$30,000 from a Tesla Robotaxi?…not as crazy as it sounds
Back in the spring, Elon Musk and Tesla unveiled detail on their autonomous vehicle aspirations during an investor-attended “Autonomy Day.” In short, the company is aiming to roll out a Level 5 self-driving robotaxi service by the end of 2020. This service will allow any Tesla owner to add their car to a network from which consumers will be able to summon a self-driving Tesla on-demand to drive them to a destination for a fee. This creates a self-driving, Tesla-specific competitor to Uber and Lyft.
We do not have full confidence in the technical feasibility of such a network. To start, many disagree with Tesla’s decision to forgo using LIDAR sensors, technology that Waymo, Cruise, and other major AV players rely upon to operate their vehicles. While some experts agree with Tesla that LIDAR is unnecessary, others believe that Tesla’s self-driving capability is not very advanced. Moreover, Tesla also claims it has a data advantage because its self driving system runs in “shadow mode” on all cars, racking up miles of simulation that can be used to improve its technology. However, the quality of this data is also up for debate, as some argue the cars are not analyzing nor uploading much of the data collected.
All in all, we have doubts about the 2020 roll-out for Tesla’s robotaxi network. But, let’s push these aside for a moment. Assume that Elon Musk is right and the technology is ready to go. In this scenario, will Tesla be able to fulfill its promise that a Tesla robotaxi network will generate ~$30,000 a year for Tesla owners? To answer this question, we created a bottoms-up model, using a combination of public data and assumptions to forecast the revenue and costs associated with the Tesla network. In this scenario, we take the point of view of a Tesla owner in Evanston, IL who is leasing a Model 3 extended range on a three-year term. The findings from our base case model are summarized in the table below ( full model here):
As you can see, we do see a path to ~$30,000 a year for Tesla owners from the robotaxi service. However, this forecast is very sensitive to several critical assumptions, some of which are different than what Tesla presented at Autonomy Day. Namely, we’d like to highlight 1) pricing; 2) owner usage; 3) utilization; 4) insurance; 5) maintenance; and 6) Tesla-owner profit sharing.
During the Autonomy Day presentation, Elon stated that Tesla assumed $1 per mile pricing for its projections, but admitted that the figure was largely arbitrary. The real takeaway from this is that Tesla aims to undercut Uber and Lyft, which are priced between $2 and $3 per mile. Users willingness to pay for a Tesla robotaxi could be higher or lower depending on the relative strength of a few factors:
Overall, the lower willingness to pay factors appear to be ones that should dissipate over time, while the higher willingness to pay factors are more sustainable. A lower price, though, also generates more demand for a service, stealing vehicle-miles-traveled share from not only human-driven ridehail, but also personally-owned cars, public transit, and other modes.
For simplicity’s sake, we assume in our base case that Tesla’s robotaxi rides are priced the same as Uber rides in Chicago ($2.48 per mile), resulting in $29,207 of gross profit per year for Tesla owners. These calculations assume a percentage of rides that result in a disengagement from the self-driving system (and require teleoperation to get “unstuck”). We assume as compensation for a faulty ride that the rider is refunded in full for these trips. Using Waymo’s market-leading rate of 1 disengagement per 11,154 miles (we hope Tesla won’t go-to-market with anything lower), this has a negligible impact on gross revenue.
If, instead, we assume the pricing presented by Tesla ($1.00 per mile), we see a loss of -$2,225 per year after charging, maintenance, financing, and insurance, suggesting that this rate is actually not feasible. We calculate the break-even price for Tesla owners to be $1.11 per mile. While this price still provides significant cushion to current ridehail rates, it does not take into account Tesla’s margin (we project Tesla would earn ~$8,000 of gross revenue per year at $1.11 per mile). The company will have significant expenses associated with the Tesla network, including depreciation, insurance, customer service, operations, data processing, and more. It will have to either price high enough to earn a direct profit on the robotaxi service OR increase sales of other, profitable products or services (e.g., more people buy cars due to self-driving capabilities).
Robotaxi profits will vary depending on how individual owners choose to deploy their vehicles on the network. On one end of the spectrum, you have the commuter, who uses the vehicle only a couple hours a day to get to and from work, and seeks to maximize the value of their car by deploying it all other hours of the day. On the other end of the spectrum you have the stay-at-home parent, who needs to use their vehicle throughout the day to run errands and to drive kids around. They might only deploy the car when they know they won’t need it for many hours in a row. The more the owner uses the car, the less that will be earned via the robotaxi.
For our base case scenario, we assume that the owner uses the vehicle for two thirty-minute commutes per weekday (plus an additional thirty minutes for the car to reach the owner). This equates to ten hours a week dedicated to commuting. In addition we assume the owner will want the freedom to have access to their car at all times during every weekend, and 10 additional weekdays (e.g., for vacations). This results in an additional 53 hours of personal owner time per week (48 weekend hours plus the allocated portion of vacation). Over the course of a year this translates to the owner using the car for 134 days (or 37%) of the year, leaving the remainder available for robotaxi use and servicing.
On Autonomy Day, Elon indicated that current Tesla owners use their cars ~12 hours per week, translating to 7% of the year. We believe that this is too aggressive of an assumption because 1) it doesn’t account for the empty miles required to have the vehicle reach its owner and 2) it doesn’t take into account the flexibility of having your car available when you want it. We don’t doubt that owners only use their car a few hours on the weekends or during vacation, but we believe that owners do value the ability to access it whenever they see fit during these times.
Only time will tell how consumers end up splitting their car’s time between personal use and robotaxi deployment, but our model demonstrates the impact of this variance on owner’s bottom line. Specifically, we calculated that a 20% change in days of user ownership per year results in a 18% shift in owner gross profit. Thus, someone who deploys their car 12 more hours per week than we anticipated will see their profit rise from $29,000 to nearly $35,000.
Utilization, or the percentage of time that the robotaxi is carrying a rider divided by the total amount of time it is deployed in the fleet, is a critical factor for profitability. Specifically, utilization is directly linked to revenue, as the time with passengers is billable and empty miles are not. Utilization has been found to be 32–50% for taxis in major U.S. cities, and 43–55% for Uber. It is unclear why a self-driving ridehail would have a different utilization than a human-driven ridehail, so we can conservatively use one of Uber’s lower city utilization rates: Boston at 46%. Our model suggests that gross profit is very sensitive to the utilization rate: if utilization decreases by 20% (to 37%), then owners’ gross profit decreases by 36% (to $19,000 per year).
We see two key factors that may negatively impact this variable. For one, we assume uniform utilization across time of day (i.e., utilization at 10pm is the same as 5am). This may be overly aggressive, but some studies have found that, “drivers achieve essentially the same utilization rate regardless of how long they work.” Secondly, we believe it unlikely that the Tesla network would be able to match the utilization rate of Uber. We have seen that ridehail networks with less drivers have lower utilization, and the Tesla network will almost certainly be smaller than Uber’s: Tesla has sold ~250k cars in the US while Uber has ~750k drivers. Tesla could mitigate these headwinds by pricing more competitively; however, this will eat into profits as discussed above and make it difficult to reach $30,000 of profit per year for owners.
Outside of technical feasibility, one of the most critiqued aspects of Tesla’s robotaxi plan is insurance, as it will be difficult for an insurance company to develop plans for such an unknown technology. A few days after Autonomy Day, Musk addressed this concern by announcing that Tesla will create its own insurance product. Critics such as Warren Buffett, who owns Geico, have stated, “the likelihood of success of auto companies getting into the insurance business is about as likely as insurance companies getting into the auto industry.” This is partially due to the fact that insurance companies rely on years of actuarial expertise to turn a profit, but is also due to the type of risk they are mitigating. Insurance companies are able to spread independent risks across a large pool of drivers so that if one driver has an accident, the premiums from all the other drivers can cover the loss. The problem with insurance owned by Tesla is that the risk is no longer independent and could be correlated. If one driver in a Tesla has an accident due to mechanical or computing error, it might be likely that many more Tesla owners have similar accidents. This would mean that the premiums from the fleet wouldn’t be able to cover the losses.
Nonetheless, assuming Tesla is able to figure out the intricacies of the insurance business and diversify its risk profile, how would it be priced to consumers? The Model 3 is expensive to insure right now, due to the high cost and long lead times of replacement parts for Tesla’s involved in accidents. According to Automotive News, “the average Model 3 insurance cost across 150 ZIP codes is $2,814 per year, which is $35 less than the cost of insuring a Porsche 911.”
But for Tesla’s robotaxi plan to work, owners will likely want more than just standard auto insurance. Many rideshare drivers supplement with additional insurance to cover incidents that are not covered by Uber or Lyft’s policies or traditional non-commercial policies. These supplemental policies average about $150 a month. It is still an open question as to who has liability in these scenarios (i.e., is a Tesla owner liable for damages caused by its vehicle when deployed as a robotaxi?), but, ultimately, we decided to include this expense for owners in our model to be conservative.
Combining the two types of insurance (vehicle and rideshare) gets us to a total cost of $385 per month or $4,614 per year. But considering Tesla has never offered insurance for its vehicles, no company has ever offered insurance for a fleet of privately owned autonomous taxis, and most insurance quotes aren’t for 70,000 miles per year, it’s really difficult to predict how the insurance will be priced. If our estimate is 30% off in either direction (i.e., insurance is ~$3,000 or ~$6,000 per year) the overall profit equation for consumers changes by +/- 10% or $3,000.
In short, if Tesla is able to pull off insurance and the prices are within the same order of magnitude as current offerings, the price of insurance won’t make or break the profitability equation for Model 3 owners.
Repair Cost & Downtime
Maintenance and repairs are factored into two different aspects of our model. The first is annual cost for routine maintenance such as tires, brakes, and other items that wear out with 70,000 miles per year. AAA estimates the cost of these items for an EV is 7.6 cents per mile which equates to $5,358 per year for our high mileage robotaxis. Because EV’s have so much torque and are quite heavy, they are known to notoriously eat through tires, so these figures make sense.
The more complicated consideration for maintenance and repairs is what happens when a Tesla is involved in an accident. A few taxi fleets have already tested ownership of multiple Tesla’s at once and have struggled to keep the cars on the road continuously. This is because Tesla’s supply chain challenges and overwhelmed service centers, result in repair times that last days or weeks, for cars that have minor problems, accident damage, or quality control issues. For a taxi fleet owner, downtime is lost money and can dramatically hurt the business. In fact, some Tesla taxi fleets have gone bankrupt for this exact issue.
In our model, we assume that the car will have repair-related downtime of two weeks (14 days) over the course of a year. For each incremental two weeks, though, overall profit to the owner decreases 9%. Thus, if a car is stranded in the shop for many months after a minor fender bender a sizable portion of the profit a consumer earns in a year could be erased.
Established OEMs that have tried-and-true repair procedures and strong dealer networks are efficient at fixing broken cars. But given Tesla’s track record of questionable quality and long repair times, the opportunity cost of broken robotaxis could really change the profit equation for Model 3 consumers.
One additional thing to highlight is cleaning. In our model, we have assumed that downtime associated from cleaning of the interior and exterior of the vehicle overlaps with charging. We believe downtime can be limited in such a way through a combination of automated cleaners, particularly for sensors, as well as financial incentives for riders (e.g., charge riders a fee for leaving a mess in the vehicle, and pay riders for cleaning up messes from previous trips). Therefore, we do not envision cleaning having a significant impact on robotaxi economics.
Tesla-Owner Profit Sharing
Ultimately the profitability of robotaxis for Model 3 owners comes down to how value is split between Tesla and the owner. Because of the complicated nature of this model, money would be exchanged between Tesla and owners across multiple dimensions: a monthly lease payment, the robo-taxi profit revenue split, and insurance costs (which we’ve already covered).
A long-range Model 3 is currently listed on the Tesla website for $712 per month. The problem with this payment is that it takes into account the current economics of leasing, but not the robotaxi future. This lease is for 15,000 miles per year for three years, but robotaxis will supposedly cover over 70,000 miles per year, indicating that the payment should be significantly higher. On the other hand, it could be argued that the lease payment should be much lower. Musk has stated that Tesla’s are actually appreciating assets now which in theory means the lease payment should be zero or negative, since the car would be worth more to Tesla when it is returned three years later. We believe this is unlikely to say the least, as it’s hard to imagine Tesla would not charge a substantial lease payment for a brand new Model 3.
The other major factor that needs to be considered is how much revenue is shared between Tesla and the owner from robotaxi fares. During Autonomy Day, Musk said Tesla would likely take 25–30% as a fee for running the platform. In our base case, we’ve assumed 25%.
Combining lease payment changes and profit sharing changes gives Tesla the ability to dramatically change the value proposition for both itself and the Model 3 owner. Here is how the net profit to owners varies:
Tesla will have to think critically about the best way to split profits with its customers to make it advantageous for both parties to deploy the robotaxi network. Perhaps the company will create innovative contracts, such as dynamic lease pricing based on robotaxi deployment, or reduced lease prices with the stipulation that the car must be deployed a certain number of hours per week. Regardless of the final structure, this will be a first for OEMs to charge consumers to use a vehicle, while simultaneously receiving commission from the owner for the value created by the car.
We created this financial model to test Tesla’s claim that a Model 3 deployed in a robotaxi network could generate $30,000 profit per year for its owner. As demonstrated with our base case model, this claim is feasible given a certain set of assumptions. But, as discussed, slight tweaks in key variables such as utilization or pricing are likely to impact owner profit dramatically. Notably, the two variables that sway profitability the most are those for which the consumer has the least control (e.g., utilization and pricing).
If Tesla is not able to establish a network that has high rider demand, is optimized for high utilization, and is priced properly, then owning a Model 3 is likely to be a losing proposition for consumers. This could be further exaggerated if value isn’t split equitably. The variables that owners do have more control over (downtime and owner usage) play significantly less of a role in overall profit.
This post was written under the assumption that Tesla can get to Level 5 autonomy and deploy a ridesharing network. But why did Tesla feel the need to make this announcement in April of 2019, when this technology appears to be a long ways away? We believe there are several explanations for Autonomy Day:
- The company genuinely wanted to inform investors about its progress and plans for AVs
- Elon Musk wanted to distract investors from Tesla’s disappointing Q1 earnings release a few days later and potentially declining sales demand
- Tesla wanted to create marketing buzz around Model 3 to sell more cars this year — through the promise of future features
If we truly believed that Tesla was able to create a fleet of robotaxis that generate significant profits without owner intervention, then it would be best for the company to simply create Model 3’s for its own robo-fleet and not sell them to consumers. But since this is not the case, we have to believe that Musk’s motivation for Autonomy Day was some combination of points 2 and 3 above — in other words, while figuring out the economics of a robotaxi network is an interesting thought exercise, it seems more likely this announcement was meant as a distraction to investors and a marketing exercise for consumers.
Originally published at https://medium.com on May 15, 2019.