An apocryphal phrase of note, in regard to Electric Vehicles, is "It's the Battery, Stupid." It no longer seems appropriate given that almost everyone who's watching the space knows batteries play a critical role in future potential. I've gone for a slight modification, perhaps better capturing the arguments that continue to rage around EV potential: 'It's the stupid battery!'
The cause of so many debates and so much doubt, critical to so many EV hopes. Taylor Anderson recently mentioned:
...the battery concern is still very much valid - for now.
This is an entirely reasonable statement, but looking deeper it is surprising to see the rate at which battery concerns, and their underlying assumptions, are evolving. I can only assume that these shifting predictions are not unrelated to the increasing EV investments of auto manufacturers including GM, Honda (HMC), Nissan (NSANY.OB), Toyota (TM), VW (VLKAY.PK), Ford (F), Daimler (DDAIF.PK), BMW, Mitsubishi (MSBHY.PK), BYD, and Renault (RNSDF.PK). Battery cost trends are of critical importance to these players' EV development plans, which means that they're also of critical importance to the strategies of investors in the EV space.
In Part 3A of my EV Myths and Realities series (which I'd suggest reading first if you haven't already), I looked at the total cost of ownership of the Nissan Leaf and Tesla (TSLA) Model S under today's commercial environment, and followed this with a quick feasibility check on manufacturers' ability to cut EV prices by more than the existing $7,500 subsidy by 2020.
Now I follow up by going into detail on batteries, reviewing the evolution in 'forecasts' over the last few years and using this as a credibility check for the assumptions I made last time. It will also form the basis of another follow-up article (Part 3C - yes, it's an asymptotic series) in which I'll look at the broader multi-variable economics of EVs drawing from the data presented here (though I'll close this article with a first sample thereof).
Also see: Part 1, which demonstrated that battery reserves and production allow tens of millions of electric vehicles per year with realistic ramp rates; and Part 2, which demonstrated that EVs are clean compared to the status quo, even after considering grid emissions.
Battery Price Trends
At the end of this article's predecessor, I presented the following table illustrating the potential for cost reduction in EVs as battery prices come down. The goal was to chop at least $7,500 and have EVs stand on their own feet without today's subsidies.
|Battery Size (kWh)||24||40||60||85|
|Battery Price Today||$12,000||$18,000||$25,500||$32,000|
|Battery Price 2020||$5,280||$8,400||$12,000||$15,200|
My assumptions above require the smaller pack in the Nissan Leaf to achieve a $/kWh below $220 by 2020, while the larger pack in the Tesla Model S40/S60/S85 must be below $210/$200/$190 respectively (actually, if we only need to save an amount equal to the subsidy, the targets are just $263/$300/$306 for the Model S range...but I predicted and expect much larger cost reductions - simultaneously improving Tesla's margins and competitive position).
I wouldn't expect you to take my word for it though - let's see what others think.
Above you see a collection of battery price estimates, in $/kWh, spanning the next decade or so. I created it by drawing on reports from several sources over the last three years. For the most part I cite cell costs, though it's worth noting that the indication for Tesla today is the fully integrated pack cost to the end customer. It's also worth noting that the cell costs comprise the vast majority of the midterm pack costs, but I'll go into that in more detail in a moment.
Comparing this chart to the table above, it seems that the estimates I used for today's prices are, if anything, a little on the low side. This is good - it means today's EVs are starting from a higher point, and the necessary reduction will therefore be easier to come by. What about 2020 though? The estimates are all over the show. This is where it gets funny.
Check the axis labels there - this not a trend in battery price reduction, rather a trend in the 2020 battery price prediction reduction. The reason it's easy to confuse the two is that they're both falling at about the same rate. In 2009, Deutsche Bank thought the 2020 price would be $350/kWh, which would make my assumptions much too low. But almost immediately they reduced their estimate 30%. Now McKinsey say $200. Roland Berger just came in with $180. The DOE, whose 2010 estimate of 2012 prices was less grossly excessive than most, think it will be $150.
When the DOE report came out, John Petersen referred to it as 'specious', but as time progresses it would seem the DOE is closer to the money than anyone else. I haven't even plotted the numbers from Lux Research for $/kWh 'usable' as they are not only an outlier, but also about 50% higher than costs today. What's doubly strange is that an annual improvement of around 8% is now widely acknowledged, even by many skeptics. 8% per annum between 2012 and 2020 would bring $400/kWh down to $200/kWh. I have no explanation for this inconsistency in mindset, other than simple denial.
Those who really, deeply, understand what the cost drivers in battery production are (e.g. the DOE) seem to be doing much better in prediction than those who call battery suppliers and try to convince them to indirectly reveal their cost structures to their customers (certain market research agencies). The common thread, however, is that as time goes by everyone is continually revising their estimates further downwards. In 2012 the 2020 estimate is around $200. In 2015 I won't be at all surprised if it's $150. This is just from already visible incremental improvement opportunities - a disruptive battery technology would lead to significant further reductions. We won't know what the 2020 price will really be until 2020 (few predicted the massive collapse in PV prices as supply overtook demand), but we can be pretty confident it won't be higher than today's estimates.
It's good that Lux try to factor in usable capacity, even though they get it wrong, but they go on to compound their error by horrendously overestimating pack cost. This brings me to the next topic (which I'm sure several more EV-skeptical readers are already chafing at the bit over)...
What about the BMS and TMS?
Indeed, what about the BMS and TMS? These, in particular the BMS, hold near legendary status in the skeptic community. Some of the numbers that are thrown around in regard to pack cost vs cell cost are totally detached from any conceivable volume production reality - I've adopted the internet meme-of-the-moment to illustrate the various perspectives.
In a midsized Electric Vehicle, a BMS is a half a handful of active and passive components totaling around $50 - $100 BOM cost, and embodying multiple millions of dollars of development work. The nice thing about that development work is that, once you've got it, you don't have to pay anything extra to use it in 10,000 cars or 1,000,000, and the nice thing about the fixed cost is that the hardware elements are only getting cheaper. When Maxim released their 11068 integrated BMS chip they said that it would reduce the cost of the typical BMS from $250 to $50. The development of cell integrated BMS which don't even require additional wiring, such as that of BatteryMan, will bring the cost lower still.
The TMS is another smart-design-cheap-hardware element that sometimes has its fundamental costs talked up as though it was made from pure palladium. When someone says "TMS", think "Heat Exchanger"... and when you're thinking "Heat Exchanger" do not think "peltier effect heat pump powered by flux capacitors", but rather "Radiator a bit like in a car, but only needing to handle around 5% of the power, coupled to a small air-con unit and a smaller heater". This is why you don't see EVs with huge air intakes - managing battery heat is a tricky topic to get right, but once you've got it right the incremental cost of getting it right many times is quite low as the heating and cooling power needed is tiny compared to a normal car. In fact, with the Leaf, Nissan didn't bother with a TMS at all (though given the issues they seem to be encountering in Arizona, I'm glad Tesla were more cautious). The highly robust Toshiba LTO cells used in the Honda Fit EV have been extensively tested under very wide temperature ranges and demonstrated excellent lifetime and performance without active cooling.
Finally, the 'pack' itself is basically just a steel or aluminum box/frame - Tesla made it a semi-structural part of the vehicle. With aluminum sheets only a few dollars per kg, the volume production cost is pretty trivial, especially once it's basically a part of the chassis (which in its entirety costs only $1k - $2k excluding closures).
A Duke University masters project in 2009 captured the structure of these additional cost elements surprisingly well - I hope the student in question is hired by a market research outfit (I'm forgiving the cell cost errors in view of the 2009 genesis of the work).
In summary - battery costs now and into the future are all about the cells; 'and the cell costs, they are declining'.
Battery capacity as a physical constraint is less relevant these days, since it's already possible (as Tesla has demonstrated) to produce a 300Mile Range full EV with existing batteries. I'm not convinced that 300Mile pure EVs are the mainstream technology of the future with even 2020 cost structures (I'd rather expect 100Mile EVs with microdiesel range extenders or similar...but new batteries could change that, and it's application dependent anyway), but the key is that if they ARE then from an energy density perspective they're already possible, and in fact driving around today.
The reason capacity improvements are interesting is that:
- They help to drive cost improvements
- There is a myth that progress is incredibly slow and has basically stalled, and such myths must be busted.
The following chart shows 18650 energy density improvements between 1992 and 2011; the majority of the data points come from this interesting article, with the 2011 data point drawn directly from Panasonic's (PC) NCR18650A Datasheet.
The average improvement over this period is 6.3%, though note that this doesn't capture improvements in battery voltage, which also positively impacts energy density. The prediction for 2020 is represented as a current density improvement, but actually combines the voltage and current density improvements. It's created by comparing the energy density of Panasonic's commercially available NCR18650A cells (245Wh/kg) with those of Envia's externally validated prototypes (400Wh/kg). Achieving this by 2020 would mean an annual improvement of, again, 6.3%. Slow and steady. Of course, Envia claims it will be in production by 2015, which would mean an acceleration in rate of improvement - by no means out of the question given the recent levels of investment.
Lithium Ion batteries are subject to limitations in a lifetime from a combination of cumulative cycle limits and calendar life limits. To further complicate the assessment, these two variables are not independent.
The real nitty-gritty of this topic is over my head, and I don't want to lead you astray. What I can say though is that there are many opportunities to improve the lifetime available, and commercially available cells already deliver far superior performance to what's widely accepted.
Toshiba's (TOSBF.PK) SCIB cells, used in the Honda Fit EV, offer minimal capacity loss after 6,000 cycles; Thermal stability and expected calendar life are also very high. LFP cells from several manufacturers are expected to deliver cycle lives of 3,000+. All cell types are seeing ongoing improvement.
Calendar life is even less my area of expertise, but the 'norm' for well managed cells today (i.e. not your cell phone left on the dashboard in the sun) is around 10 years. Some of the LTO based chemistries are already estimated to deliver approximately twice this calendar life and, as for cycle of life, calendar life improvements are ongoing.
Putting it Together
Lithium ion batteries are getting steadily cheaper per unit energy. Their cycle of life is steadily increasing. Their calendar life is steadily increasing. All great news. But what's the numerical significance? I'm going to save the bulk of that topic for my next article, but since that may be a month in coming I want to wrap up the Tesla S and Leaf-vs-Versa topics first.
In the previous article I illustrated that (even ignoring the likely significant maintenance cost savings) the cost of a Leaf vs a Versa, under various assumptions (including $4/Gallon and 12,500Miles/annum - check the last article for details) had the Leaf at an NPV disadvantage of around $1,300 after an 8 year period (for which the battery is warranted). In addition to maintenance, however, I ignored residual value; in the case of the Leaf heavily linked to the state of the battery, and replacement cost thereof.
Hence, the analysis was incomplete. Our highly numerate EV buyer will certainly be aware of the residual value. They might want to continue owning the car; they might want to sell it. Let's assume the 8 year old Leaf and Versa are otherwise similar (ignoring the significantly higher wear on the Versa engine... unlikely it will be still be getting 31MPG) and instead create a $/gallon-equivalent metric for the Leaf owner that combines their aggregate costs of battery depreciation and electricity. The behind-the-scenes assumptions here are largely the same for the first part of the lifecycle analysis, except that I have reduced the discount rate to 4% on the advice of an economist who pointed out just how unlikely an 8% risk-free after tax return was these days.
On the x-axis is the pack $/kWh replacement cost. If you want to feel extra safe, you can consider the potential for pack reuse and cell repurposing/recycling - neither of which are considered in my estimates, but both of which are likely and will reduce the effective cost still further. The y-axis is the useful life of the pack in years (I've also noted cycles for the Leafs 24kWh pack at 12,500mi/annum). The contours are $/gallon equivalent - a little hard to read, but start at $1.50 in the top left, all the way to $8 in the bottom right. $/kWh are still at 0.12.
I've marked the likely areas in which pack cost will fall both today and in the future. With the worst case assumptions, someone replacing a pack and facing 2012 costs would be paying the equivalent of $4.70/gallon to 'fuel' their EV. Not exactly a disaster, but not ideal. This is barely even realistic today though - even with existing batteries 10 years is fairly pessimistic, yet 10 years would see the future Leaf owner comfortably straddling the $4/gallon mark.
Prices in 2020 won't be those of 2012 though. A Leaf owner replacing their battery pack at a net cost of $200/kWh and expecting a (now very pessimistic lifetime) of 12 years would face the equivalent of $2.50/gallon. Optimistic estimates would take that value down to around $2/gallon. So considering resale value, the prospective buyer in 2020 would be presented with two choices. They both look similar, and have similar performance.
One, the Versa, is facing increasing maintenance costs (engine, gearbox, exhaust) and runs on oil, at a cost of around $4/gallon (barring price increases in the interim!).
The other, the Leaf, has a very low maintenance drive train and the effective fuel+battery cost is $2.50/gallon. It's also quiet, clean, and quick - and by 2020 charging infrastructure will be ubiquitous.
Which one would you pay more for?
Here's the same chart for the S40 vs the BMW 328i.
Here we see the costs that come with the convenience, speed, range, and size of the Model S. Still, however, with realistic 2020 battery costs the comparison will be between the (increasingly maintenance-hungry) BMW running $4/gallon+ gas, and the Tesla S40 requiring an equivalent of $2.50-$3/gallon.
The world isn't static though, and batteries aren't the only things under improvement - modern ICE engines, especially direct injection diesels, are increasingly efficient. Next time I'll try to close off this increasingly long series of articles with a broader comparison of future vehicle options considering various drive train types - some interesting implications. Regardless, I hope that this piece (taken together with the previous ones) at least illustrates that EVs are already highly attractive by most metrics. While I won't be surprised if range extenders play a part, the vehicles they're integrated into will be Electric Vehicles none-the-less. Performant, economic, sustainable - their future is bright.