Tesla & Robotaxi Economics: The Network That Optimizes The Machine
Credit to Author: Rahul Sonnad| Date: Thu, 09 Jan 2020 06:34:20 +0000
Published on January 9th, 2020 | by Rahul Sonnad
January 9th, 2020 by Rahul Sonnad
There is a long-standing argument about whether Tesla is a car company or a tech company. This argument is typically made in the context of how Wall Street should value the company. If Tesla is just a new kind of carmaker, it should be valued at something significantly less than its annual revenue — maybe 25% to 50%, like Ford and GM. Alternatively, if Tesla is a tech company, then it could be valued significantly higher. Google, for example, is trading at about 6 times its revenues today, Facebook at 8, Microsoft at 9, and Salesforce at 10 times. Tesla trades at around 3 times its revenues with some profits, while Uber and Lyft trade around 4 times revenues and have never been profitable.
It is starting to become clear that Tesla is building a business whose economics and business model will not be driven dominantly by car sales. Nor will they be like a classic tech company that builds high margin software or online networks that monetize an audience. Instead, Tesla is on the verge of building a new kind of business whose core focus will be to continually lower the cost of sustainable transportation. At its foundation is an electric hardware platform, with increasingly autonomous capabilities. On top of this, Tesla will both operate its own mobility network and enable others to run specialized networks of their own for freight and other transportation scenarios. While Uber and Lyft would be the closest analogues, the nature of Tesla’s robotaxi network will encompass a remarkably wider range of activities. Tesla will design and manufacture the vehicles on its network, and manage all aspects of its operation, including the generation, storage, and dispensing of the fuel. And in some cases Tesla will own the vehicles on the network. On the freight side, Tesla will enable a commercial platform designed to efficiently move cargo, that will offer the most flexible low cost movement of goods across our roads.
The automotive industry is undergoing a fundamental platform shift driven by three core technological advancements:
Because these three new technologies are all rapidly advancing as they intertwine, it’s very hard to extrapolate to where Tesla will end up — and what this means to the automotive industry in general.
It may be slightly easier to imagine what Tesla’s battery and solar innovations might mean to the global energy generation and storage industry, but here we’ll focus on the automotive business. This business is about to unleash a transformation of the automotive industry that will likely be the largest economic transformation of any industry in history.
Over the last 40 years, we have seen a few major platform shifts in other industries: from mainframes to the “microcomputers” (led by Microsoft and Intel), from desktop computing to the internet (led by Netscape, Yahoo, and later Google), from the flip phone to the internet-enabled smartphone (led by Apple), and from on-premise corporate applications to cloud computing (led by Salesforce). In all of these cases, what emerged was not a simple extension of the previous platform business, but rather a complete reconfiguration of the ecosystem, with new economics, new business models, and new consumer capabilities, and — in most cases — new corporate leaders. The idea that the fastest growing company in the ’80s would license a few megabytes of computer software to other companies that were building hardware would have been hard to imagine in the ’70s. The idea that Microsoft would lose its tech dominance to a search engine company that would offer its product for free was unthinkable at Microsoft in the ’90s. Yet, the revenue that Google captured was not even from Microsoft’s core business, but rather from the advertising industry. The idea that owning the phone interface and monetizing it was always the holy grail for mobile carriers and flip-phone makers, until the iPhone prompted a rapid reorientation of the industry around vertically integrated ecosystems.
Today, Tesla is in the second inning of a familiar situation. Those who think this will be less significant from a business evolution perspective than the microcomputer or the smartphone are likely not keeping up with the trajectory of Tesla, or may be in some state of denial like those at Nokia and Blackberry in 2006.
Arguably, the idea of Tesla as a platform was evident to some shortly after the Model S was released, with its pervasive cellular connectivity and the industry’s first open APIs. However, over the last 5 years, the shape and future configuration of Tesla’s automotive platform has become much more pronounced as autonomy, electrification, charging infrastructure, connectivity, and general car features have all been converging in a highly orchestrated manner. The result will be a platform shift more dramatic than anything that has come before.
The closest analogy might be what Apple has done with the iPhone. But if you compare what Tesla is doing to what Apple did with the iPhone, it’s striking how much further Tesla is going in terms of vertical integration. This has been driven by a combination of necessity — no one else will provide what’s needed to make EVs compelling — and a desire for control — no one else will optimize correctly. To make the analogy with Apple and the iPhone comparable, you would need to imagine Apple acting much more broadly. It would not only be designing the phone and operating system, but also:
If Apple was doing all of this, it would be a reasonable analogy to what Tesla is undertaking today, with the caveat being that the complexity and cost of manufacturing a Tesla is enormously higher than that of a smartphone.
Tesla’s aspirations extend well beyond building the best cars in the world, accompanied by the best energy generation, storage, and charging infrastructure — all manufactured in the most efficient factories. Over the next few years, Tesla is about to launch something that has never been seen before. It is a new kind of business that will transform the economics and mechanics of moving both people and freight: a fully automated transportation network — another word for a robotaxi network.
The fundamental business of Tesla as this decade progresses will be operating this network, whose main components are automobiles, energy generation, and energy storage. While Tesla might make $10,000 to $15,000 on an average car sold, the earnings potential for operating each vehicle could range from $20,000 to $60,000 per year for a handful of years.
Today, the best examples of automotive networks are Uber and Lyft. They allow a person to specify their origin and destination, and efficiently allocate resources to get them to their destination as fast as possible. Interestingly, Uber and Lyft were not enabled by any type of automotive innovations, but rather by the smartphone with GPS. Once this was in the hands of both the traveler and a car driver, Uber and Lyft could create a regional marketplace between these parties and match them in the most efficient manner. Pricing could be used to entice more supply into the market when needed (i.e., Saturday night in Hollywood) and driver incentives would create increasing supply in general. UberEats took this beyond moving people to the realm of moving food, using similar mechanics. Uber and Lyft have no direct connection to the automobile of any type, all their control is proxied through a human driver. They don’t control if a driver accepts a ride, they don’t control where and when the car is fueled, and they don’t control the music or climate in the car. They merely offer drivers earnings opportunities and suggest efficient routes that benefit all parties. Uber doesn’t care how much it costs to fuel the car, the depreciation of the car, or the maintenance costs that accrue. All of these factors and their optimizations are left to their drivers, who must be quite mathematically inclined to understand how these costs affect their earnings.
Uber’s goal is to optimize the amount of top-line revenue of the network, since it captures 25% of this along with a couple extra dollars in fees per ride. In some ways this is a beautiful business model, as free-market dynamics solve the vehicle efficiency problems. If a car is less fuel efficient, it will come out of the driver’s earnings, so this is neutral to Uber. If a car is new and depreciates quickly, it comes out of the driver’s earnings.If they need to sell the car after 20,000 miles, this depreciation becomes evident. Thus, the more driving one does, the more savvy drivers lean towards cars with low total cost per mile. Hence the proliferation of the late-gen Toyota Prius, which had the lowest operational cost of any car when Uber started.
The new generation of transportation network that Tesla is about to create has two main differences. Firstly, there will be no job of driving. Thus, it will be the car owner of a legally autonomous car that earns the money along with Tesla. Secondly, Tesla will optimize everything on a first principles basis. Tesla’s automotive network will be an extension of its vertically integrated software and hardware stack, enabling unprecedented optimization of the network. It will literally extend from when a photon hits a solar roof through the delivery of passengers and freight. It will also extend back into the manufacturing process of vehicles, batteries, and charging equipment. While manufacturing decisions today are driven primarily by driver benefits and vehicle cost, these will soon be realigned around optimizing network efficiency, and no one else is in a position to coordinate this to the level that Tesla will be able to, as it monitors in real time the performance of every aspect of the network.
The first dependency for Tesla in competing in the urban rideshare market is what’s often called “Level 5 Autonomy.” This is the gorilla in the room. Level 5 cars can technically and legally drive around any street and find safe places to drop and pick up people without the help of a human. Many people feel that this won’t be achieved for years, or in their lifetimes. Without this, Tesla’s vision for an automated network will not be possible on urban streets. As has been discussed in many places, the easy challenge is to get this to work in some places most of the time. The hard challenge is to get it to always work better than humans — the “march of 9’s,” as it’s called. When will this march get to the point where you can trust your car more than yourself? The real answer is no one knows for sure. We are not there yet. It may be 3 months or 3 years. Some think longer. It’s clearly a harder problem than some imagined, but the real questions are whether Tesla’s technical and deployment approach to the problem is fundamentally correct and if it will advance at an exponential rate.
If this strategy of equipping millions of vehicles with optical sensors and massive processing power to allow for huge amounts of data to train neural networks are the right ingredients, then Tesla will achieve superhuman autonomous driving soon enough. In the big scheme of things, it doesn’t matter that much if it’s this year or next year or three years from now, and whether the regulations take a couple of years to catch up. What matters is whether Tesla achieves this before others. It’s fair to say that working robotaxis will arrive later than Elon thought they would. His first predictions for the timing of level 5 autonomy put it at “approximately” summer of 2018, more recently updated to “certainly by the end of 2019.” But more importantly, the system has been consistently and rapidly advancing in leaps and bounds and has gotten statistically safer to use. Tesla has also developed a deployment protocol of internal testing, early-access drivers, and early adopters (who check-box that they want advanced access to new autonomy features). In combination, this has worked remarkably well to train the system, with remarkably few incidents compared to those of normal driving. It has also allowed for the rapid deployment of new releases. At this point, it seems almost a given that by the end of this year, highway driving on Tesla Autopilot will be better than that of most human drivers, as it is already better than many.
For those skeptical of Tesla’s potential to conquer city driving, they should withhold judgement until they have witnessed the recent “full self driving sneak preview” release. Recognizing everything on the road is still a big leap away from actually driving on urban streets, but it seems that this is already working on Tesla’s internal test vehicles. We should know in the next couple of months whether city driving is on track. But until we see it released outside of Tesla, the skeptics have some justification to doubt the timeline. The proof will be in the pudding.
One underappreciated aspect of Tesla’s engineering triumph is the degree to which Tesla is constantly optimizing a huge number of variables. One example of its optimization is demonstrated by the measure of EV battery core efficiency. This is derived from the quotient of [Battery kWh / Vehicle Range / Vehicle Weight] and indicates how far a car can move its weight on a certain amount of charge. But this is just the tip of the optimization iceberg (and one that doesn’t factor the manufacturing cost into the equation). Tesla is a veritable optimization machine against a huge array of factors. The design of any product requires trade-offs and optimizations across different variables, but Tesla relentlessly pursues these optimizations between cost, physics, and user experience in a manner unlike anyone else. And the optimizations for a robotaxi network are a superset of those for its vehicles. Despite the fact that Tesla does not yet operate a robotaxi network, many vehicle design decisions seems to be proceeding with an intuitive understanding of the implications they will have on the economics and operation of a robotaxi network.
The fundamental end goal of a transportation network is conceptually simple: make the most profit on a given amount of capital and expenses. If a $50,000 car can generate $60,000 of revenue each year for 4 years, with total expenses of $15,000 a year over that period, then you’re making $130,000 of profit.
But there are a lot of factors that play into these numbers. Clearly, a Tesla Cybertruck costing the same as a Tesla Model 3 will be able to earn more, because it is bigger, fits more people, and fits more cargo. It will also likely have less damage over time, and thus fewer expenses, such as insurance costs and repairs.
Additionally, faster charging creates a more efficient network. If you only need to charge for 90 minutes a day vs. 3 hours (for the same number of miles), you’ll be able to drive more hours and earn more money, maybe about 10% more. Time spent waiting to charge has a similar effect, and the distance you need to drive to charge a car has both time and expense factors. This is an area where Tesla has a huge lead that is likely to get much bigger.
Of course, how long the car lasts is a major factor, along with the repair and maintenance costs. A vehicle that costs $60,000 would yield about $60,000 more in profits over its lifetime if it lasted a million miles vs. 500,000 miles.
Fuel costs are also a significant factor. At the rates that Electrify America often charges today, you might spend another $60,000 over the lifetime of a million-mile vehicle (which could be as low as 3 years with long-distance round-the-clock driving).
Vehicle versatility is another key economic factor. Compare Cybertruck to Model 3. Cybertruck can drive people around Los Angeles during the day, and then tow trailers of cargo between LA and San Diego during the night. Model 3 has insufficient passenger demand to keep it fully utilized at night, creating a lower ceiling on overall utilization. If a Cybertruck can be loaded with 100 square feet of cargo locked in the back, while it picks up passengers headed from LA to San Diego, it can get paid twice for the same trip. And if it can be used to go to Big Bear Ski Resort in a snowstorm, it will have another economic advantage over other vehicles. For this reason, the Cybertruck is significantly more economically powerful than any previous Tesla.
These are the characteristics of the transportation network for which Tesla is laying the foundation. As Tesla trends towards million-mile batteries and increases charging speeds, they will achieve roughly a three times cost advantage vs. gas vehicles. And once Tesla achieves autonomy, it will give another roughly 3× cost advantage vs. human drivers of a comparable EV. Thus, it will be impossible to compete with this network unless you also have an autonomous EV network. At first, this will not have a huge impact on existing rideshare companies, because the number of Tesla vehicles operating as robotaxis will be small. However, if Tesla continues to grow manufacturing at 50% year over year, it will only be a few years before there is enough vehicle supply to render all legacy mobility networks obsolete.
But even once another company creates a competing electric autonomous robotaxi network, it will be nearly impossible for the company to match the scale and efficiency of Tesla’s network. Tesla has the advantage of being able to centrally manage and optimize nearly every component of the network to support overall network health. For example, Tesla’s control of the charging network allows it to load balance demand using a variety of techniques, including pricing and charge percentage limits. It has already implemented defaults in some high-demand areas that stop charging at 80% of battery capacity. This enables significantly more throughput on a charger. While for any individual it may be better to charge to 100%, limiting charging to 80% for everyone is a win for almost everyone on the network, because it dramatically lowers wait times in peak situations. Today, this can be manually overridden, but it is an easy change to enforce this and offer different policies to different categories of fuel consumers.
On a network where different companies manufacture the vehicles and the charging networks, this type of optimization becomes much more challenging on a variety of levels. A multi-vendor network with different brands of cars, going to different charging networks, conceptually could be competitive over time, but pragmatically there is no way to make the proper optimization decisions when these decisions are in the hands of multiple parties. Thus, the latest “Tesla killer” EV on an automated Uber-like network using an Electrify America–like fast charging network is going to have a significantly higher cost structure, possibly approaching twice the cost. At this cost, it is nearly impossible to compete. Even at a consumer price that is 20% higher, it will be nearly impossible to compete become most consumers are extremely price sensitive on transportation costs. This same price sensitivity extends to commercial cargo.
Beyond this, Tesla’s competitors will likely deliver a significantly less compelling user experience because they do not have the vertical integration required to maintain a cohesive experience from the consumer app into the guts of the vehicle. When you use Tesla’s app to turn your heated seat on, it will consistently work. Tesla designs every line of software and every piece of hardware on the chain. When you have a network running multiple brands of EVs, with widely differing programmatic interfaces (APIs), it is much less likely to work consistently, and when it breaks, it may take a long time to create a fix involving many parties with differing priorities.
When you examine the potential scale of Tesla’s transportation network, some good comparables might be airlines and Uber. The total number of commercial airplanes in the world is about 25,000. These move about 4 billion people a year, or approximately 450 people per plane per day. Uber is servicing about half that number of people (2 billion a year) but at much shorter distances for shorter periods of time.
Tesla envisions manufacturing over 10 million vehicles a year by the end of the 2020s. If just 10 million of these were utilized in an automotive network, with each car moving 27 people a day, this would be about 100 billion trips a year, about 50 times the size of Uber. Beyond urban mobility, Tesla will be competing with trains, buses, and airlines in the 50–300 miles travel segment (a roughly $20B market today) for the huge number of people who drive this distance ($225B market today). In most cases, robotaxis will be both faster door to door and dramatically less expensive than the legacy alternatives, and will thus rapidly displace these existing networks.
As costs decrease, the size of the robotaxi market will likely continue to increase well beyond today’s rideshare and airline markets. Perhaps, its biggest obstacle will be the road traffic it creates that limits its own growth.
When considering how dominant Tesla may become in this market, we can look again to the analogy of Apple with regards to competition and profits.
Over the last several years, Apple has been able to garner the majority of the profits of the smartphone industry, often 80% to 90% (while selling a minority of phones). Tesla, however, may be in an even better position. If you consider the utility and economics of an iPhone vs. those of Samsung or other leading vendors, it would be hard to argue that the iPhone is somehow significantly more economically compelling, optimized, or technologically advanced. Tesla, on the other hand, has a much bigger lead and is innovating across dimensions that are much harder for others to copy. Others can attempt to recreate this vertical integration, but in order to beat Tesla economically, they will need to be innovating, optimizing, and scaling significantly faster than Tesla is in order to catch up on both operational efficiency and manufacturing economics.
However, the interesting analogy here might be to imagine if an iPhone plan only cost $40 a month, while an Android plan cost $70 a month. If this was the case, Apple would be in an enormously more leveraged position. Due to Tesla’s fundamental economic optimizations, this is the type of landscape that may soon emerge in the automotive industry. On one side you have Tesla, the first mover with significantly more scale, better core economics, stronger network effects, faster innovation cycles, centralized design, and surgical decision making; a company that is attracting better talent and reinvesting large amounts of profits as it increases its vertical integration, scale and efficiency. On the other side, you have an alphabet soup of carmakers, autonomy providers, service providers, and an array of technology venders that are unlikely to formulate and execute a cohesive plan. The potential for a long-running monopoly in the space akin to IBM mainframes or Google Search may be very realistic. It’s too early to know, and things can change fast, but as we sit here today, there is no comprehensive alternative vision being offered by the market. Until we see someone with a plan to create extremely efficient electric cars at scale, with low-cost autonomy serviced by a ubiquitous fast charging network that extends across the world, it will be hard to imagine what beating Tesla’s transportation network efficiency might look like.
Follow CleanTechnica on Google News.
It will make you happy & help you live in peace for the rest of your life.
Rahul Sonnad Rahul is cofounder & CEO of Tesloop. Based in Los Angeles, Tesloop offers a new one-way rental model for semi-autonomous Tesla vehicles, designed for travel to nearby cities. visit: tesloop.com contact: rahul@tesloop.com