Batteries: how cheap can they get?

How dirt cheap batteries will completely transform our electricity grid, paving the way for solar and wind and replacing grid reinforcements with grid buffers.

Sodium batteries will become ridiculously cheap. That in turn will revamp our electricity grid: local demand response will be key, resilience and grid stability will improve, grid reinforcements will become less of a costly bottleneck, and solar and wind will thrive. In terms of our energy system, batteries will change everything.

How this epiphany came to me

Batteries have been central to my research for over fifteen years, but their impact continues to surprise me. For me this journey started in 2007 when I concluded that solar was underappreciated. So I gave my opinion in a newspaper, created a chart that gave me some notoriety on twitter, and took a sabbatical. During my sabbatical it dawned on my that it was not just solar and wind but that batteries were just as big a game changer because they could wean us of oil for cars. So I wrote a book about electric vehicles and started a new career (now heading a large research initiative at the Eindhoven University of Technology). But I never fully appreciated the impact of batteries on the new energy system. Until now.

In my first career I experienced the learning curves for PCs, TCP/IP, WWW, mobile phones and smart phones so applying that to batteries came natural to me. It made me conclude that EVs (and in 2017 eTrucks) where unavoidable. But what never ceases to amaze me is the amount of ways in which you can improve batteries. There are endless ways in which you can improve production methods, material composition and packaging and each time you make them lighter, longer lasting and cheaper. And every time this opens up new business cases and applications.

For almost a century we were complacently using lead acid batteries, until laptops and PCs forced us to develop something better. In 2008 batteries were over $1500/kWh but now NMC battery cells (lithium batteries with a cathode from nickel, manganese and cobalt) can be bought for less than $100/kWh and the somewhat heavier lithium iron phosphate cells that are better in every way than those first cells can be had for just $47/kWh. All these batteries can still become cheaper but sodium batteries promise to become cheaper still. And all of a sudden we will put stationary batteries everywhere.

Material costs for batteries: running the numbers

It’s hard to predict how good and cheap the batteries of tomorrow will be. But we have reason to believe they will continue to become much cheaper.

I use a method in which I look at two things:

1. The learning curve that the technology exhibits

2. The materials you need (as a reality check)

1. Calculating the learning curve

Regarding point 1. I should tell you about Wrights Law, because that is how learning curves work: for every doubling of production the price will drop by a certain percentage. I will use a wonderful open source article from Way et al on learning curves (that also shows how accelerating the energy transition saves us money). This graph is especially relevant to us:

You can ignore the colored lines and focus on the dots. The dots show how observed prices have come down (vertical axis) as experience through production increased (horizontal axis). Because both axis are logarithmic the result of Wrights Law is an almost straight line. Do you see how predictable the dots follow one another?

Now let’s calculate a learning rate based on this graph. We see that going from e.g. 10 to 1200 GWh of batteries reduced battery prices from $1200 to around $150. To calculate the learning from this the simple way you first determine how often production doubled. So let’s double 10 a few times: 10>20>40>80>160>320>640>1280. That’s seven doublings. Now the next question becomes by what percentage the price went down for every doubling. The formula for this is: starting price x (100% - learning rate) ^ 7. And the answer is that the price comes down by around 25% for every doubling.

So in order to predict the future price we first have to predict future growth. As I said in the introduction I’ve learned predictions are always too conservative so I like to look at actual developments and extrapolate that. A good source for that is the recent battery report from the IEA. So I took their data on battery production from 2015 to 2023 (page 20) and produced these graphs with them.

The dots are their data. I added the trend line by fitting an exponential growth of 59% per year. As you can see the trend line fits very well. It actually has a correlation of 99.9%!

So let’s assume this battery growth will continue for a couple of years and let’s look towards 2030. (All other forecasters assume the growth will level off, and they do this every year. For batteries it’s just as extreme as it was with solar. But that’s the topic of another blog.)

If we start with 2410 GWh in 2023 and grow with 59% per year that gives us 61.917 GWh in 2030. That would mean almost exactly 8 doublings in 2030. Each time the price would be reduced by 25%. If we started with $80/kWh in 2023 and subtracted 25% eight times in a row, the end result would be battery cells costing just $8 per kWh. That seems crazy talk. But wait! This is where method number 2. comes in.

2. Calculating the materials we need

I always thought that a battery price of less than $50 per kWh on the cell level was wishful thinking. I often calculated the cost of NMC batteries (see above). I compared the first version called NMC111 (where you have equal amounts of nickel, manganese and cobalt) with the new NMC811 batteries (where nickel is 80% of the weight and manganese and cobalt just 10%) but it didn’t matter that much for the cathode price. The raw materials for the cathode would always cost roughly $50 per kWh when energy density peaked at around 300 Wh/kg. Add to that around $10 per kWh of lithium and you can’t get lower than $60/kWh.

But then LFP batteries made a comeback and somehow CATL managed to boost them to over 200 Wh/kg. Believe me: a few years ago predicting that LFP would be that light would have made you look insane. Now a very decent 60 kWh battery (good for a range of around 350 km in frugal EV) would only weight 300 kg if made from LFP.

And then I looked at the raw materials: iron and phosphate cost less than 20 cents per kWh. The same is true for the carbon for the anode. In theory cathode and anode materials could all of a sudden become less than one dollar per kWh! All of a sudden lithium is 90% of raw material costs but these costs would be only $11/kWh or so.

That’s crazy talk of course! I mean, getting below $50/kWh is already crazy. But then we got clarity on battery prices in 2024 and it turned out LFP was already at around $50/kWh! In 2024! Not 2030! Fully installed grid batteries (so not just cells) are rumored to be offered at below $100/kWh (which the IEA expects after 2050 by the way).

But now it’s time for sodium. In LFP batteries, the raw material costs are dominated by lithium while sodium costs about 30x less. The result would be raw material costs for the cathode and anode crash again to around $1/kWh! And energy density is pretty good. Sodium is already at 160 Wh/kg, which would e.g. mean a 60 kWh battery still weights less than 400 kg on the cell level.

So extrapolating the learning curve gives us $8/kWh in 2030 while material costs could become a few dollar per kWh. And this is without even talking about e.g. Lithium Sulfur batteries that would cost just as little but would also be extremely light.

My prediction: stationary batteries in your house and everywhere will turn the stormy electricity grid into a calm swimming pool

The renewable energy community I’m part of has proven by now that we can have an affordable energy system on mostly wind and solar. But the community of practice that I’m part of in the Netherlands is faced with enormous problems due to grid congestion: over 10 000 companies cannot get the electricity they need and the number is growing fast. We are planning to spend 236 billion on the grid in the next ten years. I’m now convinced that cheap batteries (e.g. 5 hours of storage for the entire country or 7 TWh costing ‘just’ 5 billion euro or so) will replace most of these grid investments.

I predict that long before 2030 we will have long lasting batteries for less than $50/kWh everywhere. Your house will get a 20 kWh battery costing just $1000 that will earn itself back in less than three years and make sure your electricity use never peaks during the day, that your voltage hardly fluctuates, and that you will never experience a blackout again. Companies and business parks will buy larger batteries that quickly solve their grid congestion woes.

Wind and solar will get an almost constant price during the day because batteries simply absorb the excess electricity they produce when it’s a bit cheaper, to give it back when it’s a bit more expensive. So wind and solar will continue to grow quickly.

At the same time these batteries will also make sure that blackouts, voltage fluctuations and grid congestion due to peaks are things of the past on the wider grid. Everywhere the peaks and dips in the grid will be flattened by cheap batteries.

Preparing for the future: plan for an open and secure grid around local demand response that is like the Internet

In order for cheap batteries to solve our grid congestion woes and pave the way for solar and wind we need to prepare. I see many similarities with the Internet.

The global grid needs something akin to the OSI model of the Internet, that will allow everybody worldwide to develop hardware solutions that are interoperable with each other. Not just on the hardware level and in terms of the protocol (probably TCP/IP) but also in terms of how the equipment that regulates the energy flow automatically communicates with each other. It would be similar to the W3C standards that make sure you can read this in your browser, but specifically for energy. Finally I think we need to make sure that all devices use public-private key cryptography so we can be sure we get information from a trusted device that we know should exist and has known capabilities. I think this is also an area where distributed ledgers with low energy requirements (so not Proof of Work but Proof of Stake) could shine by creating an ‘trustless’ system (meaning the system justs works, also if there is no ‘trusted’ party that plays the boss).

To conclude

I think I severely underestimated how cheap and ubiquitous stationary batteries could become with the advent of modern sodium batteries. I think they will turn our grid upside down from something that is managed top-down to something that is mostly decentralized and bottom up. You will use batteries to make the electricity in your home dependable and cheaper, your neighborhood will use batteries to share local electricity (that way saving on grid costs and grid construction delays) and all in all the grid will become cheaper, more resilient, and able to deal with massive amounts of solar and wind.