ICE versus Electric Cost per Mile

The best-case economics are there, but the long life of the packs will indirectly scare away price-sensitive consumers for the next 5-10 years.


Tesla made quite a splash last week with data showing their cost per mile would be that of diesel. The specific graph they showed was this:


This puts the cost at electric at roughly 17% below that of diesel. In a previous post we took a look at fixed costs plus electricity and repair costs to arrive at the battery depreciation cost Telsa might be seeing. 

It's often believed that the move to electric will happen only if governments are there to make it happen. But in fact, there isn't much governments can do because the numbers get so big so quickly. Even the most progressive states and countries are having trouble keeping up with subsidies. And this shouldn't be a surprise: A nation spends perhaps 8% of its GDP on energy, and subsidizing that by overspending (which is what subsidizing effectively is) would cost a fortune. For the US, our energy spend is roughly $1.6T/year, and if subsidy required you to overspend by 25%, then that would take another $400B to change behavior. That's not going to happen. The US budget doesn't have anywhere near that much to spend. That is approaching the annual military spend.

And so, there's no much governments can do to change policy more than a year or two here or there. That is, without any government intervention, maybe half of all cars sold are electric in 2030. And with LOTS of government subsidy, perhaps that might be 2028. But in the end, it's noise. We like to think it's not. But it is. 

The tipping point

The tipping point happens if and only if electric cars are better. Period. And part of being "better" means you get more for your money. And once that occurs, then there's not any subsidy required. It will happen on its own.

Today, electric cars are on the cusp of costing less per mile than ICE, as I'll show below. The final barrier to an electric car is that you must effectively pay for 10 years of gasoline up front. Instead of a weekly $50 fill up for 10 years, you pay an extra $20,000 or so up front to purchase a battery, and in return get a much cheaper weekly gas bill. 

That's an easy sell if you are trying to sell to someone that has lots of savings. But it's a very tough sell if you are trying to sell to someone just getting by. The former sees it as a great long-term savings. The latter cannot afford it, and so they will fall back to ICE simply because it has a much lower acquisition cost EVEN IF long-term it's more expensive.

This is the reason furniture rental exists at all. If a new couch costs $500, but renting a couch costs $50/month, there are those that will opt to rent a couch for 2 years even though they are overpaying for the couch. They have no choice. 

Now, the good news is that financial products can fix this to some degree. A car maker could sell you a car and then lease you the battery. But there are some sticking points with that. There's a reason there aren't 10 or 15 year car leases today. Why would there be 10 or 15 year battery leases? Leasing businesses tend to gravitate towards 3, or maybe a 5 year lease. 

Let's start by figuring out the cost per mile for gas and electric. This will be a first-order estimate, ignoring purchase price and maintenance. 

Comparing ICE and Electric

Let's compare the new Tesla Model 3 with something similar in terms of size, weight &  performance. Rather than going through an exhaustive sorting process, let's just pick a car like the Acura TLX. They compare favorably:

Tesla Model 3 Std Acura TLX V6 P-AWS
0 to 60 (sec) 5.6 ~6
Curb Weight (pounds) 3549 3600
MSRP (no subsidy) $35,000 $34,000
Length (inches) 185 191
Width (inches) 82 73
Headroom (inches) 39.6 37.2
Front Legroom (inches) 42.7 42.6

The TLX V6 FWD has a highway rating of 30 MPH, and a combined of 23 MPG. 

Battery Cycle Life

The previous post discussed the importance of depth of discharge on cell longevity. If you discharge a cell completely each time you use it, you'll see 500-600 cycles. If you discharge it just 70-80%, you'll see over a thousand full-cycle equivalents (in terms of total Wh delivered)

Tesla offers an 8-year unlimited mileage warranty on their 85 kWh battery, and their 65 kWh pack is warranted for 125,000 miles. At 344 Wh/mile, this is 43M total Wh delivered over the 125,000 mile life of the pack. This is 661 full cycles. 

Now, what is interesting is that Tesla didn't just scale the 65 kWh pack to 85 kWh when contemplating their warranty. If a 65 kWh pack gets a 125,000 mile warranty, why not give a 85 kWh pack a proportional 163,000 mile warranty? Perhaps the data is showing 65 kWh customers are doing a much deeper discharge than the 85 kWh customers? This does have implications for the smaller pack on the Model 3.

But the pack is capable of much more as we know. User-reported data on Model S is suggesting 150,000 miles to 90%. And since the usable life of the cells is to 80%, this suggests that 250,000 miles is readily achievable. In fact, Musk has claimed they have measurements in the lab that have achieved 500,000 miles with 80% remaining. But let's assume that 250,000 will be readily achieved by most users on the Model S.

Tesla assumes the following baseline figures for efficiency when supercharging:

Model Efficiency
Model S 344 Wh/Mile
Model X 369 Wh/Mile
Model 3 237 Wh/Mile

The above figures correspond closely to the EPA rated cycles. The Model 3, for example, claims 220 miles from the 50 kWh pack, which is 227 Wh/km. From the Model S data, this means that Tesla can likely achieve 250,000 / 344 = 726 full cycle equivalents on the Model S, which as you'd expect would be higher than the 661 derived from warranty estimates above. Your average lifetime must always be greater than your warranty lifetime, otherwise you are going to go broke. 

With 726 cycles from the pack, this suggests the Model 3 lifetime would be 159,000 miles. In fact, it could be worse if Model 3 users are discharging the battery much more between charges than the Model S users. But for now, we'll go with that figure. 

The last piece we need is pack cost. Assume in 2019 Tesla is at $150/KWH, and they are selling that the consumer for $250/kWh (40% gross margin, which includes warranty, R&D, SG&A, etc), then the user is getting 159,000 miles of driving for 250 * 50 = $12,500 which is 7.8 cents per mile. Adding in electricity cost at $0.12/kwh adds another 2.8 cents per mile, for a total of 10.6 cents per mile.

The Acura TLX, at $2.50 gallon and 24 MPG would yield $2.50/24 = 10.4 cents per mile. This nets out at:

Tesla Model 3 Acura TLX
Cents per mile 10.6 10.4

OK, so today, the costs are very close. Maintenance on the ICE would be higher, of course. But remember, this assumes the Model 3 owner squeezed all the life out of the pack by driving the pack until exhausted. For a typical car owner, this would be nearly 12 years. If you own the car for 6 years, and sell it for 40% of its purchase price, then the operating cost per mile of the gas car remains the same, but the cost of the electric battery depreciation moves from 7.8 cents per mile to 9.3 cents per mile. 


If you are willing to buy an electric and drive it into the ground in 10 years, and much of your driving is city, then you will probably be able to beat gas on cost handily. But if not, then gas will still be more economical for you AND you don't have to pre-pay for a decade of gasoline the way you do with electric. But the trend is clear: Electric economics will continue to improve every year. 

What is so potent about Tesla's Semi offering is that the battery is fully consumed in just 5 years. With Tesla's consumer vehicles, the fact that the battery lasts 13 to 18 years of normal driving, combined with the fact that you must pay for a long-life battery up front when you purchase the car makes the sell a bit tougher because you've shut out many potential buyers that are focused on monthly economics. Leasing a pack might be an option, but as the pack lifetime increases, it gets harder to lease because the first year costs must be increased to cover the likelihood that there won't be any takers to lease the aged pack down the road. It's the same with cars today: You must overpay to lease a car, and leases aren't usually considered for anything other than a new cars. 

Battery technology that delivered 1/5th the number of cycles for 1/5th the cost would alleviate much of this. You'd just replace the pack every 3 years for a modest cost that was roughly equivalent to 3 years of gasoline spending.

Benchmark: Solar Costs

NREL publishes cost benchmarks on solar. One metric that is very useful is the LCOE, or "levelized cost of energy." This is NRELs best estimate of the cost to generate electricity via solar. It includes not only the upfront installation costs, but it also takes into account the expected lifetime (30 years), taxes, working capital required as well as the debt service, construction, etc. In other words, it's very comprehensive. 

The appendix B summary is very useful, indicating that utility-scale generation is achieving $0.03-$0.06 $/kWh, depending on region of the US. 


Tesla Semi Battery Economics

Tesla appears to have hit on a recipe that will permit them to lease a battery that will completely exhausted in 5 years. This is potentially game-changing because it means payback comes quickly.

This week Tesla announced the details on their semi truck. Some key metrics below:

  • 500 mile range (800 km) @ 60 MPH (96 kph)
  • 7 hours of driving time
  • 400 miles of range in 30 minutes of charging
  • $0.07/kWh network energy costs
  • < 2 kWh per mile, which is 1250 Wh/km

This post will look into the economics in more detail, and also determine if the application of the cells used in the Semi is that different from that of the Model S today. If you can't wait to the end, the economics look promising, and taking the cell tech straight from the Model S/3 would probably work just fine. In fact, the Semi is probably more gentle on the cells than the Model S.

Cost per Mile

Most interesting this week was Tesla's graph on cost-per-mile, which is $1.26/mile.


Tesla is assuming a $0.07/kWh cost for energy, and they will presumably deliver on these numbers via a combination of strong utility agreements (eg. charge megacharger batteries during off-peak) married with some local generation and storage. But the energy costs, as we'll see, are a fairly small part of the overall per-mile cost (roughly 11% of the $1.26/mile figure). There's no reason to doubt Tesla's claim here.

Battery Lifetime

With 500 miles of range, and 2 kWh/mile, we can deduce the battery capacity is about 1 MWh. This is larger than many had been thinking prior to this announcement. But there's a good reason for this.

The lifetime of lithium batteries is related to the depth of discharge, among other things. With reasonable charge and discharge rates, the typical specs from cell manufacturers usually suggests that at very high depths of discharge (nearly 100%) you will get about 500 cycles over the life of the cell. That means that if your battery capacity is 1 watt-hours, then you have 500 watt-hours of total energy delivered over the life of the cell. Assume that you only discharge to 50%, though. How many watt-hours of energy will be delivered over a lifetime? You might think you'd get get twice as many half cycles, which would deliver the same 500 watts hours of the life of the cell. But in fact, something magical happens: The total energy delivered can grow nearly 3X, meaning the cost per delivered kWh is now 1/3. That is, the 1 watt-hour cell can deliver as much as 1500 watt-hours of energy over its life if you take care of the cell and follow some very simple rules.

Musk made an important point during his presentation: 80% of trucking routes are 250 miles (400 km) or fewer. This starts to give us insight into where Tesla might be headed with this: Start with a 1 MWh pack, which could deliver a max range of 500 miles (800 km) at 2 kWh/mile. Derate that so that you'd seldom consume more than 70-80% of the total charge (70 to 80% Depth of Discharge). That gives you 1200 to 1500 full cycle equivalents, and a total range of 625,000 miles (1M km) over the lifetime of the pack. 

Diesel and Electric Economics

In a diesel truck, fuel costs are about 40% of the total operating costs per mile, and repairs and maintenance are about 10%. This means that half of the $1.51/mile figure goes towards fixed items such as driver salary, tires and insurance. This breakdown is in the table below.


Since Tesla claimed they can hit $1.26/mile and the fixed costs are $0.76/mile, we can subtract out repair cost and arrive at the battery depreciation costs. Let's assume that since electrics are simpler than ICE (fewer moving parts), that the repair costs is only half as much. And Tesla said the cost to travel a mile at speed would be $0.14. This gives us $0.18/km left over for battery depreciation costs. 

In the table below, we break down the battery costs, and arrive at a "operating cost per km". This figure doesn't include the electricity cost. This is simply the battery cost divided by the total number of miles delivered. And once we have the operating cost, we can derive the pack cost.


This suggests that Tesla will initially be leasing (selling?) the pack out for about $187/kW to hit the published per-mile figure they claimed. No matter how you slice it, the pack is expensive.

Remember, too, that Tesla stated their pack cost on Model 3 was "below $190/KWH". In fact, in 2014, Elon Musk noted in an investor call "I would be disappointed if it took us 10 years to get to a $100/kWh pack." With production slated to begin in 2019, some penciling around the quote above (assuming 8% YoY) suggests a pack price around $150/kWh for 2020. Yes, the pack is expensive. Probably $150,000 to $200,000. But remember, it's an asset that generates revenue. Your monthly pack lease bill should be less than your monthly fuel bill. It's that simple. 

If a typical truck is logging 130K miles per year ($180,000 annual operating cost/$1.38 per mile = 130K miles = 208 km), then this means a pack has a 5 year life time. Keep that figure in mind, it's important later.

Battery Kindness

Model S P85D

We'll look at this in more detail in a future post, but it's important to note just how gentle Tesla is being to these cells. 

The Model S can hit about 300 miles of range from its 85 kWh battery at a constant 60 MPH speed. This is 5 hours of driving, and suggests 17 kWh consumed per hour and thus 17 kW is required from the battery to move the car. This is a 0.2C rate discharge. Most driving is probably occurring around this rate, and this rate is very comfortable for 18650-type cells to deal with. 

Supercharging in the Model S delivers 50% charge in 20 minutes. On a P85, this means around 45 kWh delivered in 20 minutes (ignoring losses), which means the battery was hit with 135 kW for 20 minutes. This is about a 1.6C charge rate. In order to charge at this rate, the battery needs to be nearly flat. As the battery state-of-charge increases, the charging must be tapered down. 

The Model S P85 pack can experience two different types of discharge extremes. The first is a short-term discharge experienced during hard acceleration. With a peak power of 310 kW, this is a 3.6C draw from the pack. At this level of draw, the cells temps will quickly rise. Worse, the temp rise is happening deep in the cell before any thermal solution can help. For this reason, this mode of operation can only happen when the cells are at a reasonable temperature to start AND for a few seconds at a time.

The second discharge extreme is sustained during high speed driving. Tesla's graph shows the Model S P85 with a range of 200 mile at 80 mph (about 0.4C discharge).  Other reports on the internet indicate a P90 on the track thermally limiting at about 100 MPH. Eyeballing from the Tesla graph, the range at that speed would drop to perhaps 150 miles, which is about 0.6C. It's not clear if the thermal limit is due to the pack or the motor or both. In a perfectly designed system, both power train and battery would limit at about the same point. Regardless, we'll go with that figure. 


The numbers for the Semi appear to be much kinder to the pack, primarily because the pack is so large. If the Semi indeed has 2 kWh/mile energy usage at 60 MPH, then this means the semi is consuming 120 kW to move at 60 MPH. This is a discharge rate of 0.12C. This is quite a bit lower than the P85. But that's good, because it makes the battery happier, and happy batteries live longer.

To deliver 400 miles of range in 30 minutes of charging means 400 * 2 kWh = 800 kWh delivered in 30 minutes. This is 1.6 MW for 30 minutes, or about a 1.6C charge rate. This is the same as the P85 at the Supercharger. 

But how about hills? We saw above it takes 120 KW to move the loaded Semi at 60 MPH. What does a 7 degree hill mean? High-school physics says the power adder required would be:

P = m * g * sin(theta) * v

Where P is the added power required (assuming 100% efficiency), m is the mass (kg), theta is the hill angle in radians, and v is the velocity in m/s. At 80,000 pounds and 60 MPH, 7 degree hill would be a doubling of flat-ground cruise power. This doubles the cell draw to 0.24C, which is still very comfortable. It will indeed cut your range. But if the backside of the hill is properly handled, the impact might not be so bad. 

A bit more study and speculation is needed for the acceleration rates. But we'll take a swing at that later.  

New cell tech?

Tesla has a decade of 18650 learning under their belt. No doubt, they are probably running thousands of driving profiles on individual cells 24x7 to learn just what the chemistry likes and doesn't like and what it will and won't tolerate. They don't need a full pack to learn this. They can run a scaled down profile on a single 18650 cell, taking into account hard acceleration, cruising, regen, fast and slow charging--all of the parameters you might ever want to understand. Their engineers can then make statements along the lines of "With a sample size of 20 cells and confidence of 98%, Supercharging at 1.7C every 5 standard charges reduces cell cycle life by just 1%." In other words, this is a numbers game. And to win the game, you need to know the rules better than your competitor. 

So, is Tesla using some brand new cell technology for the Semi? Some in the last few days have suggested that. But it seems that they could meet the top-level requirements easily with the current cells used in the Model S (or Model 3). And this is key, because if you are Tesla, you'd like to bet everything on a single core chemistry (perhaps with a few tweaks) if you could. 

P85 Semi
Average Draw at 60 MPH 0.2C 0.12C
Supercharge Rate 1.7C 1.6C
Hard accel (2 sec) 3.6C ???
Sustained High Speed ~0.6C ???

From the above, we can see that some new, exotic technology isn't needed from a charge/discharge rate perspective. What would be nice, obviously, is a longer cycle life. If you can get 10% more cycle life from the cells, then you reduce your costs by 10%.

Much has been written about cities and governments mandating the end of ICE. At the end of the day, ICE is dead REGARDLESS of what cities and governments decide to do. Purely because we've seen per-mile costs destroyed by electric in passenger cars like the Model 3, and now, per-mile costs destroyed by electric in Semis. It's the economics. And it's irrefutable. 


Tesla appears to have hit on a recipe where they can generate an OK return on a very expensive asset from the beginning. Roughly, in 2020, it seems reasonable that Tesla could be looking at a $150/KWH pack cost, and 5 year lease on a pack could generate $187/KWH. So, out of the gates that would be a 4.4% return beginning in 2020. Meh. OK. 

But when you take into account the fact that gas prices over the next decade will probably remain where they are today or go up (meaning Tesla doesn't have to drop their per mile price and might be even able to raise it), and when you also factor in Tesla's claim they can readily hit $100/KWH pack price in 2024, this suggests than a 5 year lease on a pack would start generating 13.7% annual returns. Wowza.