#19: Waymo the Leapfrog - by John Loeber
When people talk about self-driving cars, they usually talk about them as though they’re in the future, still a little opaque and unproven. But as so often, the future is already here: it’s just not evenly distributed.
For the last few years, Waymo has been operating fully driverless cars in Phoenix and San Francisco. They’re commercially available. Anyone can download the app and hail one: I was recently passing through SF, and took the opportunity to make a few trips by Waymo. I found the experience hugely impressive, to the point that it made very clear what some parts of our future will look like.
In this essay, I will talk about my experience riding Waymos, predict the impacts that self-driving vehicles will have on our society, sketch out Waymo’s unit economics and discuss its competitive positioning vis-a-vis Uber and others, and finally argue that Waymos will totally overhaul how we think of public transit — offering a rare technological “leapfrog” opportunity to urban America.
I hadn’t thought much about how it would be to ride in a self-driving car. Surely it’d be like taking an Uber, but without a driver — functionally, it’s all the same, right?
It’s actually totally different. And it’s wildly superior.
The feeling of the drive is much better. Waymos accelerate very gently, and drive in a slow, defensive way such that they will never suddenly speed up or hit the brakes. In a city with lots of hills and stop-and-go traffic like SF, this is a godsend. If I were stuck with an impatient Uber driver who floors gas and brake, I’d feel nauseous after a few minutes. On the other hand, the gentleness of the Waymo means that I can read for the entire drive, and not even notice any traffic.
The Waymo is your own personal space. It’s really nice to have it entirely to yourself. That makes it much more relaxing than taking an Uber. When you share a small, confined space (like a car) with a stranger, it causes a very subtle stress: you’re paying attention to what they’re doing, and you think about your own actions so as not to bother them. It’s slightly taxing. Waymo obviates this.
This is really noteworthy, because it means that even having a great driver is worse than having no driver in this respect. One’s first-glance intuition might be that driverless is better than a bad driver and worse than a great driver, but this is not so. Driverless is better in both cases.
A Waymo is clearly a much safer driver than a regular human. Waymos actually keep their legal minimum distance of three feet to bicycles on the road! As a cyclist, I’ve had trucks speed past me at over sixty miles an hour with maybe six inches of clearance. I’d feel much safer sharing the roads with Waymos that are actually programmed to follow the law.
In short: the Waymo experience is great. It’s hard to imagine how good it is if you’ve never been inside one, because you’ve been used to a different paradigm your entire life. Taking a Waymo is relaxing in a way that taking an Uber (let alone driving yourself and having to actually navigate traffic) simply never is. Even if a Waymo travels a little slower than a human driver, the ride is so gentle that you’re better off on net. It turns into quiet personal time or reading time, instead of being waiting time.
There’s a saying in surfing: slow is smooth. Smooth is fast. Google has played it right: fifteen years of slow-is-smooth work on Waymo, delivering an impeccable safety record and a magical experience. They are far ahead of any competition, ideally positioned to roll-out nationally as quickly as regulators will allow.
I’d expect for large fleets of Waymos to become commercially available in major US cities over the next few years. Approval for roll-out will become easier and easier as the cumulative safety record speaks for itself.
Waymo is still a rare luxury today. The SF fleet has only about 250 cars, of which 100 are on the road at any time. This makes Waymo slightly more expensive than Uber. As more Waymo vehicles become available, you might expect Waymo to undercut Uber on pricing, and take the rest as margin. But I think Waymo will race to drive down its pricing much further. Why? Because Uber is not nearly as big as the terminal market size. For comparison:
In other words, there are ~58x more personal rides than rideshare rides in the US. The true TAM is that Waymo replaces all travel by car (and perhaps even all travel by bus, subway, etc.). To get there, Waymo has to bring its price down to the point that anyone with a car would pay to not have to drive themselves. Maximizing business value here is a volume game: the winning recipe is massively scaled deployment, with prices lower than any competition, and taking the thin margin.
Not having a driver significantly lowers OpEx. Take a look at Uber:
Uber claims that its drivers get roughly 70% of the ride fare whereas drivers claim they get about 50% of the ride fare;
Uber’s cost of revenue is about 16% of gross bookings, and their cost of operations and support is about 1.8% of of gross bookings;
Clearly there is an opportunity to reduce the marginal cost of operating the ride by between 50 and 70%, perhaps more if the removal of driver operations, payouts, etc. further reduces costs. But Waymo’s safety record makes for even larger savings. Fewer crashes mean:
Less need for expensive repairs;
Lower insurance rates;
Longer vehicular lifespans over which they depreciate.
Long-term, you’d expect the price of a Waymo ride to be the sum of:
The cost of gas (or electricity);
The per-mile depreciation of the vehicle;
The marginal cost of any supporting software or operations (e.g. cleaning);
Whatever the average consumer is willing to pay to not drive themselves, which might be a few dollars an hour, or a little more if the experience is really great and they win back some of the opportunity cost of driving.
I think a decent first-order approximation, before factoring in any next-generation technology advances or economies of scale, is that Waymo could bring down operating costs 70% relative to Uber. The world looks very different when a $40 half-hour Uber ride suddenly becomes $13 — that’s a lot of increased mobility.
Waymo vehicles are not cheap! Each one costs about $200K to deploy. Recapturing this CapEx is a tall order, and it makes Waymo’s prospects less clear.
Suppose a Waymo vehicle runs a gross profit of $10 per hour, and it drives for 20 hours a day on average. That’s $73,000 a year in gross profits. Even if the cost of deployment comes down to $150K, it takes 2.1 years to recapture the CapEx. If it drives an average of 25 miles per hour, then it’s putting on 182,500 miles per year, or about 383,000 miles by the time the CapEx is recouped. That’s a lot of miles! I don’t know what the life expectancy of the vehicle is after 383K miles, but it’s probably not great. Further, if any of the expensive self-driving hardware (Lidar, sensors, etc.) has to be replaced on any regular schedule, it’s possible that an individual Waymo would take many years to recapture its upfront cost, if it can do so at all.
This suggests that Waymo would have to make very big, long-term bets, risking significant CapEx with uncertain payoff profiles. Waymo might be severely unprofitable at significant scale before it arrives at viable unit economics. Importantly, Waymo kind of has to do this, because it must (1) use its current market-leader advantage to build distribution while competition is scant and (2) scale up in order to have economies of scale that can bring the costs down.
But never forget: Waymo is owned by Google, and making very big, very long-term CapEx bets is the Google style. This is the true nature of the moat in Google’s search engine: Google has invested heavily for twenty-five years in building web crawling infrastructure. If you wanted to compete with Google on its turf today, and create a web crawler and search engine on par with Google’s, the level of capital expenditure would render it basically impossible: it would take billions of dollars over many years of engineering just to get to par with Google. Then you’d have to spend tens of billions of dollars trying to take market share away from Google, and only then would you be in a position where you can fight a price war on ads against Google, at the end of which neither of you will make any profits. It’d be economic insanity to try. So nobody does it, and Google’s search engine enjoys a natural monopoly and prints money.
The reason to tell that story is that there’s an analogous case for Waymo. Google could perfectly rationally deploy many billions of dollars into Waymo to obtain:
Proprietary vehicles that are a superior form factor, both in terms of rider experience and unit economics (i.e. minimizing maintenance costs);
Massive scale, creating superior ride liquidity — i.e. when a customer requests a ride, they are able to obtain one faster than with any competing service;
Superior economies of scale in purchasing, outfitting, and maintaining vehicles;
Superior technology and data, creating a better and safer drive experience;
A superior safety record translating into regulatory capture.
Google has plenty of money: $108B in pure cash on hand, and $69B in free cash flows for 2023. Self-driving cars are among the few markets with TAMs big enough to move the needle for Google, so it would make sense to burn a few billion dollars a year on Waymo for the next decade, and to start capturing profits only once the barriers to entry are too tall for competitors and Waymo is in a position of natural monopoly.
There are two important notes here:
To believe there can be a natural monopoly, you must believe this is a winner-take-all or winner-take-most market. We’ve had a great live experiment with that over the past fifteen years: even though there is little differentiation among rideshare services, it has been very hard for new ones to compete. Ten years ago, I expected many vendors to emerge, but the reality has been that acquiring customers is so expensive and the margins are so thin that Uber has been able to take over the market, with small (shrinking) slices for Lyft and for local alternatives (primarily outside the US). By analogy, this suggests that the market for self-driving vehicles will have winner-take-most characteristics.
Google has already shown tremendous long-term thinking: they have been investing in Waymo for fifteen years now. Consider the level of conviction required to keep up this level of investment for so long, especially in the early years when deployment seemed ever-so-far away. Now that the goal is in sight, massive spending is more rational than ever.
Uber might go down as a strange story in business history. They’ve burned $31.4B in operating losses since 2014 to win this market. For many years, Uber (Travis Kalanick) recognized that human drivers were a transitional step, and that autonomous vehicles would eventually gobble up the space. TK had the right instinct to focus on Uber ATG starting at 2015, when it was still possible to catch-up. It’s not anymore. I expect that as soon as Waymo deploys a thousand cars in a big taxi market like LA, Vegas, or NYC, the public markets will put an expiration date on Uber.
Startups are a tough game, and the path from disruptor to disrupted can be surprisingly short. Uber was founded in 2009. If Waymo disrupts Uber by 2029, then it will have been a twenty-year-journey of posting huge losses to seize a market, scoring a few years of comparatively small operating profits, and that’s it. Over a twenty-year journey, you really do have to keep innovating — in fact, it’s so long that you may be forced to bet on, and pivot, the very definition of your business. This will have been the case for Uber. TK and team saw this; I’m not sure if their successors did.
The once-dreaded long commute is about to come back in a big and pleasant way. I would have no issue at all sitting in a Waymo for 45 minutes each way every day. It’s just a nice time to myself that I can use to nap, work, or read.
This means that the suburbs and exurbs stand to benefit meaningfully. Working in a high-cost-of-living city and then commuting from a lower-cost-of-living suburb is a common strategy, but it comes at a certain lifestyle tax: you lose a lot of time on the commute, and you have to plan around availability of transport. Inexpensive, widely available Waymos fix this.
This brings us to public transportation: so far in this piece, I have chiefly compared Waymo to Uber. But Waymo is most compelling not as an alternative to a taxi or Uber, but as an alternative to driving yourself — or to being on public transit.
Waymo becomes most interesting as an alternative to public transit. It favorably changes the economies of scale: more riders per vehicle imply better economics, and deals with local governments can stabilize Waymo financially with long-term contracts.
Waymo as public transit is particularly attractive, because virtually all cities in the US have wound up in a position where building public transportation infrastructure is well-nigh impossible. In San Francisco, it cost $346M over 6 years to install a new set of north-south bus lanes on a straight stretch of two-mile road. In NYC, a new subway line is costing $2.5B per mile. Trying to “build” anything, even if it’s just sectioning off an existing part of a road with some paint, invites months or years of local political debate, environmental litigation, and runaway costs far beyond any reasonable imagination.
But we have tons of roads. And thankfully, self-driving cars do not require any more infrastructure. This sidesteps the quagmire of trying to build new public transit infrastructure, and allows us to provide great public transit in the form of driverless cars/small buses without all the administrative overhead. It’s an easy winner:
The fact that these cars can be called by app to any destination and routed efficiently by algorithm means that they offer huge accessibility and environmental advantages over public transit that always follows a fixed route and schedule.
Many people who currently drive themselves would probably be happy to carpool in a self-driving vehicle if it’s reliable and easy, which would improve congestion.
Unlike all other public transit, maintaining the infrastructure for self-driving vehicles (simple roads) is relatively easy and inexpensive.
It is ironic that for so long, every lane of road for cars was seen as zero-sum competition for public transit, and now we might achieve broad, high-quality public transit precisely via all the road space that we’ve already made available.
In fact, this might bring American public transportation to a leapfrog moment. Many pundits have lamented that developing cities elsewhere have “leapfrogged” the US on public transportation — building subways and rail networks that put ours to shame. Over a hundred years ago, we built first-generation public transit. Over the last forty years, other countries built second-generation public transit. Now we have the opportunity as a nation to lead the world on third-generation public transit, and in that course develop products and expertise that can be exported.
More granularly, cities all over the US are about to have a fantastic opportunity to redirect budget from artificially expensive transit infrastructure projects toward driverless cars and small buses as next-generation public transit. A billion dollars doesn’t buy you a lot of subway stops these days, but it’ll probably buy you a terrific long-term, at-cost contract for the future of transportation.
BEST-PRACTICE DISCLOSURES
I don’t have any financial positions in Uber or Google other than indirect investments in their common stock via low-cost index funds.
ncG1vNJzZmikn5qvpr6NrKybq6SWsKx6wqikaKhfZ31uw8CypKhlpJ2ybrjEmqefqp%2Bc