Hacker News Reader

TBH, I'm not really following your logic here. While obviously a 15% reduction in cost cannot happen yearly, the Moore's law comparison is still relatively accurate.

Those Lithium Ion batteries that haven't changed in decades? They were only commercially released in 1991, less than three decades ago.

I feel like people have a hard time seeing that a curve is still exponential when it moves at ~4% a year, as batteries have for the past century or so. But that still means that the technology could effectively double in ~15 years. How many ICE applications become irrelevant at that point? And which ones become irrelevant at various points in between?

The battery revolution is in full swing.

> Those Lithium Ion batteries that haven't changed in decades?

I mean the basic chemistries haven't changed in decades. Li-ion batteries were understood and experimented on long before they became commercially viable.

The basic chemistry is the same as it was back then. There have been various improvements in packaging and chemistry/anode tweaks that have maybe resulted in a doubling of capacity in that time.

> I feel like people have a hard time seeing that a curve is still exponential when it moves at ~4% a year, as batteries have for the past century or so

It is actually a hard problem to see if something is exponential at very low growth rates. Like the economy used to grow at 5%, then 4%, then 3%, now 2% - is it really growing "exponentially" or are we trying to fit an exponential curve to something sub-exponential?

After all, if you take the measurement of anything at times T0 and T1, you can calculate the rate of growth in % terms, which by its very units implies exponential growth, but it may not be.

So if Li-ion batteries keep dropping in price by say 15% +/- 5% for the next five years, you won't really have enough info to say who was right. You can fit other curves to that data. Only when you have many years with a high enough growth rate, like Moore's law, can you really be sure exponential is the only curve that fits.

> The battery revolution is in full swing.

I don't disagree - batteries are everywhere and becoming cheaper and better in many respects. It just won't look anything like Moore's law long term.

I see your point now, and I largely agree with it. I do think that some of the same problems apply to Moore's law, though. At any given point, it could have been off by a significant margin and occasionally looked really dicey.

Basically, predictions are hard. :)

The reason the ongoing drop in battery prices matters is that they are essential for the saving the environment by getting the planet off of fossil fuels.

That being the case, the relevant question is not whether the prices will continue to drop forever at the present rate. Rather, the question is whether they will drop low enough to make the transition possible.

Well, they are already low enough to do that for some uses, and a good deal more price reduction is expected in the coming decade, so it seems the answer is yes, they will drop low enough.

There are still orders of magnitude between current batteries and the theoretical hard cap. Current batteries throw away enormous capacity due to the safety constraints in preventing thermal events. Lots of folks looking at how to solve that problem.

And what then? In term of energy contained in atoms, current batteries are many, many orders of magnitude away from the theoretical maximum energy density of matter. (And so is liquid hydrocarbon fuel.)

Batteries don't need to hit theoretical maximums, they only need to get within the ballpark of liquid fuels (maybe .25-.5x)? This will basically require the use of air in the reaction, since one of the reasons liquid fuel can store so much energy in such a compact and lightweight form is that it doesn't require storage of one of the primary reactants (oxygen).

last time i check liquid fuel has 50x the energy density of battery.

so to get to /4, you would need a 12x increase.

>since one of the reasons liquid fuel can store so much energy in such a compact and lightweight form is that it doesn't require storage of one of the primary reactants (oxygen).

the main reason is the covalent bond is stronger. using air will only double the energy density for fuel( 1x fuel instead of /2 fuel /2 oxygen.)

Hydrogen is a lot lighter than Oxygen, so you can increase the density a lot more than by a factor of 2.

Also important in some applications, such as aerospace, is that liquid fuel depletes in weight as you use it. Batteries are enough of a closed system that you still have to carry all of the weight around even when you're out of power.

> There are still orders of magnitude between current batteries and the theoretical hard cap.

Not at all. Existing cathode materials have a theoretical coulombic capacity of about 200 mAh/g (mostly less than this value, a few more have more). So for a 45g cell like an 18650, you are looking at 9,000 mAh maximum in the impossible world where your battery is 100% cathode, no anode or electrolyte. Those cells are already > 3000 mAh, so there is no way there are orders of magnitude between current batteries and the hard cap.

On the contrary, for contemplated chemistries the practical hard cap is probably less than 2x current capacity/weight values (since you need a significant amount of anode and electrolyte in practice), and almost certainly less than 3x.

> Current batteries throw away enormous capacity due to the safety constraints in preventing thermal events.

If by "current batteries" you mean current chemistries like Li-ion and existing and contemplated cathode materials, then this is not correct. The batteries have essentially the ideal material ratios within the existing manufacturing capability. That is, if you were willing to have a much less safe battery with the same materials you would gain almost no capacity. The main concession to safety is when a safer cathode material is chosen, like LiFePo over cobalt or whatever.

The main reason semiconductors scaled the way the did, was due to economies of scale. The more refined the process, the more R&D money had to be put behind the next node to make it work, the more production you needed to justify the investment. The end result being that we currently only have 3 (?) companies that are actively working on bringing out the next-gen semi-conductor node.

The same principally applies to almost anything. In fact, for typical physical products the saying is that double the production will halve the cost. This is no different. Moore's law was much the same, just that it's scaling was much faster than anything seen before, such that every 18 months we were able to halve the costs. And of course semiconductors revolutionized almost everything.

For li-ion batteries, this seems to be every 4-5 years and has been for the past 20 years or so. Rumors have it that Tesla is even beating this...

Call it "Ric's law": every 4-5 years li-ion battery prices will halve ;)

Moore's law was very different because the underlying process was being scaled. CPUs just process information, and the inherent limit on the amount of "stuff" needed to perform say an addition has an incredibly small physical limit [1] which existing chips don't approach. So through scaling the feature size of chips, CPU manufacturers were able to take the same amount of silicon, the same size wafer, and get "double" the computational power (roughly speaking) out of it. So with about the same amount of "assembly line" work you can churn out something that is twice as fast in 18 months. After a decade its 100 times as fast, but your "assembly line" looks about the same.

That's not "efficiencies of scale" - that's fundamentally making something much better due improved physics, with the same amount of work. It particular, it would apply even to "small" producers.

Now don't get me wrong, CPU manufacture was also subject to traditional efficiencies of scale: the biggest fabs got bigger, and a few large players squeezed out the rest and were able to sell more and average our their R&D costs over more sales, but that effect is small compared to the million-fold improvement in the underlying physical design.

Batteries have no such scaling. The power stored is basically related to number of lithium ions and the capacity of the anode to accommodate them all. There are some small efficiencies: you can increase efficiency from 80% to 90%, but never to 2000%. You can make materials thinner or cheaper. You can change form factors to use materials more efficiently. You can standardize on battery sizes to make more use of a single production line. You can secure long term lithium contracts and open more mines.

These are all the traditional "efficiency of scale" things and they all hit a wall pretty quickly. You could probably easily sustain a 15% reduction for a while longer, but certainly not 40 years like Moore's law.

> The same principally applies to almost anything.

It doesn't, just look around.

What else has decreased in cost by a factor of a million over the past few decades? Cars have increased in popularity by many-fold since the 40s , and after an initial period of traditional "efficiencies of scale", costs have remained relatively fixed.

Look at any random food product that suddenly increases a surge in popularity, perhaps reaching a 10x sales multiplier: final cost and production costs don't drop 10x.

If all of a sudden we start eating 2 avocados every meal they aren't going to start costing 10 cents.

> In fact, for typical physical products the saying is that double the production will halve the cost.

I can imagine this rule is true... up to a point!

That's the "traditional efficiencies of scale" at work: it's usually an S-curve [2]. If you want some custom widget, you are probably going to have to pay $100 for the first one, and $1 each or whatever for your run of 100. When you order one million, maybe it drops to 1 cent. When you order a billion, they don't cost 0.001 cents though.

You don't have to take my word for it though: just look at any two big companies, but where one is bigger than the other, and look at their costs of production. Let's say Coke sells 5x as much as Pepsi: does Pepsi cost 5x as much to produce? Not all, they are virtually identical.

Many models of vehicles sell 10x or 100x of some unpopular rivals, but the production cost is about the same.

---

[1] https://en.wikipedia.org/wiki/Landauer%27s_principle

[2] Approximately, at least - although it depends heavily on the product. For example, the initial part of the curve might not be very flat for some things (e.g., with large fixed one-off costs) - but they almost all share the "rightmost" flat part of the S-curve.

While the scalability behind silicon was perhaps easier and faster, there are similar effects at work for batteries.

Li-ion batteries are typically spooled in layers [1] (see https://en.wikipedia.org/wiki/Lithium-ion_battery#/media/Fil...). The thinner the layers, the higher the capacity as you have a smaller distance between positive and negative layers. From my discussion with battery chemists, this is what is primarily driving the annual ~15% reduction, economies of scale also obviously playing a role.

Unsure what the theoretical limit of battery size layers is, and how far off we are from that, but the thinner we get those layers, the higher capacity the batteries per weight and also presumably price.

[1] https://en.wikipedia.org/wiki/Lithium-ion_battery#Shapes

> While the scalability behind silicon was perhaps easier and faster, there are similar effects at work for batteries ... The thinner the layers, the higher the capacity as you have a smaller distance between positive and negative layers.

I'm arguing it's not at all all similar. The capacity of a lithium ion battery is fundamentally limited by the amount cathode material it contains, along with sufficient electrolyte to support it, just like any other battery. Making other materials thinner, allows you to stuff in more of this stuff, and other changes in the arrangement may make the process more efficient, but up to a limit. There is hard cap to usefulness of all of these processes, at the "theoretical efficiency" and in principle softer caps before that point long before you approach the theoretical caps.

If you have a chance, ask the battery chemists you know what the best-case storage is for a particular chemistry and cathode material, compared to what is available today. I don't think it is more than 10x aware and is probably much less.

That's very different than CPU scaling where you started out many billions of times away from the physical limits of the computational capacity of the material, and many trillions away from the theoretical physical limits of computation.

So no, thinner stuff isn't going to sustain a 15% annual increase in efficiency [1]. In fact, I don't think think it will even sustain a single 15% increase from today until the end of time.

Of course, there are many other vectors along which battery efficiency can increase when the denominator is "cost", even if the storage/size doesn't increase much. You can increase manufacturing efficiency. You can introduce new form factors. You can tweak the chemistry. You can carefully match the required discharge characteristics to the application. Outside of the battery itself, you can improve charging and discharging algorithms, you can use finer-grained control over smaller groups of batteries, you can improve thermal management. However, these are just in the range of normal industrial optimizations that apply to any product. You can replace "batteries" with "lettuce" in the above and come with a similar list.

---

[1] To be clear, there is no 15% increase in battery efficiency per year, when measured by volume, weight or other physical characteristics: that stat must involve "price". Panasonic, the best and biggest Li-on player out there, has barely budged in efficiency on their headline battery, the 18650.

> ask the battery chemists you know what the best-case storage is for a particular chemistry and cathode material

It's entirely possible that with an exponential increase in production volume (i.e., funding) we could see some new chemistries appear.

Compare to Silicon Carbide (blue LEDs) and Gallium Nitride (power transistors).

It is possible. In fact, I definitely expect to see new chemistries and anode/cathode materials appear.

The question is how much it will buy us.

Unlike other domains where theories or algorithms or medicines or whatever appear almost "out of thin air", the periodic table is limited and the mechanism of battery operation is well understood, so I think there is already a pretty good grasp on the possible materials that can be used, even in theory.

For example, Lithium is used for a reason, something like it's electron carrying capacity per unit weight. There are no other elements waiting to be discovered that are going to be better. Cathode materials are more complicated, but I don't think there is any order of magnitude improvement hidden out there.

I could certainly be wrong.

Well it happened in the solar module market[1]. And batteries are similar in many ways. They have a lot of government/policy support (rebate programs, pilot programs). Solar technology didn't change much during this period either (and in fact, the more advanced/novel technologies failed rather dramatically), a lot of the cost reduction was due to better manufacturing efficiencies and scale. I also think that because both of these technologies are older, they are less likely to be impacted by patents and monopolistic enterprises that can force prices higher. So that helps as well.

[1] https://www.greentechmedia.com/articles/read/tracking-reside...

> Well it happened in the solar module market[1].

What is "it" that happened? If you are talking about sustained exponential cost reductions over a period of decades, like Moore's law, then maybe it did happen from a very high starting cost, but it won't continue.

Any market can undergo a rapid reduction in prices over a short period of time, as a product increases in adoption by one or more orders of magnitude.

You could fit an exponential curve to that growth, but ... it will almost never be sustained. The future is what we are talking about here.

Let's take solar modules as an example. From something like [1] prices have dropped from about $4.00/watt to $0.30/watt in 10 years. So that's a price drop of about 23% per year (each year is 0.77 the price of the previous). So today, at 30 cents/watt a 300W panel (the big ones you see on houses) is about $90.

If that is really a sustained exponential drop, those same panels will be $6.75 in 10 years. Possible? Yes. That said, I don't think you can find almost anything that big and heavy almost anywhere for $6.75. Maybe bricks or gravel or something, I'm not sure.

10 years later, those same panels will cost 50 cents. Possible? No way. You aren't going to ship a 300W panel anywhere for 50 cents. You aren't going to have space in your warehouse for something that big for 50 cents. You aren't going to be able to buy even the basic raw materials like glass and aluminum for 50 cents.

There's a common theme here: for something to scale exponentially, at a fixed size, every part of the process has to scale exponentially. The cost of shipping was irrelevant when the panels were $1,000, but it becomes pretty important when you are trying to ship something that weighs 100 lbs across the world or even a country and sell it for less than a dollar. The basic raw materials are a pittance with $500 panels, but they dominate the cost if you are trying to sell them for 50 cents.

The only way this exponential scaling works is if you can make something roughly the same size exponentially more efficient. That's that happened with CPUs. The CPU itself was more or less the same size for decades. They made them in similar fabs. There were some "traditional" efficiencies of scales thrown in, like moving from 200mm wafers to 300mm wafers, but the only thing that allowed exponential improvement was that the chips themselves became a million times more efficient without increasing in size or (mostly) the consumption of any raw material.

CPUs didn't scale up a million times by building a giant fab the size of a country, and using wafers miles across and mining silicon from the moon, which would be how a traditional process would scale up: they scaled up by becoming internally a million times more efficient.

Solar cells don't have that type of scaling available to them. There are hard limits on the efficiency (which is already above 20%) based on the physics involved - but even without any reference to cell physics there is a hard limit at 100% which means the efficiency upside is at most a one time 4x gain from here.

---

[1] https://commons.wikimedia.org/wiki/File:Price_history_of_sil...

> as a product increases in adoption by one or more orders of magnitude.

/aside With world human population forecast to decline, efficiences and economies of scale will be less in the future.

The real reason we need strong AI - as consumers.

Oh come on! No one here, other than yourself, has tried to claim that exponential cost reductions will continue indefinitely. But it is a fact, that solar experienced that type of cost reduction for a period of 2-3 decades. It's very unlikely to continue on that trajectory, but it doesn't really matter because solar is now one of the cheapest forms of generation on the market at this point.

Batteries are in a similar position as solar was a decade or two ago. The grid operators desperately need storage capabilities, and li-ion is one of the more promising solutions, but it's still too expensive. But another decade of cost reductions can change that. No one knows for certain if it's possible, but like I said in my original post, no dramatic technology changes are really needed, just better logistics and scaling could be enough.

> Oh come on! No one here, other than yourself, has tried to claim that exponential cost reductions will continue indefinitely.

The OP did, which is what I was responding to. He mentioned the current exponential decrease in battery costs, and said "I see no reason for that not to continue".

They then drew a direct analogy to Moore's law to explain how such exponential gains can be sustained. I disagree.

Of course, even the GP probably agrees they won't continue "indefinitely" but my claim is that they can't even continue for very long (the higher the rate of cost reduction, the shorter it can be sustained).

> But it is a fact, that solar experienced that type of cost reduction for a period of 2-3 decades.

Kind of, there were certainly long periods of stagnation in solar panel costs, but also periods of large drops. It also depends how your pick your starting point.

One has to ask (as you did) whether today's batteries are more like the low-volume niche product of 70s solar panels, or more like today's solar panels. I'm willing to bet the latter, both because batteries have seen substantial investment to date (I'd wager much more than solar panels), and because the existing scale of battery manufacturing is already massive (c.f., gigafactory). Furthermore, there are no apparent revolutions in battery technology on the horizon.

So rather than being at the start of a precipitous drop in battery prices, I think we are somewhere near the middle, and the drops from here will mostly slow down not speed up.

That is a fair point re existing battery scale. However I would counter, that to date, the vast majority of that development and investment has gone towards batteries for small devices (phones and laptops, basically). Compared to utility scale battery packs (and to some degree electric vehicle packs), I imagine the manufacturing challenges are very different. For comparison, the battery in my phone is about 13Wh, whereas the Tesla battery in question is about 130MWh, 8 orders of magnitude bigger. And we have only very recently started building systems like that. While a 10x reduction in cost certainly might be a stretch, I wouldn't be surprised to see at least a 50% reduction in the next 5-10 years.

You might be right about these large scale systems.

Are they using significantly different technologies for the individual cells? I always found it weird that the large battery in a Tesla, for example, was just made up of thousands of 18650 cells that are probably smaller than the battery in your phone. Is the Powerwall is the same?

Does it get any different at larger sizes? Obviously stuff must be different outside of the cells, even if the cells are the same.

This is just based on my memory, but I think they use different battery chemistries but use the same cell size (18650). I also thought it was weird that they were using 18650's when I first found out about it, but I think the need for thermal management actually makes the smaller cell sizes ideal.

Of course this is also why I think there are some cost reductions on the horizon, as current EV and grid tied batteries are basically just a bunch of laptop batteries thrown into a fancy box.

> but I think they use different battery chemistries but use the same cell size (18650)

I also read something like that. I suspect it is more of a tweak than anything major. No doubt it is the same basic chemistry but perhaps with minor tweaks to the various ratios and material thicknesses and so on. The Tesla batteries are a single-application cell, so you know exactly the maximum discharge current, the maximum charge current, the temperature parameters, and so on, whereas a generic 18650 has to balance those for a "typical application" or whatever. See for example the Samsung high-discharge INR18650 cells which trade off capacity for higher max discharge current.

OTOH the other hand I also heard the early Teslas were using exactly the Panasonic 18650B cell. They could both be true: maybe it was a stock cell in the early days and they tweaked the formula later in concert with Panasonic as their volumes rose. There was no big evident spike in capacity vs weight though...