Tesla robotaxi service becomes very widespread across the United States by end-2026

I think self-driving cars is essentially a solved problem at this point. And Tesla has rolled out robotaxi service in a few cities and will be very, very widespread by the end of this year within the U.S.
Elon Musk, 2026-01-22 — World Economic Forum, Davos

Restated 2026-05-18: We already have some vehicles operating with no people inside and no safety monitors in three cities in Texas and we probably will be widespread in the U.S. by the end of this year.

Sealed at 8%

Locked 2026-08-31. Confidence 74%. Uncertainty ±5 pts. Resolves 2026-12-31. Status: PENDING.

How the engine got there

This statement shares the technical foundation of ELN-03 but adds a scaling requirement, and scaling is where the time available becomes the problem. The starting point is known and small: the speaker’s own May restatement puts it at three Texas cities without safety monitors, and Texas is among the most permissive US states for autonomous deployment — the easiest possible first three. Reaching a genuinely national footprint by 31 December requires expansion across states with materially different rules; California alone requires separate DMV and CPUC authorisations, each with its own evidentiary process, and several states have no framework at all, which is slower to navigate than a strict one. Regulatory clearance is serial and per-jurisdiction and does not compress the way software does. One further signal carries weight: the same forecast was issued in January and restated unchanged in May, with the end-of-year date held constant while four months of the window elapsed. A deadline that does not move as its runway shortens is weak evidence about the world and strong evidence about the forecaster’s prior. None of this is a judgement on the driving — "essentially a solved problem" may prove right about capability. The seal is about metro count on a date.

What counts as correct

"Very widespread" is vague, so it is operationalised toward the generous reading: CORRECT if, on or before 2026-12-31, Tesla robotaxi service is available to the general public — not an invite-only tester cohort — in at least 25 US metropolitan areas. Twenty-five is chosen as approximately the threshold at which a US consumer service is ordinarily described as widespread rather than regional; a literal reading of "very, very widespread" would demand considerably more.

We are wrong if

at 2026-12-31 public Tesla robotaxi service operates in fewer than 25 US metros, including where the service is excellent in the metros it does cover.

Causal chain

Driverless operation was demonstrated in a permissive jurisdiction

Three Texas cities now run vehicles with no occupant and no monitor, under the most accommodating state rules available.

Existence proof established; the remaining question becomes replication elsewhere.

Each additional metro carries its own approval process

Authorisation is granted per state and often per municipality, with independent evidentiary standards and no reciprocity between them.

Expansion rate is governed by administrative throughput, not fleet size or software readiness.

The deadline was restated without movement

January and May statements both terminate at end-2026 despite four months elapsing between them, indicating the date is anchored to intent rather than to observed expansion rate.

The forecast carries less information about the world than its confidence implies.

Historical precedent

2020-2026 — Waymo US expansion

The closest available comparison: a working driverless service expanded to a handful of metros over roughly six years, gated almost entirely by approvals.

2016-2023 — Ride-hailing regulatory expansion

Even without autonomy, city-by-city regulatory entry took years per major market for a far less novel service.

2021-2022 — Cruise San Francisco scale-up

Rapid technical progress followed by regulatory pause, demonstrating that approval, not capability, sets the pace of the ramp.

Directly related seals

More on this desk

Browse the Technology & AI desk, the full projection registry, the calibration ledger, or the record by subject. Scoring is explained in how a prediction is made.