Boeing, COMAC open facility to transform gutter oil into aviation biofuel
Ballard and Van Hool launch European service center to support fuel cell buses

CMU/MIT study finds large-scale battery manufacturing will do little to reduce unit costs past a 200-300 MWh annual production level

A new techno-economic analysis by researchers at Carnegie Mellon University (CMU) and MIT has found that economies of scale for manufacturing current Li-ion batteries for light-duty EV applications (in this case, prismatic pouch NMC333-G batteries and packs) are reached quickly at around 200-300 MWh annual production. Increased volume beyond that does little to reduce unit costs, except potentially indirectly through factors such as experience, learning, and innovation, they determined.

That’s comparable to the amount of batteries produced for the Nissan Leaf or the Chevy Volt last year,” said CMU’s Dr. Jeremy Michalek, the corresponding author of a paper on the research published in the Journal of Power Sources. “Past this point, higher volume alone won’t do much to cut cost. Battery cost is the single largest economic barrier for mainstream adoption of electric vehicles, and large factories alone aren’t likely to solve the battery cost problem.

The cost of Li-ion batteries is arguably the single largest barrier to mainstream adoption of EVs. Thus, battery cost is a key factor in addressing oil dependency, global warming, and air pollution in the United States. We investigate the role of battery design variables on the cost and performance of Li-ion batteries by first characterizing the tradeoffs in battery design and subsequently using this knowledge to optimize and assess technical and economic implications.

Existing studies on the economics, adoption potential, and emissions reduction potential of EVs typically treat Li-ion batteries as though they are all the same, with a single estimate of cost per kWh of storage. In practice, Li-ion technology encompasses a wide range of alternative chemistries (e.g.: LiMn2O4, LiFePO4, LiNi0.33Mn0.33Co0.33, etc.), electrode designs (e.g.: thin/thick), packaging alternatives (prismatic, pouch, cylindrical), and capacities (size, number of electrode layers, etc.) of the individual cells that make up the pack as well as differences in pack configuration, thermal management, and control electronics. Each of the potential combinations of these alternatives has different performance, cost, weight, volumetric, thermal, and degradation characteristics that interact with the constraints and needs in the design of a vehicle powertrain system. For example, short-range PHEVs require cells with higher power-to-energy ratios because they have less battery capacity over which to distribute peak power demands. Thinner electrodes deliver higher power per unit capacity, but they also require more of the inactive materials, and this has implications for cost, volume, weight, and life.

… We aim to produce a transparent, bottom-up assessment that explicitly accounts for the battery design changes needed to meet requirements for various EV applications at minimum cost while identifying key factors and characterizing uncertainty.

—Sakti et al.

The team built an optimization model to identify the least-cost battery and pack design that satisfies energy and power requirements representative of PHEV10 (16 km AER), PHEV30 (48 km AER), PHEV60 (96 km AER), and BEV200 (320 km AER) vehicles, where the subscript indicates the vehicle’s all-electric range (AER) in miles.

They calculated cell capacities for different designs, then pack energy using capacity times average cell voltage as estimated by using Battery Design Studio (BDS) simulation software. BDS was also used to simulate the hybrid pulse power characterization (HPPC) test—defined by the United States Advanced Battery Consortium (USABC)—on a set of 48 virtual LiNi0.33Mn0.33Co0.33/Graphite (NMC333-G) cells varied over a full factorial of selected electrode thickness and cell capacity levels. The single side electrode coating thickness was varied from 25 mm to 200 mm in intervals of 25 mm and the cell capacity was varied from 10 Ah up to 60 Ah in 10 Ah intervals.

To compute cost, they modeled the process of manufacturing the Li-ion battery pack using a process-based cost model (PBCM) to simulate production operations in a manufacturing plant, using data at the individual machine level for each of the process steps. They adopted information on equipment cost and processing rates for most of the many process steps from Argonne National Laboratory’s Li-ion battery cost and performance model, BatPaC.

Process steps involved in Li-ion battery manufacturing (adopted from ANL’s BatPaC). Process steps with sufficient data on machine processing rates are assumed to be dedicated, and have been shown in green boxes with italicized fonts. Blue boxes show the process steps for which sufficient information was not available, and they were therefore considered undedicated. Credit: Sakti et al. Click to enlarge.

They assumed a yield of 100% for all process steps except Cell Stacking (#7 in the diagram above), in which defects may be incorporated as the bi-cell layers are stacked on top of one another. Their base assumption for cell stacking yield was 95%.

Comparing the cost of a battery and pack design sized for a PHEV20, for example, using BatPaC vs. the PBCM with base case, optimistic, and pessimistic assumptions found results from the base case PBCM comparable to BatPaC at a volume of 100,000 packs, the level at which BatPaC is calibrated.

Total breakdown of the sample PHEV20 (36 km AER) battery pack at 20,000 packs per year, with a further breakdown of the material cost. Sakti et al. Click to enlarge.

The PBCM results are lower cost than BatPaC estimates at low production volume and comparable cost at higher volume.

Results from the PBCM suggest that economies of scale are reached at about 200-300 MWh of battery capacity production—much sooner than suggested by the BatPaC model. This early attainment of economies of scale is observed across a wide range of battery pack specifications.

—Sakti et al.

They found that the specific cost of the optimal design decreases with the increasing electric range—from $545 kWh-1 for the PHEV10 (16 km AER) to $230 kWh-1 for the BEV200 (320 km AER). Part of this cost decrease is due to increased cathode thickness for larger AER applications that have lower power requirements per unit energy.

However, they noted, the PHEV30 (48 km AER) design is constrained by the upper bound for cathode thickness, and larger packs cannot take advantage of thicker electrodes. Additional reductions in specific cost for the PHEV60 (96 km AER) and BEV200 (320 km AER) result primarily from spreading some of the packaging, battery management and thermal control costs over a larger pack energy.

In general, results suggest that the lowest cost designs use the thickest electrode coatings that satisfy the power requirements and large cell capacity and a preference for more cells per module instead of more modules per pack (because additional modules incur more module regulation costs, primarily from the module state-of-charge regulators). There is a marginal cost difference between achieving an active material target via increasing cathode thickness vs. increasing the number of bi-cell layers.

—Sakti et al.

The results showed that pack-level specific cost ($ kWh-1) for these designs varies almost linearly with power-to-energy ratio.

Specific costs are pessimistically as high as $680 kWh-1 for the PHEV10 reducing to $330 kWh-1 for a BEV200 (320 km AER) or optimistically as high as $480 kWh-1 for the PHEV10 (16 km AER) reducing to $190 kWh-1 for the BEV200 (320 km AER). Overall, the effect of pack size on specific cost is larger than the uncertainty represented by our optimistic and pessimistic cases.

… The reduced specific cost for larger packs is due to the ability to use thicker electrodes for applications with larger energy requirements (larger AER), and new technology enabling cathode thickness values up to 200 mm could further decrease costs of larger packs by up to 8%.

—Sakti et al.

The results of the study raise questions about whether increasing vehicle sales is the best way to continue to spend limited resources—as opposed to, for example, more research on battery technology,said co-author Dr. Jay Whitacre. Whitacre pointed to the study’s finding that a way to make batteries with thicker electrodes could lower the cost of long-range electric vehicle batteries by up to 8%, and noted that increasing production beyond current levels may only cut costs by less than 3%.

If economies of scale in battery production are achieved at relatively low volume, as our process-based cost model suggests, then policies attempting to achieve reduced EV costs via subsidies for EV sales may have limited effects on battery costs beyond levels of ~200-300 MWh per year. … Additionally, our results emphasize that different cell and pack designs are appropriate for different applications. Customizing battery designs for each application may save costs (assuming adequate production volume), and policymakers should be careful not to assume that achievement of cost targets for one application necessarily enables cost targets to be achieved for other applications.

Further, any cost estimate for automotive Li-ion batteries should be viewed in the context of the application (AER), the scope (cell vs. pack level costs), and the unit (cost per nameplate capacity vs. cost per usable capacity). Comparing cost estimates may be misleading if differences in context are not accounted for.

—Sakti et al.

NMC-G. The study only considered the popular NMC-G chemistry, which is used either solely or in combination with other active material chemistries in the Ford C-Max Energi, BMW ActiveE, BMW i3, BMW i8, Mitsubishi i-MiEV, Volvo C30 EV, Honda Fit EV and Honda Accord, according to the team.

Nor did the study explore the 18650 cell format that Tesla uses, opting instead for the prismatic format that everyone else is using, Michalek said. Although manufacturing cylindrical cells involves a few different steps, Michalek said he would expect economies of scale for these cells to be comparable. Indeed, with manufacturing for that format having already been cost-minimized for decades, there is likely less room for improvement in that format, he suggested.

Although high volume alone may not provide the cost savings Tesla is looking for from its Gigafactory, the company may get additional savings from other factors such as supply chain integration, he suggested.


  • Apurba Sakti, Jeremy J. Michalek, Erica R.H. Fuchs, Jay F. Whitacre (2015) “A techno-economic analysis and optimization of Li-ion batteries for light-duty passenger vehicle electrification,” Journal of Power Sources, Volume 273, Pages 966-980 doi: 10.1016/j.jpowsour.2014.09.078


Account Deleted

So the million dollar question is who should we trust the most? Panasonic and Tesla saying they can reduce production cost of battery packs from current state of the art production by 30% minimum with their 50GWh factory or a few academics who never got their hands dirty with actual production?

I think Musk said the savings would primarily be from vertical integration and the fact that when a factory buys 10 or 30 times more machines, raw materials and semi-manufactories than other battery factories out there they get significant discounts. Something the researchers did not analyze.


Tesla claimed that they would be able to get around 30% cost savings in a few years, but AFAIK did not specify how much of this was from high volume manufacture and how much to other factors.

The researchers also specifically say:

'Although manufacturing cylindrical cells involves a few different steps, Michalek said he would expect economies of scale for these cells to be comparable. Indeed, with manufacturing for that format having already been cost-minimized for decades, there is likely less room for improvement in that format, he suggested.

Although high volume alone may not provide the cost savings Tesla is looking for from its Gigafactory, the company may get additional savings from other factors such as supply chain integration, he suggested. '

However this is primarily a study into how cost scales with production levels in batteries, and should not be taken as some sort of advocacy exercise either for or against a particular company.

If it has to be related to Tesla, then I would see it as essentially neutral, as although it sees limited effects from pure scaling, clearly the effect of using a bigger battery pack is very favourable, and scales much more sharply.

Since big low power batteries have always been Tesla's approach I cannot conceive how this can possibly be taken as any sort of hatchet job on them.


@Henrik: One of the "academics who never got their hands dirty with actual production" on this study invented a new stationary battery technology, built a company from the ground up to produce that technology, and now employs several hundred people. He also helped design the batteries for the Mars Rover.


I said 4 or 5 plants would be better for risk reduction. They could have put one plant in each of the competing states. We will see how this "gigafactory" works out, it could be Musk's Waterloo.


I vote for BYD and Foxconn to to it (30+% cost reduction) first followed by LG, Sanyo, Panasonic etc.

Account Deleted

The CEO of a 200 man strong company doing specialist jobs for Nasa and other high tech companies has not anything to do with mass production. What is the most units of anything they have done. 10, 100, or a 1000? Musk and Panasonics 6500 man strong battery factory is going to make billions of 18650 cells per year. The Mars rover battery is nuclear decay based as a recall, not chemistry based. I am sure it is one of a kind and I do have a lot of respect for people doing basic research but it does not make them experts in mass production and its costs.

My problem with this study is that it will be used by some people in large automakers to argue that the Tesla way is wrong and should not be copied as it is a dead end because batteries will always be too costly to be relevant for mass market vehicle production. That would be a pity IMO.


Why don't they define a 22750 (or whatever) battery with twice the capacity (and volume), but essentially the same class of battery. Since they are making so many, they can define their own standards.
I understand Musk went for smaller batteries as they have a larger surface area and are easier to cool. I wonder would this still be the case if you doubled the volume - 1/2 as many batteries.

Account Deleted

One of Musk's main arguments for using small cells is that they contain less energy per cell relative to a large cell and the net effect is that Tesla can pack the cells more efficient will less need for heavy fire protection material between the cells in order to prevent a thermal runaway in one cell from propagating into other cells. Tesla ends up with a battery pack that is as safe as it should be but also has much higher Wh/kg at the pack level. That is another reason Tesla is the only carmaker with well over 200 miles range for their BEVs. But you are quite right that Tesla can have a cell made in any size they desire as it will become the most mass produced cell in the world. I would not be surprised if the next version of Tesla's battery packs will use something else than 18650 if that could improve on the wh/kg at the pack level. Maybe something even smaller than 18650.


Yes, he has said that, but he picked a pre-existing cell size which is not necessarily optimum, just the best available.

Now that he is the 500 pound gorilla, he can define any size he wants, including making them larger in any(or all) dimension.
If he increases each dimension by a factor if 1.3, he doubles the volume.


What is the ideal cell size and shape to:

1. maximize energy storage density?
2. reduce installed cost per kWh capacity?
3. allow less cooling?
4. allow quicker discharge/charge?
5. reduce total weight?
6. maximize longetivity?


A VW exec was interviewed, he believes that there will be 300 mile EVs by 2020, they need a 2x improvement in density and cost for that to happen, IMO.

Pouch cells probably cost less to manufacture, one person said they have 20 AH, 40 AH, they are working on a 60 AH. You can pack and stack pouch cells, look at the LEAF pack to see an example.


The author suggests that government incentives to consumers may not be as effective as investing that money in battery research. This seems to reflect the same old mindset that consumers will only buy BEVs if their costs are comparable to ICE. He ignores the reality that people routinely spend lots more money for a car that makes them feel good (aesthetics, performance, technology, etc.). Once people experience BEVs, they buy them for a much better driving experience (and other non-quantitative factors).

He also ignores the fact that you can already buy a BEV for not a lot of money (Mitsubishi iMiev, Nissan LEAF, etc.). He also implies that BEVs should have a 200-mile range, and that battery cost needs to be significantly lower for that to happen. The fact is that most people drive less than 50 miles per day, and that's not going to change in the future. BEVs do not have to satisfy 99% of the population's transportation requirements.

The study apparently is based on current generation battery technology, predicting several years ahead to the economics of Tesla's mega-factory. There are any number of battery chemistries in the pipeline, with production techniques that have not yet been developed. Why would this study even be relevant? I have more faith in the backers of the mega-factory, given that they're putting their money where there mouth is.


This is a "near term" study, as in studying battery packs that are being produced right now, for a form factor and chemistry that Tesla doesn't use.

USC Vitterbi has already demonstrated a viable (as in it can be cheaply manufactured) battery with 2.5x the energy density, and 600 cycles with a 20% drop (in the context of a 200 mile pack that's 120k miles - good enough)

What do you know, Tesla has an intern program called ViterbiConnect with that school.

I think we will be seeing the gigafactory making these cells by 2022 or so - so only 2 years after it is fully online and complete.

So this analysis has very little to offer on clarity in Tesla's business model. Worse that very little, it tends to deceive if you're not very well read up.


I would try to make it as easy as possible for people to have Evs as second (or first) cars, then you don't need great range, you can use an ICE on the days when you need it.

This could be done by reducing taxes on EVs, by making ti no more expensive to insure an ICE+EV as an ICE only (or only marginally more expensive).

It could be done by the car companies providing rental vouchers to EV buyers, it could be done by "someone" providing an EV-ICE swap service.

It could be done by enabling people to drive rentals on their own insurance, and by having a valet car swap service.

The EV people have to accept that the range is not enough for all drives, and figure out ways to make it easier for people to do the occasional long trip, rather than just pretending that the problem does not exist.

There are lots of non-technical ways to increase the usefulness of EVs (until we get the battery thing licked).

You should tackle the simple thing first, then the harder things (or both simultaneously).


You have defined 6 valid metrics for battery performance. Adding cost makes it 7.
So, as you point out, it is a non trivial problem: how do you compare cost vs energy vs power vs longevity.

So you would have to build a model, and initialise it with current 18650 cells and then start adding 5 or 10% to each dimension and see where it leads you.

Some metrics will be continuous: a 21 kwH battery is better than a 20 kwH one (by 5% !). Others will be binary or have a threshold - a 16 year battery might be no better than a 15 year one.
Once you get above 12 years for the life, nobody might care, however an 8 year one might not be acceptable.

Anyway, build a model (of the whole car) and see what happens as you increase the battery cell size.


Cost is in No. 2.

There are 101 various new battery technologies in development. A few will be mass produced by 2020 or so.

Twice the energy density and half the cost should be reached by 2020 or so.

Post 2020 EVs will certainly be more performant and affordable.


Whitacre's battery is a water based low cost materials battery for stationary storage. His thinking is that low cost is most important to stationary as one is competing not against gasoline, but rather low cost electricity. This is likely correct and a good approach to solar and wind storage. Lithium ion in it's current incarnations will be hard pressed to meet the low cost requirement for solar and wind storage.

However, certain chemistries (one mentioned in another comment) have the potential to alter that thruth. Si anode/sulfur cathode batteries if improved upon enough will be extremely low cost. Partly because the materials themselves are cheap and abundant, but also because the energy density is higher and fewer batteries will be needed for the same range in an EV. The thing is that Whitacre's battery will never make the transition to vehicles technologies, since it will never have enough energy density. However, the Si/S batteries could make the cross to solar and wind if higher cycle numbers can be achieved. In vehicles the switch is from relatively high cost gasoline to low cost electricity the financial motives both for the customer and the manufacturer makes it far more likely that mass manufacturing will be put in place for vehicle batteries in advance of anything specifically for grid applications. The hurdle for grid storage batteries using an alternative chemistry of low cost is that regardless of the cost, the incentives (even though present) to put in place mass manufacturing are less than those for vehicle batteries. Therfore any breakthrough in chemistry that can lower the cost of vehicle batteries will also potentially take the cost low enough to make them viable wind and solar storage batteries. Once that manufacturing is in place and the cost meet the solar and wind applications there will be little incentive for huge investments for alternative grid storage chemistries.

Finally, the authors of the paper study say "Increased volume beyond that does little to reduce unit costs, except potentially indirectly through factors such as experience, learning, and innovation, they determined."
Experience, learning, and innovation are exactly the things that differentiate good competitive companies from failures and without a doubt is what the winners of this capitalistic competition will have to do. It is not that they may do it or should do it or could do it, they will have to do it and someone will.


I really think that tesla will go bankrupt someday. For now they are not making profits and the future might be worst. I never heard anyone saying this in all blog sites im in. The cost per battery will go higher because it's like petrol if the demand increase than the price increase. Lithium and grafite prices will go higher with an increase in demand. Battery cars are subsidized for now but it won't last because it's inneficient. Tesla lose money with their 90 000$ model s and they won't be able to sell a cheaper car. They will lack money to invest for the future. A slight drop in petrol price will slowdown the sales of tesla. Notice this im the only one predicting the future bankruptcy of tesla. Everything will collapse like battery 123 or fisker or tucker. Car business is not profitable even for gasoline cars, look at gm, Volvo, Ferrari, chryler, mazda, Suzuki, fisker,etc. This is a high investment low profit margin business. Look at petrol company, billions of profit everywhere till 100 years, they can sustain a lower price for a long time. Sales of suvs are on the rise. This is the end of clean air, because of gm, exxon, shell, ford, tesla, DuPont, air liquid, Toyota, customers, goverments, bloggers, scientists, universities, citgo, Aramco, shale oil and gas, pg&e, hess, Nissan, opec, Venezuela, tar sands, nuclear, highways, airports, truckers, bp, washinton, London, farmers, bosch spark-plugs and fuel pump and gasoline injectors, Chrysler, general-electric locomotives and jet engines, etc.


Cheap gas did not last long. Price has gone up from $1.23/L to $1.35/L in less than a week.

It could be back up to $1.50/L by the end of the month.

The comments to this entry are closed.