Companies that don’t make money eventually run out of it. Even AV companies.
Authors note: The following post is my response to the Ars Technica article “Self-driving startups are becoming an endangered species.” The Ars article observes, correctly, that quite a few AV companies have failed or been acquired over the last few years. But I think there’s a key underlying factor that explains what we’re seeing — and it’s a basic economic truth that applies to small and big AV companies alike.
Companies that don’t make money eventually run out of it.
The most important race in the AV space is the race to a scalable and sustainable business model. Who can actually make money by providing services using autonomous vehicles? Once you’re making money, you can begin to scale without spending investor dollars. And that scaling allows you to grow your brand, understand your riders and how you create value for them, and collect more data which allows you to improve your technology even faster. Having a viable business is the beginning of a positive feedback loop, leading to sustainable and profitable growth.
Obviously, one of the biggest levers to profitability is the autonomy technology itself: safety drivers are expensive! And so, many companies have focused on developing that technology as a prerequisite to building a business. In short: first make the technology work, then make the business work. But this model is risky. What if the technology works but the unit economics don’t — the vehicles are too expensive in comparison to the revenues the vehicles are able to generate? How do you know if you’re making the right technology investments if you’re not testing your ideas in the market?
Indeed, it’s hard to think of historically-successful businesses that started out by investing billions of dollars in R&D before earning any revenue or having any customers. In fact, the opposite appears to be true — even in very capital-intensive industries, those that prioritize commercial traction tend to win.
Being a space nerd, SpaceX comes to mind as an example of a company with fantastical R&D ambitions — but their approach was not to hole up and build a massive research organization. Instead, they aggressively pursued customers and revenue, using those experiences to accelerate the development of their product. While they’re famous for their reusable rockets, remember that they began with simple expendable ones that generated revenue.
Our company, May Mobility, has always been focused on building the business in parallel to building the technology, and we’ve reached a signal moment. With our launches this year, we have profitable sites. Our site operating costs have been steadily dropping since our first commercial operation in 2017, and our per-vehicle revenues have been gradually increasing at the same time. And now, those lines have crossed, despite the fact that our costs still include a safety driver. (Those safety drivers, incidentally, also help expand our market, by helping to educate riders and to build our city partner’s confidence.)
Profitable sites allow us to grow faster — we don’t have to subsidize our routes using expensive equity dollars. Every vehicle we operate teaches us more about riders, partners, and operations. And critically, it’s a positive feedback loop for our technology as well: we have more data for our machine learning systems, and more intervention data to help us focus our technical investments in the most important places. We also discover where operational requirements have significant implications on the design of the technology as well. For example: while extremely precise sensor calibration might be great in the lab, it is an operational nightmare. Operating vehicles in the real world quickly teaches you that relaxed calibration requirements are a necessity.
Operating real-world services is giving our teams a much better understanding of riders’ needs and cities’ goals. Arlington, Texas, for example, has an particularly innovative approach to providing transportation for its citizens — they have invested in on-demand services and have no traditional bus services at all. We are learning a lot from working with Arlington that we couldn’t otherwise.
Let’s return to the key point: companies that don’t make money eventually run out of it. Smaller companies run out a bit faster than bigger companies. Many smaller AV companies have run this course, leading to consolidations and acquihire-style mergers.
But even big companies eventually run out of capital. Uber ATG and Zoox have undergone consolidation, unable to raise the capital they needed to sustain their R&D on their own. And even the massive players — Waymo, Cruise, Argo — have moved away from a single-investor model to a more diversified investor base that allows each investor to limit their risk. To wit: when Waymo brought in $2.5B in external investment, Alphabet was sitting on over $110B in cash. Most of these companies have generated zero revenues, which is a significant business risk!
Focusing on profitable sites involves trade-offs. New and exciting (but expensive!) technologies — like better LIDARs, more powerful GPUs — would instantly expand our addressable market by allowing us to operate at higher speeds. But they’d also make the vehicles too expensive. And so we’ve optimized for routes that are a good fit for the vehicles that make economic sense today. But the technology is evolving quickly — we are ready to expand our markets as sensor costs come down.
Our strategy reflects our belief that the most important race in the autonomous vehicle industry is the race to a scalable, sustainable business model. And while autonomy itself is one of the most important levers, companies that can scale will have huge advantages in growing their business and gathering the data that fuels their technology. And when you’re making money, you’re a lot less likely to run out of it.
