Pony.ai Has About 1,975 Robotaxis Running. Tesla Has About 14.

That's the comparison that should frame the September 3 event, and it isn't close.

As of June 30, Pony.ai's robotaxi fleet stood at 1,975 vehicles, scaling toward 3,500 by year end. Tesla, according to independent tracking, has roughly 42 cars in Austin, an active driverless fleet nearer 20, and only about 14 vehicles operating unsupervised across all its markets — down from a peak of 25. The majority of the Austin fleet still carries safety monitors.

Tesla has driven more than 380,000 unsupervised miles across six cities in two states. Waymo does that kind of volume in a good week.

So yes — Tesla is arriving late. But "late" undersells it. The gap isn't a product cycle. It's roughly two orders of magnitude in deployed autonomous vehicles.

One correction, and it matters

The Waymo cost figures need fixing, because the whole cost argument rests on them.

The Ojai did not take per-vehicle cost from $150,000 to $50,000. Morgan Stanley estimates the Ojai at roughly $125,000 per unit, against about $200,000 for the Jaguar I-Pace it replaces. The base Zeekr vehicle is around $38,000; Waymo's autonomous hardware adds under $20,000; the fully equipped, service-ready vehicle lands near $125,000.

That's still a serious cut — roughly 37% — and it's the right story. It just isn't a 3× reduction to $50,000, and a mobility analyst will catch that instantly.

Where the field actually stands

Status

Zoox First-ever NHTSA commercial exemption for a purpose-built robotaxi, granted July 30. Covers eight federal safety standards, up to 2,500 vehicles annually for two years, commercial service in Las Vegas. Conditions: crash reporting, US-based remote operators, published service maps, and no sales to private buyers.

Waymo Ojai open to all riders in SF, LA and Phoenix, expanding toward San Diego, Las Vegas and Denver. Built with Zeekr / Geely — in China.

Pony.ai 1,975 vehicles operating as of June 30; 4,000+ overseas units contracted but not yet deployed, including 2,000+ with Uber for Europe. Rollout gated on permits.

Tesla ~14 unsupervised vehicles across six cities. Production Cybercab unveiling September 3, Austin. Invite-only, livestreamed, guests may not be content creators.

Note the Pony.ai distinction, because it cuts both ways: 4,000+ is a pipeline, contracted and awaiting permits — not cars on roads. The genuinely impressive number is the ~1,975 already running.

Now the part I'd push back on

The manufacturing-scale argument is the strongest case for Tesla, and I think it's also the one that needs the hardest look.

Manufacturing has never been the bottleneck in this industry.

Nobody is vehicle-supply-constrained. Waymo isn't limited by how many Jaguars or Zeekrs it can buy — it's limited by per-city validation, regulatory approval, remote-operations staffing and depot infrastructure. Zoox's cap of 2,500 vehicles a year is a regulatory ceiling, not a factory one.

And here's the uncomfortable proof: Tesla can already build millions of cars a year, and it is running fourteen unsupervised robotaxis. If manufacturing were the binding constraint, Tesla would be winning already. It plainly isn't the constraint, and Tesla's own numbers are the evidence.

Manufacturing scale is a real advantage — eventually. It's the advantage that matters at vehicle number 50,000. It does nothing for you at vehicle number 15, and vehicle 15 is where Tesla is.

What Zoox won is harder than what Tesla is unveiling

A vehicle with no steering wheel and no pedals cannot legally operate at scale in the US without an exemption from federal motor vehicle safety standards. Zoox spent from September 2025 to July 2026 obtaining one, and it covers eight separate standards.

Unveiling a car without controls is a design milestone. Getting permission to charge people for rides in it is a regulatory one — and those are separated by years, not weeks.

That's the specific thing to listen for on the 3rd. Not the design. Not the FSD demo. The regulatory pathway — under what authority does a Cybercab carry paying passengers with no controls, in which jurisdictions, and by when.

One wrinkle that genuinely favours Tesla

The Ojai's cost reduction depends on Zeekr — a Chinese OEM. In the current tariff and policy environment, a US robotaxi operator whose unit economics rest on Chinese-built vehicles carries a risk Tesla simply doesn't have.

So the manufacturing argument isn't wrong. It's just narrower than stated: Tesla's advantage isn't that it can build more cars, it's that it can build them domestically, at cost, without a supply chain a single executive order could reprice.

My read

"A concept car in 2024 was a promise; a production vehicle in 2026 has to be a plan" is exactly the right frame. I'd add one line: it has to be a plan whose first constraint is a regulator, not a factory.

And the small detail about excluding content creators from guest slots is worth noticing. A company confident in a live demonstration usually wants more cameras, not fewer.

Question for you: if manufacturing scale is Tesla's advantage but regulatory approval and per-city validation are the actual bottleneck — is Tesla's edge worth anything before 2028? I'd argue the Cybercab is a bet on a constraint that hasn't started binding yet, which is either visionary or three years early.

Address

Boston, Massachusets

Contacts

news@techproduit.com

Get your weekly newspaper