Tesla sells humanoid robots to the general public by the end of 2027
Probably later this year, by the end of this year, I think they’ll be doing more complex tasks and still deployed in an industrial environment. By the end of next year, I think we’ll be selling humanoid robots to the public. That’s when we’re confident that it’s very high reliability, very high safety, and the range of functionality is also very high. You can basically ask it to do anything you’d like.
Elon Musk, 2026-01-22 — World Economic Forum, Davos — in conversation with Larry Fink
Report of the quote
Davos session video
C-SPAN recording
Sealed at 26%
Locked 2026-08-31. Confidence 58%. Uncertainty ±13 pts. Resolves 2027-12-31. Status: PENDING.
How the engine got there
Our sealed TEC-05 / CAL-14 sits at 55% for the industry shipping 50,000 humanoid units worldwide in a single year — a statement about volume across all manufacturers. This statement is narrower and, in one respect, easier: consumer availability does not require scale. A few hundred units delivered to retail buyers would satisfy it, so manufacturing throughput is not the binding constraint. The binding constraint is the one the speaker names himself inside the quote — "that’s when we’re confident that it’s very high reliability, very high safety". A bipedal machine of this mass operating in unstructured homes, around children and pets, is a product-liability instrument as much as a robot, and safety certification for an entirely new consumer device class has no fast path through any major regulator. The gate is also self-imposed and elastic, which historically means it moves. The base rate is the decisive input: Optimus was announced in 2021 with production repeatedly forecast one to two years out, and each of those dates passed with the programme real and advancing but the timeline reset. That is a pattern of genuine engineering progress combined with persistent temporal optimism, and it is the most predictive single feature available here. This seal encodes a disagreement about sixteen months, not about the product — we would price Tesla selling a humanoid to consumers by 2032 well above 70%.
What counts as correct
CORRECT if, on or before 2027-12-31, Tesla makes a humanoid robot available for purchase by non-enterprise members of the general public in at least one jurisdiction, with orders accepted and at least one unit delivered to a retail buyer. Deliberately steelmanned: a limited release, a waitlist that actually ships, or a very high price do NOT prevent this resolving CORRECT. Deployment confined to Tesla facilities, enterprise partners or employees does not count — the statement says "to the public".
We are wrong if
at 2027-12-31 no member of the general public can order and take delivery of a Tesla humanoid robot, including where the programme is live but restricted to industrial or enterprise customers.
Causal chain
Humanoid hardware crossed from demonstration into low-rate production
Sub-$20K platforms and real 2025 shipment tallies (~18,000 units, IDC) established that the category manufactures, which is what underwrites our 55% industry seal.
Hardware readiness stops being the reason a consumer launch would slip.
Consumer sale imports an entirely different regulatory regime
Industrial deployment operates under workplace safety rules with trained operators and controlled environments. Consumer sale requires product safety certification, liability cover and post-market surveillance for an unprecedented device class.
The launch date is set by certification and insurance, neither of which compresses to engineering timescales.
The reliability gate is self-set
The quote conditions the launch on internal confidence in "very high reliability, very high safety" — a threshold the company defines and can move without any external event.
Historical slippage on exactly this programme becomes the dominant prior for the date.
Historical precedent
2021-2026 — Optimus programme timelines
Announced 2021 with production forecast one to two years out, restated repeatedly. Direct, same-programme evidence of real progress on a persistently resetting clock.
2002-2010 — Consumer robotics: Roomba to humanoid expectation
A genuinely successful consumer robot arrived, but as a single-function device. General-purpose home robots forecast to follow within the decade did not.
2014-2020 — Honda ASIMO retirement
Two decades of capable humanoid demonstrations never reached a consumer product, and the programme was wound down. Demonstration capability and shippable consumer product are separated by more than iteration.
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.