texas combined accounted for 73% of total US utility-scale battery capacity in 2024.
Far behind price arbitrage, the other big uses of battery capacity were in that order: frequency regulation, excess wind and solar generation, system peak shaving, load management, co-located renewable firming (pairing wind and solar with battery systems at the same site), and others.
The EIA obtained this data from its annual survey of power plant activity, where it asked battery operators to identify the primary use case for their battery systems.
And in case you missed it: Demand for Electricity Takes Off. US Power Generation by Source in 2024: Natural Gas, Coal, Nuclear, Wind, Hydro, Solar, Geothermal, Biomass, Petroleum
Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the mug to find out how:
“Buy electricity when demand and prices are low, such as at night, and sell electricity when prices are highest such as during peak demand times in the day, and profit from the spread.”
Cali utilities effectively do this via net metering (with NEM2, and even more so with NEM 3). They’re soaking up all that excess solar production at the wholesale rate from rooftops, and then selling it back at retail. The spread is crazy wide — we were paid less than 2 cents/kwh for our annual net overproduction. Time for an EV, to soak it up myself! Or, to run the heater a few degrees warming all winter.
I find the electricity mkt fascinating – especially with data center mania. It’s nice perspective to see the Primary Case chart, and get a feel for the mix of activities that are parts of electricity prices — and how the options can contribute to profit, as well as rate setting.
I knew batteries were part of backup storage, just as is natural gas and even coal, but to see the batteries as a frontline mechanism for spinning cash flow, adds to my thoughts about data center buildout.
I can’t help think that the Worldcom overcapacity story, for fiber, is playing out exactly the same way with processing capacity going forward. As I recall from a recent story, there’s still a lot of dark fiber — still unused from the Dotcom buildout — and those shareholders are still waiting for their investments to return something other than empty hype.
I can see the beauty of arbitrage and think the batteries are a game changer — but I hope our electricity providers don’t get caught up in mania, that ends up in financial disaster.
“overcapacity story, for fiber, is playing out exactly the same way with processing capacity going forward.”
The AI hype (and related data center/electricity hype) is worth a series of posts all on its own.
Although it is fairly hard to imagine the pretty smart people at Google, Microsoft, Facebook (although…Metaverse…), etc. foolishly burning tens/hundreds of billions in over-hyped capex…it is also fairly hard not to have some serious doubts (which is more and more a common perspective).
Leaving aside the whole “what is really being pitched as AI” debate and “show me the business model/breakeven analysis” (which is a pretty spicy meatball to leave aside. too)…a fairly huge chunk of software development history demonstrates advancing *efficiency* of code – lowering the hardware demands.
And lower cost hardware rendering prior generations of hardware fairly obsolete very quickly.
Sure, there is astronomical bloat in things like MS’ operating systems – but that is because they are trying to interoperate/add functionality in ways not remotely dreamed of 30 years ago. And MS has valued “speed-to-market” over code efficiency for a long, long time.
A more relevant example might be software/hardware driven video compression – with today’s vast online streaming options being utterly impossible 25 years ago (but the tech shift happened a lot faster than that).
Yet we’ve got hundreds of billions in planned capex (ditto data centers and increased electricity supply) based on a “technology” that can easily be described as over-hyped, with a very shaky business model.
There’s a dark fiber line in front of my house. Based on what I’ve been able to find out, it has been there since the 1990s.
Apparently, it was laid to be ready for all that telephonic traffic that would require, well, dark fiber. And then cellular telephony took off.
Dark fiber? I can’t remove it. And I don’t know who can at this point. I just landscape around it.
Well of course dark fiber is a meme from the past.
Batteries are great in Texas for filling the 5PM-9PM gap when solar generation drops off, but AC load is still there. Recharge Midnight to 6AM from wind power. (One typically has decent wind blowing at night in West Texas.)
By having batteries and customers that will load follow (BTC Miners, etc.) ERCOT is able to reach higher levels of renewables generation as a percent of overall generation. ERCOT has been a great innovator on this front.
I have read several local articles (I am a Texan) where private equity firms are making an absolute financial killing putting batteries in for power arbitrage. With the general public supporting the effort since such projects helping to drive the overall renewable power percentage up. One is getting paid big returns and driving a social good. Does not get better than that.
Not sure the general public really understands that they’re helping to pay for these new data centers (and crypto mining, although it seems most of the huge new loads are data centers). Additional load requires additional generation and lots of transmission lines. New generation and transmission is at a much higher cost than embedded (“old”) facilities, meaning the average cost is going up, and all utilities in ERCOT set rates based on average cost (and transmission costs are actually spread out over all the distribution utilities in Texas, by each distribution utility’s contribution to the the ERCOT monthly system peak load during each of the four summer months). Both data centers and crypto miners use a LOT of power (and water), but don’t really provide a lot of post-construction jobs. The general public would probably be better off financially without either data centers or crypto miners, not having to build as much renewable power, and likely lower rates, although ERCOT is seeing quite a bit of co-location of solar/wind with the data centers, and also with batteries, which can reduce the need for new transmission, and the loads are essentially buying directly from the new solar/wind farms, thus reducing the impact of new generation costs.
I don’t think the people understand at all that they are the fools paying for the surveillance business plans of the successful companies.
Buh-buh-buh we’re being watched for our safety! And our security!
(Sarcasm off.)
If electric power is being “shelved” and inventoried for arbitrage trading, is there a “shelf-life” issue, or can this unused inventory be held indefinitely.
Does battery-stored electricity leak or degrade?
Thanks for the interesting post.
Just read the article instead of just the headline. It says:
“…the top use by battery operators is price arbitrage: Buy electricity when demand and prices are low, such as at night, and sell electricity when prices are highest such as during peak demand times in the day, and profit from the spread.”
It’s not like they’re putting electricity on a shelf and letting it sit for three years. The cycles are hourly and daily.
Any power pulled from the grid and stored in the batteries should be used later that day when power is more expensive. Storage time would be less than 24 hours.
The round trip efficiency of these systems would be somewhere in the 90-95% range. So yes, there are losses. Anytime you do something there’s always some kind of losses… friction… drag… Those losses should be baked into the profit and loss cake business case for these systems.
The primary loss in this case is from the internal discharge rate which every battery has, and which is worse for rechargeable chemistries. Which is why your device batteries die just sitting on a shelf.
So yes, the cycle time should be in hours.
This is a great article and I’m glad that it appears in your site… a business, finance and money site. California and Texas are #1 and #2 in solar, and now you’re telling us they are #1 and #2 in battery energy storage as well. These two technologies go together exceptionally well as they override the negative affects of each other.
I love that you point out that the #1 reason battery energy storage is being put on the grid is… to make money. Not to be green. Not to fulfil some government mandate. Not to appease shareholders. It’s to make money.
California and Texas couldn’t be farther from each other regarding political leaning, and yet they’re #1 and #2 in solar generation and battery storage. What’s the commonality? Deregulated electrical markets. No power monopolies. In this environment cost and profit reign supreme. The technologies of solar and battery have gotten so cost effective, that they are exploding in terms of # of projects being installed. And why is that? You make the most money off of these systems, because the #1 cost of electrical production has been removed. No fuel costs.
I was thinking the same thing. Florida has a regulated, monopolistic energy market and I have been happy about that after hearing horror stories out of Cali (Enron) and Texas (lousy maintenance). But it appears to be the key to going green in this country. So the process looks like deregulate prices -> install batteries -> install solar for lowest costs.
Natural gas and coal are not going away, but the increased use of renewables benefit from battery technology, which can increase profits — but the gold rush for all utilities yo overbuild is the greatest threat, in this arms race.
The threat comes from overcapacity and too many stranded assets. That end up raising rates for utility customers, who don’t require cutting edge super computing calculations, like, using a, “supercomputer to incorporate a time dimension into seismic imaging and to examine how oil reduces the strength of sound waves passing through it”.
Between the greed of shareholders and regulator inaction, we can expect massive AI based data center overkill, which will astronomically explode electricity rates for customers who will be the casualties of this arms race — and that’s criminally disgusting!
Redundant, Amen to your thoughts. Like all industrial/electronic U.S. innovation the profit motive drives the circus train. We do need to include innovation for the sake of human existence, polio vaccine, as not so much profit/money driven. Very few if any venture capitalist were involved in polio research. Perhaps, mental disease doesn’t have the cost to return ratio accepted by the money grabbers or it is the symptoms.
“Between the greed of shareholders and regulator inaction, we can expect massive AI based data center overkill”
Agreed – it is fairly un-nerving to watch guaranteed-profits-regardless-of-results utilities drool at the thought of promised massive “AI” electricity demand – when the AI itself is pretty massively over-hyped and discussed in extremely fuzzy terms (actual software processes, business models, etc.)
The thing about being green is that it’s been sold as a feel good thing when it really does come down to money. Environmental clean up has cost the taxpayers vast amounts of money after a lot of businesses made their money and got out of dodge (see mining, lumber, chemicals, etc..) Those businesses wouldn’t have been anywhere near as profitable and their products no where near as cheap if the clean-up costs had been added in.
Being green was a poorly thought through marketing campaign and frequently had nothing to with being green.
Traditional electric energy production from coal, nat gas and nuclear is “thermal power.” Steam or gas turbines convert thermal energy to mechanical energy. Then it’s turned into electrical energy.
But, connecting renewable energy from wind and solar creates volatility and randomness which present challenges to the stability of the power grid’s frequency.
Using large-scale battery packs, which are connected together in series and parallel can reduce the time needed to bring back frequency stability faster than trying to do so with the thermal power alone. Batteries can quickly release power into the grid at the beginning of a frequency drop, and make up for the slower response time of thermal power units. It is a complex process with a lot of variables.
“Learn something new everyday @ WolfStreet!”
Until reading Wolf’s report on this last evening, I didn’t know that battery packs were used for frequency regulation in this way. But when reading WolfStreet, I wanted to learn how it’s is done.
A paper published on 7 December 2022, “Research on the Frequency Regulation Strategy of Large-Scale Battery Energy Storage in the Power Grid System”; which has Bin He from the Collage of Energy and Power Engineering, Inner Mongolia University of Technology, as the lead author addresses this in detail. It is an interesting read on the application of this fairly new technology.
“Large-scale batteries help stabilize the grid.” Should not be news to anyone in the back half of 2025…
R C T C,
Yes, that’s a good point about stabilizing the grid with batteries.
Yet, I had not looked at, or though about how this happened until reading this report.
With batteries, the time to restore precise frequency stability control can be achieved in about one-third the time it takes with thermal energy alone. We’re talking frequency deviations of 0.125 Hz.
But there are quite a few trade-offs. Battery life is one of them. You don’t want to overcharge or over-discharge them. Integrating a State Of Charge feedback loop that’s put into the control mechanism, can improve the utilization of the battery-pack. Utilities need a blend of thermal and battery sources to maintain the grid system and a control strategy to coordinate everything.
It is not a simple task to use batteries with thermal power to keep the grid in perfect condition. The paper I referenced above should be an instruction manual for utilities that employ battery packs together with thermal energy!
An edit for the last sentence:
… an instruction manual for utilities that employ battery packs together with wind or solar, and thermal energy.
While batteries can help address frequency changes, inertia from rotating generators (fossil fuel, not wind) is actually the first line of defense against decaying frequency due to a sudden loss of generation. Inertia is passive, and actually acts to dampen frequency changes from even happening, while frequency response from whatever source is active and is in response to frequency changes, and takes longer to have an impact than does inertia. Having a sufficient amount of inertia is critical, and why grids that think they can get rid of all rotating mass generation are likely going to not be reliable in the long-term.
Legal Economist,
The paper referenced has this to say about inertia from rotating generators, and it confirms your reply:
“The grid frequency is related to the rotational speed of conventional thermal power units in some way. In case of power shortage and frequency fluctuations in the power grid, the conventional thermal power units usually respond to to frequency changes within 5 seconds and gradually increase active power within 10 to 15 seconds, due to their characteristics such as large rotational inertia and slow response speed; until the frequency returns to stability.”
“The battery energy storage system is used to compensate for the power shortage of thermal units in the first 5 seconds to achieve the purpose of regulating the frequency stability of the grid system.”
L E,
Your last sentence is so true.
“Having a sufficient amount of inertia is critical, and why grids that think they can get rid of all rotating mass generation are likely going to not be reliable in the long-term.”
Sorry if I seem exasperated and preacher-like, but AI is in my crosshairs.
I’m fascinated by the mix of battery technology and the clever, profitable cycling between energy production and storage, and then turnaround arbitrage — but, it’s the underlying core narrative of AI that pisses me off.
We’ve all seen our smartphones become faster and cooler, but as that miracle has evolved, electricity consumption has been kinda flat, primarily because of technological advances — but, 3 years ago, the ChatGPT narrative mutated into mythical status, based 100% on future expectations, that are manufactured by the people selling snake oil fantasy.
The utilities are excited about building out massive capacity — to meet future, unrealized demand, but, if big customers don’t provide enough revenue, mom and pop are gonna pay for this stupidity.
I’m guilty of sharing grandkid videos and using capacity for stupid stuff, but, just as with internet buildout, how much capacity is needed for AI to be dramatically necessary?
The numbers don’t add up:
1: Predictions indicate that generative AI could generate $1.3 trillion in total revenue by 2032 across the entire technology ecosystem.
2: A McKinsey & Company analysis projects that global demand for AI compute power alone will drive $5.2 trillion in data center capital expenditure by 2030
The massive, transformational trillions going into everything AI seems very misguided, primarily, because, AI by its nature will become more efficient, far faster than data centers can be built — this is Woldcom 2.0, but systemically dangerous in terms of the scale of future failure.
Short term, maybe a few energy companies can be good investments as they adapt to smart technologies like batteries, but are they smart enough o protect little customers? There’s a lot of crazy dominos in this, that can tip the wrong way.
“The numbers don’t add up:”
You’re just listing crazy predictions that these people pulled out of their @ss. These are just fantasy fart numbers. Back off a little here before you, by taking these predictions seriously, fall off the edge entirely.
There are two possible ways that AI bubble plays out:
1. Their hype about AGI is real and the country is about to lose a crap-ton (millions) of jobs. Economic devastation follows.
2. The hype is totally just smoke blown up the posteriors of investors to get them to fund the massive build out of infrastructure that normally would take years of PowerPoints and charts. Eventually, the curtain gets pulled back and the AI investment bubble pops – investment devastation follows.
I’m pretty sure the bet is for door number 2.
I use AI every day and while it’s super useful, it isn’t replacing millions of jobs anytime soon. Not even close. What it does do is provide easy cover for companies looking to downsize anyway. They can blame it on AI and not shareholders.
The good news is that the possibility of AI eating a bunch of entry level jobs has gotten the kids interested in the trades again. No AI is replacing plumbers and electricians, so in that sense the hype machine is helping us regain the losses in skilled trades from decades of deciding that every kid needed to learn to code.
Amen to the trades.
There is basically a labor shortage in every trade. If you can get more supply of trade workers, then theoretically, construction costs and other services will become more affordable. At least that is my hope.
Is there any real downside of using batteries for short-term storage of energy on a mass scale?
From what I’ve been reading so far, it’s literally a win-win scenario for *everyone* involved, especially if electricity demand keeps growing. I wonder then why more states / countries / electrical systems are not jumping onto this (or maybe they do, but just do it quietly).
The only issue I can possibly think of (and this is just my uneducated guess) would be the environmental impact of producing batteries and then recycling them when they degrade / happen to not be needed anymore, but I have no clue how massive it is.
There is no free lunch in energy. Every option is bad. What differs is the degrees of bad.
So the FRED chart showing 35% increase in kwh cost in the past 4+ years has zero/some/a lot to do with arbitraging storage?
Don’t spread homemade BS here.
1. Price arbitrage with batteries brings electricity cost DOWN because operators (often NOT utilities!!!) offer supply when prices and demand are highest, thereby helping put a lid on those price spikes.
2. The battery business is minuscule compared to the total electricity generated and the costs needed to do that. What you’re paying for is tens of billions of dollars in endless settlements from the endless wildfires, the expense of endless wildfire mitigation efforts, interest expense on the bonds the utilities issued, fuel costs (natural gas in CA), maintenance on old equipment (that is often responsible for the fires), huge capital investments in the grid, etc. The LA fires were likely caused by a powerline. There are going to be huge settlements, including with insurance companies, and rate payers are going to pay for that. You’re paying for the decommissioning costs of the San Onofre nuke that is costing billions of dollars (which leaked contaminated steam because the pipes were corroded and could not be fixed at a reasonable cost). Now there is a surge of demand from data centers. That’s the kind of stuff that drives up your bills.
3. Investor-owned utilities are monopolies and gangster operations, encouraged by the PUC. There should not be any investor-owned utilities. They should be owned by rate payers (co-ops) or municipalities.
I deleted your first comment to that effect because I didn’t want to waste my time with that BS. But now, by reposting it, you forced me to waste my time with your BS.
Wolf,
“3. Investor-owned utilities are monopolies and gangster operations, encouraged by the PUC. There should not be any investor-owned utilities. They should be owned by rate payers (co-ops) or municipalities.”
That being said, XEL is one of my largest equity holdings. It’s a “Buy and hold.” stock for me.
They just had a settlement with the Denver & Boulder area, for a 2021 fire, before Jury selection was about to begin. $640 million will go out to pay for this.
“Minneapolis-based Xcel has maintained that its equipment did not cause or contribute to the fire. It said it does not admit any fault under the settlement.”
“Investigators found that a sparking power line owned by Xcel was one of the causes of the fire that was fanned by high winds.”
“Embers from a smoldering scrap wood fire set days before on a nearby property used by a Christian religious communal group was also found to have been another cause.”
XEL is up 6.7% today.
Like I said, gangster organization, operating by screwing the rate payers.
This brings us to the small scale version of this topic, EVs. The Ford Lightning F150 could buy electricity every night and sell it back (or save home-owner electrons) every day. Teslas cannot, and will void the warranty if you hack it. (disclosure, i own a 2013 Tesla and a 2025 EV6 Kia). The fact that the government hasnt mandated manufacturers to adopt standardized electricity protocols to optimize the grid is a financial and national (grid) security failure. The article is indicative of a belated step forward using mostly “bleeding edge” technologies.
Recycling batteries is not a huge problem. There are stadiums in Japan using old Nissan Leaf batteries even if they are well below initial capacities. Their are various other solutions and commercial operations arising, its a growing business.
So batteries are taking over the job of gas powered turbines providing energy when the need is high. Not necessarily a bad thing.
Yes and no.
Right now renewables are bad for the inertia problem (like in Spain). You still need the turbines and the cost has skyrocketed on those because they are in high demand.
Bingo. As I commented above, having sufficient inertia is critical to grid reliability.