Waymo leads Texas's robotaxi count, but data is the real story
Waymo runs 577 robotaxis in Texas to Tesla's 42, but the bigger lesson is what happens when a government forces a private industry's data into the open.
The new Texas autonomous vehicle registrations data is being read as a scoreboard: Waymo with 577 robotaxis, Tesla trailing at 42. That framing is fine for headlines, but it buries the part that actually matters to anyone who builds software. The interesting event here is not who is winning. It is that a state government switched a private industry from "trust us" to "show us," and then published the numbers for free.
I'm writing this for the Sri Lankan engineer or student who will never register a robotaxi but does need to find real datasets to learn on. This story is a small case study in why public, structured data is worth more than the cars it describes. (Reporting via TechCrunch.)
📊 What the Texas numbers actually say
A new Texas law took effect on May 28, 2026. It requires autonomous vehicle companies to register with the Department of Motor Vehicles and disclose fleet sizes and safety information. The state then launched a public tracker website with that data. Here is the robotaxi side of the ledger:
| Company | Registered AVs | Note |
|---|---|---|
| Waymo | 577 | Launched commercial service in Austin, March 2025 |
| Avride | 317 | Second by a wide margin |
| Nuro | 47 | Delivery-focused |
| Tesla | 42 | Robotaxi service launched in Austin, Summer 2025 |
| MOIA (Volkswagen) | 12 | Electric autonomous microbuses |
And the self-driving trucks, a quieter but serious category:
| Company | Self-driving trucks |
|---|---|
| Aurora | 91 |
| Gatik AI | 64 |
| Kodiak AI | 33 |
| Waabi | 13 |
Key takeaway: None of these numbers existed in public, comparable form a week ago. The law created the dataset. That is the real product launch.
🔍 Why fleet size is the wrong thing to measure
It is tempting to rank these companies by row order. Resist it. As the original report put it:
"The size of an autonomous vehicle-fleet only reveals so much about where a company stands on the leaderboard."
That line is doing a lot of work. A registered vehicle is not a vehicle carrying paying riders. Fleet count says nothing about:
- Miles driven per car — 42 hard-working cars can serve more trips than 577 parked ones.
- Disengagement and safety record — how often a human had to take over.
- Geographic spread — one dense city versus several.
- Revenue per ride — the only number that tells you if the business survives.
This is the same trap junior analysts fall into with any dataset: counting the rows you were handed instead of asking what the rows leave out. The honest move is to treat fleet size as one weak signal, not a verdict.
🌐 The reusable lesson: regulation can manufacture open data
Strip away the cars and this is a data-governance story. A regulator decided that a fast-moving industry could not keep its operating footprint private, mandated disclosure, and shipped a public website. That pattern repeats everywhere, including here.
Sri Lanka already publishes more open data than most builders use. Think about what is sitting in public PDFs and portals right now:
- Central Bank rate sheets and inflation tables.
- IRD tax brackets and gazette notifications.
- CEB tariff revisions and fuel price circulars.
- Election Commission and Department of Census results.
Most of this lands as a PDF, not a clean API. That is the gap a builder fills.
Bottom line: A dataset becomes valuable the moment someone makes it queryable. The Texas tracker did that for robotaxis. You can do the same for a Sri Lankan dataset nobody has cleaned up yet.
🛠️ A free-tier portfolio project hiding in this story
If you are a student building a CV, copy the shape of what Texas did. You do not need self-driving cars, a cloud budget, or anything that costs money.
- Pick a public Sri Lankan dataset that only exists as a PDF or a clunky table.
- Extract it into structured data — CSV or JSON. (If it's locked in a PDF, our free PDF to Word converter gets the text out so you can clean it up.)
- Build a small tracker — a single static page that loads the JSON and renders a sortable table and one chart.
- Deploy on a free tier — GitHub Pages, Cloudflare Pages, or Vercel's hobby plan. Zero rupees.
- Write up your method so a reader can trust the numbers and reproduce them.
That last step is what separates a portfolio piece from a screenshot. The Texas tracker is credible because the data has a clear legal source. Your project is credible for the same reason.
| Robotaxi industry | Your portfolio version |
|---|---|
| Law mandates disclosure | You choose an already-public source |
| DMV collects the data | You scrape or transcribe it once |
| State hosts a tracker | You deploy a static page, free |
| Public reads fleet sizes | Recruiters read your method and code |
The skills on display here — data cleaning, schema design, an honest chart, a written methodology — are exactly what a remote employer screens for. A robotaxi fleet cannot fit in your repo. This can.
💡 What this means for you
The Waymo-versus-Tesla scoreboard will be forgotten in a month. The durable lesson is that transparency is a feature you can build, and the raw material is often already public, just badly formatted.
- Don't rank anything by the easiest-to-count number. Ask what the dataset omits before you draw a conclusion.
- Treat messy public PDFs as opportunities, not obstacles. Cleaning one into structured data is a complete project.
- Cite your source and show your method. That is what made the Texas numbers worth reporting, and it is what will make your work worth hiring.
You will not out-spend Waymo. You can out-clarify it on a dataset that matters closer to home, for nothing but your time.