Robotaxis run on paperwork too.
That’s the part nobody puts in the promo video. When you watch a car glide through an intersection with nobody in the driver’s seat, you’re watching software make decisions. But you’re also watching the outcome of city council votes, state regulations, permit approvals, and a lot of conversations in rooms with no cameras. According to the Financial Times, Waymo has doubled its lobbying spending as its rivalry with Uber over robotaxis intensifies. That single detail tells you more about how AI actually reaches the public than most technical explainers do.
I write about AI agents for people who don’t build them, and this is one of my favorite teaching moments. An AI agent is software that perceives its environment, decides what to do, and acts on that decision without a human approving each step. A self-driving car is the most physical, most obvious version of that idea. It’s an agent with wheels. And what’s happening right now with Waymo and Uber shows that getting an agent into the real world is only partly an engineering problem.
What the money is telling us
Stack up the recent reporting and a picture forms. The FT reports Waymo finalised a $16bn funding round at a $110bn valuation. Uber pledged $10bn to win the robotaxi race, described as a strategy shift. And Waymo is reportedly exploring a split with Uber as tensions between them deepen. Alongside all of that, Waymo’s lobbying budget doubled.
Those numbers aren’t just competitive posturing. They’re a rough map of what it costs to deploy an autonomous agent at scale:
- Capital to build and operate the fleet
- Capital to buy your way into the race if you started late
- Political spending to shape the rules the fleet operates under
- Partnership decisions that determine who controls the customer
Only the first item is about technology. The rest is about permission and position.
Why the partnership strain matters more than it sounds
Waymo and Uber have been partners. Waymo has the autonomous driving system; Uber has the app hundreds of millions of people already have on their phones. On paper that’s a tidy division of labor. In practice, it’s a question of who owns the relationship with the rider.
Now Uber has committed $10bn of its own to robotaxis, and Waymo is reportedly weighing a split. This is a pattern worth recognizing because it repeats across AI, not just in transportation. Whoever controls the interface — the app, the chat window, the assistant on your phone — has enormous power over which agent you actually use. And whoever builds the agent would rather not be a supplier to someone else’s storefront forever.
If you use AI tools at work, you’ve probably already felt a mild version of this. The model that answers your questions and the product you open to ask are often made by different companies with different plans for you. Those arrangements shift. Robotaxis are the same tension with a much bigger price tag attached.
The lobbying detail is the real story
Doubling lobbying spend is a signal about where the hard problems are. If your remaining obstacles were purely technical, you’d double your engineering budget. Doubling political spending suggests the binding constraint is regulatory: which cities allow driverless operation, under what conditions, with what reporting requirements, and what happens legally when something goes wrong.
For non-technical readers, that reframes how you should evaluate news about autonomous systems. The question isn’t only “does it work?” It’s also “who decided it was allowed to work here, and what did they ask for in return?” Those decisions happen in public meetings and legislative sessions, which means they’re one of the few parts of the AI buildout that ordinary people can actually influence.
What I’d watch next
A few things would tell us how this shakes out. Whether Waymo and Uber formally separate, and how riders find a robotaxi if they do. Whether Uber’s $10bn goes toward building its own autonomy stack or buying partnerships with others. And whether the increased lobbying produces visible policy changes in specific cities, because that’s where the effect becomes measurable rather than theoretical.
The lesson I keep coming back to is simple. We talk about AI agents as if their arrival is a matter of capability — once the software is good enough, the future shows up. The robotaxi fight suggests otherwise. Capability gets you to the starting line. Money, permission, and control of the customer relationship determine who actually crosses it. When a company doubles what it spends on lobbyists, it’s telling you plainly which race it thinks it’s running.
🕒 Published: