NORBERT BIEDRZYCKI

Autonomous Driving: The Billion-Dollar Bet on an AI-Powered Future

Autonomous Driving: The Billion-Dollar Bet on an AI-Powered Future

Autonomous driving is the most expensive experiment in the history of applied AI. The question is who can afford to wait for the result.

There is no better illustration of the promise and the danger of AI investment than the self-driving car.

The thesis is compelling, and I understand why so much capital has been drawn to it. Transportation is one of the largest markets in the world. If software can replace the driver, it will change the economics of taxis, logistics, insurance, urban planning and car ownership itself. Whoever controls that software layer could control a share of value comparable to what operating systems captured in personal computing.

2020 has given the believers some genuine milestones. In October, Waymo opened fully driverless rides to the public in parts of Phoenix, with no safety driver behind the wheel. Amazon bought Zoox for a reported $1.2 billion. Lidar companies such as Luminar and Velodyne went public through SPACs, with Luminar’s founder becoming one of the youngest self-made billionaires on paper.

And yet, in the same year, the cracks became visible.

The timing problem

A few years ago, several industry leaders promised fully autonomous vehicles on public roads by 2020 or 2021. Those dates have come and gone. This month Uber agreed to sell its self-driving unit, Advanced Technologies Group, to Aurora, and invested $400 million into the combined company in the process. In plain language: one of the most aggressive players in the race decided that it could not carry the cost of the journey on its own balance sheet.

This is the core investment dilemma. Autonomous driving is not a technology problem that can be solved in a typical five to seven year fund cycle. It requires billions of dollars, millions of test miles, regulatory approval in every jurisdiction and public trust that can be lost after a single accident. The tail risk is not that the technology fails. It is that it succeeds too late for the people who financed it.

Three ways to play it

From an investor’s perspective I see three distinct strategies.

The first is to bet on the full stack: a company building the entire driving system. The potential reward is enormous, but so is the capital requirement, and the realistic owners of that stack are increasingly large technology companies and car manufacturers with deep pockets.

The second is to invest in components: sensors, lidar, mapping, simulation, compute. These businesses can sell to many players regardless of who wins. However, their valuations already assume a market that does not yet exist at scale, which is exactly what many of this year’s SPACs are pricing in.

The third is to invest in narrower applications where autonomy is commercially useful today: mining sites, ports, warehouses, agricultural machinery, highway trucking corridors. These environments are more controlled, the regulatory path is shorter and the customer pays for productivity now rather than for a vision of the future.

If I had to allocate capital this year, I would be much more comfortable with the third category than the first. Not because the robotaxi will never happen. Because the investor’s job is to be paid within a reasonable time for the risk they take.

What boards should be asking

For the car manufacturers themselves the question is different and more existential. If value migrates from hardware to software, what is a carmaker worth without the software? Several have chosen partnership or minority investment rather than building alone. That is rational, but it may also mean surrendering the most valuable layer of the future vehicle to someone else.

The lesson beyond cars

Autonomous driving is teaching investors a broader lesson about AI. The distance between an impressive demonstration and a reliable commercial product can be very long. In safety critical domains, the last few percent of reliability can cost more than everything that came before.

Investment question: How should investors value a technology with extraordinary long-term potential but uncertain commercial timing?

Read next

  1. The Physical AI Opportunity: Investing Beyond Software

    The Business of AI series, written in 2026

  2. OpenAI and the consultants: the last mile of AI is organizationalLinkedIn

    When OpenAI announced partnerships with McKinsey, BCG, Accenture and Capgemini, it didn't feel like a breakthrough moment. At first glance, it looked like another predictable move: big tech teaming up with big consulting.

  3. Solow's déjà vu: AI everywhere except in the productivity statisticsLinkedIn

    I've been participating in technology revolutions since the early days of my career, and in AI since the moment it became deployable at scale in real commercial environments. And yet, I increasingly feel a sense of déjà vu.