I have written a fair bit recently about electric cars and electric trucks. Note that I'm not talking about autonomous driving, although some of the ADAS technologies are coming along at the same time. The assumption tends to be that the limiting factor is the cost of batteries and, perhaps, the energy density. With modest decreases in cost and modest increases in density, the price point at which the total cost of ownership of an ICE (internal combustion engine) and EV (electric vehicle) become comparable. Anywhere close and electric vehicles should take off since people like them for being fashionable, having great acceleration, quietness, and aura of greenness. Battery Derangement Syndrome In a fascinating article that I read over the weekend, but written last year, Mark Mills writes in Forbes about what he calls "battery derangement syndrome". A large part of his article is devoted to looking at subsidies and other issues. But the heart of the piece is that there won't be enough batteries. Here's the meat of it. Global lithium battery factories collectively manufacture enough capacity to store 100 billion watt-hours (Wh) of electricity annually. Sounds like a big number, but here’s the rub: the world uses over 50,000 billion Wh every day , with America alone using about 10,000 billion Wh daily. To achieve the ‘pure’ green solar-battery vision, quite obviously each home needs on average at least 12 hours of storage any given day. (We’re being generous here ignoring issues like cloudy days.) Thus, do the math on what’s required to manufacture a total of 25,000 billion watt-hours of storage systems to hold that half-day’s worth of electricity: it would take 250 years of production from all of today’s global battery factories. Yes, we could build more factories, but these are very big systems with enormous capital costs that already use astronomical quantities of materials. It is an understatement to say a 100-fold kind of manufacturing expansion for an already huge industry would be a very heavy lift. Now consider the scale needed for batteries to replace gasoline. At any given moment the fuel tanks in more than 1 billion vehicles on the world’s highways and in garages hold about 10,000 million gallons of gasoline (and diesel). That quantity of energy expressed in electrical terms totals 400,000 billion watt-hours. If we reduce this to take into account the efficiency advantage of electric motors, typically 4X better than an engine at converting stored energy into motion, then one needs only 100,000 billion Wh of batteries. That’s a quantity 5X greater than the already massive number needed for the solar vision. This. Won’t. Happen. This. Won't. Happen. That's a pretty negative conclusion. A Thought Experiment In semiconductor, we have Moore's Law, so if we measure things in transistors, we have a nice exponential trend going that makes all sorts of problems relatively easy to solve on decades-long periods. So let's not cheat, and let's just look at square meters of silicon, or wafers. Back in, say, 2000 if we looked at the requirement for silicon manufacturing in 2017 we might well see an outlook that is similar. How are we going to get there? We might find ourselves saying something very similar to Mark: Yes, we could build more fabs, but these are very big systems with enormous capital costs that already use astronomical quantities of materials. It is an understatement to say a 100-fold kind of manufacturing expansion for an already huge industry would be a very heavy lift. The switch to 300mm started in about 2000, so all the big fabs in the world were built since then, both for memory, foundry, and IDM. In fact if we just focus on 300mm, we could stand there in 1999 and say that the entire world's 300mm manufacturing are a few pilot fabs, how are we going to get to millions of wafer starts per year, look at the capital investment, the land, the size of the concrete shells. Will. Not. Happen. But it did. I know very little about what it really takes to build a state-of-the-art fab that can manufacture at, say, 50KWPM. I knew even less about what it takes to build a battery factory. But I do know that the capital costs are not going to be the limiting factor if people really want chips or batteries. Something else might be, such as a shortage of silicon. Oh, wait, it's the most abundant element on earth. Lithium doesn't seem in short supply. Apparently (which means according to Wikipedia but makes you sound more intelligent) in 2016, we used 32,000 tons of Lithium and there are reserves of 14,000,000 tons. There may well be some critical element in both semiconductor and batteries that becomes a severe limitation, but in both cases tweaks will be made to the process or chemistry. Is It Possible to Get There? It's not quite as bad as the Mark's analysis sounds. Let's look first at what the numbers really mean. Let's say that as existing cars come off the road, they are replaced with electric vehicles, and that the fleet gets replaced over almost 20 years between now and 2035. We don't have to convert all the vehicles next year. If we take the 2B vehicles by 2035 number as given (but see below) then this is around 100M vehicles per year. So each year we need ~50 times current world production. Is that a lot? Mark thinks production is already high, but that might be like saying silicon production was already high in 2007 but smartphones were still going to require a massive increase in the existing world capacity for memory and foundry. Let's look at what the predictions are for growth in place already in the short term. Just as with fabs, pretty much anything that is going to be in place by 2020 has broken ground, presumably. The above chart (from here ) shows that between 2016 and 2020 (let's say, five years) that capacity will increase nearly 6X. Let's make it 6X since that means that for the next five years, each year the world will build capacity equial to the entire 2016 capacity. If that rate carries on without any acceleration, then between now and 2035 we will increase capacity by about 20X. If we see the sort of learning and returns to scale that we have in semiconductor, then some of the factories will be bigger still, making Tesla's Gigafactory look small (in fact there are already larger factories under construction in China). If you take the other extreme, and work out the CAGR between 2016 and 2020 then capacity is roughly doubling a bit faster than every two years and so between now and 2035 it will go up by...1000X. So that's pretty silly and isn't going to happen. A plausible number is probably somewhere in the middle, with capacity increasing 50X. I pulled that number out of thin air mainly because if capacity increases 50X, and we have 20 years of production, we are at half of what we need (only half because we only get to 50X of today's production by 2035 so we are short in the early years). That doesn't guarantee it's going to happen, but it looks a lot less implausible than when it was just "we need the next millennium of production to get there." Other Aspects Everyone assumes that some level of shared ownership will happen, at least in big cities. And, in case you haven't noticed, another trend is that everyone is moving to big cities. There are over 100 cities in China with over 1M people, and you have heard of maybe a dozen of them. Same in other parts of the world. The biggest city in all the Americas is, I believe, Mexico City. There is a strong likelihood that far from there being 2B cars on the road in 2035 that there are fewer than today. If there are, they will require fewer batteries, because the current assumption is that most of the batteries are going to be sitting in garages or parking lots at work rather than being used, whereas an Uber/Lyft kind of vehicle has a much higher duty cycle. There don't seem to be any breakthroughs on the horizon that are likely to produce an order of magnitude of efficiency increase in batteries, but over the next 20 years it seems reasonable to guess an improvement of 30-50%. On the negative side, these numbers assume that when a new EV ships, it never needs its battery pack replacing. By 2035 that might be true, but it isn't today. If the life of the vehicle requires the battery pack to be replaced, say, three times, then that will require three times as many batteries. But the batteries in the old packs are not useless. They will only charge to, maybe, 70%. Using them in a car makes no sense since you can't compensate for the fact that they won't hold a full charge by adding more cells, because of space and weight. But for static applications, such as grid use with solar, they are perfect. You just need to add 30% more cells, or maybe 50% to be safe. It makes no sense to use cells that can charge to 100% for these applications, they are too valuable for that and should go in vehicles and computers where volume and weight are critical. This was something I'd never thought of until Hanho Lee of Samsung SDI pointed it out in his presentation, that batteries that are no longer good enough for cars are still good enough for a lot of static applications. For the purposes of these calculations, that probably means you can take the static applications out, at least in the long-term, and assume they are supplied with batteries coming out of cars. There is a big mismatch between solar and charging cars. In general, people want to charge their cars at night when, I expect you noticed when you were about two years old, the sun doesn't shine. There doesn't seem to be a good way to charge cars from solar, unless it is done during the day, which largely means at work (at home only for cars that stay home most of the day). It clearly makes little sense to charge up static batteries during the day, and then discharge them into the car at night. Conclusion When I started writing this post, I was convinced by Mark's argument that there battery capacity was going to be so mismatched to requirements that it was obvious that only a small fraction of the world's cars could be electric. It is not clear if the numbers work, but they are in the right range. But if shared ownership becomes a thing and we only need one-third the number of vehicles, or less, then I think it is clearly possible. So a definite...maybe. Sign up for the weekly Breakfast Bytes email:![]()
![]()
↧