What if the most important feature in a robotaxi has nothing to do with the car?
I review AI tools for a living. I poke at APIs, break things on purpose, and write down what actually happens versus what the marketing page promised. So my instinct with self-driving is to ask about the model, the sensor fusion, the disengagement numbers, the edge cases at night in the rain. Those are the things you can test.
Then you read that Waymo has doubled its spending on lobbying as its fight with Uber sharpens, and you have to sit with an uncomfortable idea: some of the most consequential engineering in this category is happening in legislative committee rooms, not in a garage.
Following the money tells a different story
Look at the recent run of headlines together. Waymo finalized a $16bn funding round at a $110bn valuation. Uber pledged $10bn to win the robotaxi race. Waymo is reportedly exploring a split with Uber as tensions between them deepen. Volkswagen is pushing into robotaxis to challenge Waymo and Tesla. And Waymo doubled its lobbying spend.
Notice what’s missing from that list. Not a single one of those items is about the driving getting better. They’re about capital, partnerships, distribution, and rules. The technology is treated as a settled input. The fight is over everything wrapped around it.
That pattern is familiar to anyone who has watched a tool category mature. Early on, reviews are about capability, because capability varies wildly between products. Later, the products converge and the reviews become about pricing, lock-in, terms of service, and whether the vendor will still exist in three years. Robotaxis appear to have crossed that line while most of us were still arguing about lidar.
Why a Waymo and Uber split matters more than it sounds
The reported exploration of a split between Waymo and Uber is the part I keep turning over. Those two were partners in a distribution sense: one had the autonomy stack, the other had the app that hundreds of millions of people already have installed. That arrangement papered over an obvious conflict. If Uber is committing $10bn to win the robotaxi race itself, it isn’t a neutral channel anymore. It’s a competitor that happens to own the storefront.
Anyone who has built on a platform knows how this movie goes. You integrate with a partner because their reach is worth the dependency. Then the partner decides your layer is where the margin lives, and suddenly your integration is a liability. The specifics differ, but the shape is the same whether it’s a ride-hailing app or an AI API you wired into production last quarter.
The practical lesson for anyone evaluating tools: partnership announcements are not architecture. They’re temporary alignments of interest, and they unwind when the interests do.
Lobbying as a product feature
Doubling lobbying spend isn’t a footnote in a quarterly report. For a service that can only operate where regulators permit it, permission is the product. A vehicle that drives beautifully in a city where it isn’t allowed to pick up passengers has a utilization rate of zero. Regulatory access is the distribution layer, and Waymo is clearly treating it that way.
I find this genuinely hard to score as a reviewer. If I’m assessing an AI tool, I can measure latency, accuracy, cost per call, and how badly it fails when I feed it garbage. But that second thing may determine whether the tool is available to me at all.
Volkswagen entering the fight against Waymo and Tesla adds another layer. A legacy automaker brings manufacturing scale and existing relationships with governments in a way a software company has to buy. Different companies are optimizing different constraints, and none of them are purely technical.
What this means if you’re building on top of any of it
A few things I’d hold onto:
- Valuations and pledged capital tell you who can survive a long fight, not who has the better system. A $110bn valuation and a $10bn commitment are both statements about endurance.
- When your partner starts funding a competing product, treat the relationship as time-limited and plan the exit before you need it.
- Availability is a feature. A tool that works everywhere beats a better tool that works in three cities.
- Watch where the spending shifts. Money moving from engineering into government relations is a signal that the company thinks the hard part has changed.
None of this makes the underlying autonomy work less impressive. It just means the scorecard I’d normally use is incomplete. The best robotaxi may not win. The one with the most permission, the most capital, and the fewest dependencies on rivals probably will.
That’s a less satisfying conclusion than a benchmark chart. It’s also closer to how this actually gets decided.
🕒 Published: