Tesla Optimus
Where it stands: Tesla has begun initial mass production of its Optimus humanoid at Fremont, on the former Model S and Model X lines, aiming for 20,000 robots a week by late 2027 and outside deliveries around the end of that year, though no outside customer has been named. Musk says the hands, durability and AI still need work, and China's magnet export restrictions have set production back by months. Watch for a promised demonstration video and the first named outside customer.
Latest
- 2026-11-01(Shattered.io — shattered.io): the Fremont conversion gets its duration — the Model S line teardown took 46 days to make room for Optimus production, with the reporting re-running the 1-million-robots-a-year capacity target. Development of the Q3 earnings-run record that Optimus mass production began at Fremont on the dismantled Model S/X lines. Watch: actual output off the rebuilt line against the 20,000-a-week late-2027 target and the million-a-year plan, and the first external customer.
AI image - 2026-10-31(Tesla North — teslanorth.com/2026): development — Musk answered "No" on X to a question about Optimus ordering opening in December, shooting down the October 30 report that Tesla planned to open customer reservations that month (the underlying report came from The San Francisco Standard citing an unnamed source, with no Tesla confirmation in its own reporting). The denial answers the reservation-rumor watch directly. Watch: whether a reservation program surfaces at all, deposit amounts and terms, pricing beside the $20,000–30,000 consumer target, and the late-2027 outside-customer timeline holding. tags: tesla, humanoids
AI image - 2026-10-30 (San Francisco Standard report carried by Basenor — basenor.com/blogs): Tesla is reportedly planning to open Optimus customer reservations in December 2026 — a first named month for a step in which buyers outside Tesla's own facilities can commit to the robot, against the consumer-sales-targeted-2027 line and the end-2027 external-delivery figure from the Q3 call. Single-outlet report with attributed sourcing; no Tesla confirmation in the reporting. Watch: a Tesla confirmation or denial of a December reservation opening, deposit amounts and terms, and any named pricing or capacity beside the $20,000–30,000 target.
Images in this thread
Claims to watch · 9 (7 pending, 0 slipped, 1 hit, 1 missed)
- Elon Musk, Mar 11, 2026: Digital Optimus / Macrohard "user experience ready" September 2026 due overdue Source
- Elon Musk, Sep 16, 2026: humanoids outnumber humans within 10 years review by pending Source
- Tesla Optimus, Oct 22, 2026: 20,000 robots/week around late 2027; external customers by late 2027 review by pending Source
- Tesla, Sep 2, 2026: commercial Optimus sales at $20,000–30,000 target beginning 2027 due pending Source
- Emad Mostaque, Oct 8, 2026: Tesla will sell more robots than cars "within two years" review by pending Source
- Elon Musk, Sep 27, 2025: Optimus annual production run-rate of 1 million by 2030 due pending Source
- Elon Musk, Sep 24, 2026: 1 billion robots "will outproduce humanity" in 10 years review by pending Source
- Elon Musk, Jan 29, 2025: Optimus: ~10,000 units produced in 2025 due missed Source
- Elon Musk: Optimus Gen 3 production start "before end of 2026" due hit Source
October 2026 · 23 updates
- 2026-10-29 (Yahoo Finance — finance.yahoo.com/markets): development — Musk dismisses an estimate that humanoid robots will use about 0.04% of 2026's DRAM output, saying robot memory needs "will exceed everything else on Earth" — the analyst side of the dispute now carries a number, where the October 26–27 reporting carried only the qualitative claims, with the article reading Micron and its >200GB-per-humanoid case as the beneficiary if Musk is right. Watch: a concrete Musk-side figure in a transcript or filing, the estimate being restated in Counterpoint Research's own robot-memory work, chipmakers quantifying humanoid demand in guidance.
- 2026-10-28 (~12:56 PM ET, MSN — msn.com/en-us): development — Musk showed Optimus folding clothes while admitting the robot was not operating autonomously, with the clip set beside Figure AI's spoken-instruction humanoid demonstration in the same coverage. The household-task instance of the hype-gap storyline, days after the Fremont mass-production claims and the late-2027 outside-customer timeline from the Q3 call. Watch: an autonomous version of the task shown, the outside-customer timeline holding. tags: tesla, humanoids
- 2026-10-27 (~4:45 AM ET, India Today): the claims layer gains a memory-demand rebuttal — Elon Musk rejects Counterpoint Research's forecast for humanoid-robot memory demand, arguing robots will need far more memory than the analyst projection, with the headline framing "robots could take all your RAM" (indiatoday.in/technology). No figure from Musk in the reporting; the dispute puts the CEO against Counterpoint's per-robot DRAM line and on the same side as Micron's >200GB-per-humanoid bull case already on this thread. Watch: a concrete Musk figure surfacing in a filing or transcript.
- 2026-10-27 (~8:45 AM ET, International Business Times — ibtimes.co.uk): the memory-demand rebuttal gains its fuller wording — Musk says humanoid robots' RAM needs "will exceed everything else on Earth" as analysts eye chipmakers, published beside Counterpoint's 20-fold-by-2030 DRAM forecast (24/7 Wall St., ~8:45 AM ET — 247wallst.com/investing) and Micron's >200GB-per-humanoid estimate on file. Development of the October 27 claims-layer rebuttal already recorded from India Today. Still no per-robot figure from Musk to score against either side. Watch: a concrete Musk-side number in a transcript or filing, and Counterpoint's figure restated in its own research.
- 2026-10-26 (~4:48 AM ET, 36kr English edition): the training layer gains a price tag — coverage reports Tesla building a training cluster of 20,000 robots with a total investment of approximately ¥14 billion (≈US$2 billion) (eu.36kr.com/en). Development of the October 22 Optimus Academy record (10,000–20,000 robots learning by imitation because the program lacks general-task training data): the fleet-size ceiling now carries a first reported investment figure. Announced build-out, not footage — no site on camera, and the standing Academy watches hold: autonomous vs teleoperated practice, and the fleet materializing against the 20,000-a-week production claim's history of promises outrun by reality.
- 2026-10-23 (~8:47 AM ET, MSN syndication): the production-claims layer gains a headline-level unit target — a production-start re-run headlined "Tesla starts Optimus production, targets 20,000 robots weekly by late 2027" (msn.com/en-us). The captured body detail carries only the production-standards framing already on record, and no unit count appeared in the October 21–22 earnings coverage on this thread — so the 20,000-weekly figure is headline-grade until corroborated. Claim ledgered in predictions (due 2027-12-31). Watch: corroboration for the figure, the still-absent Optimus demonstration video, the external-delivery date holding.
- 2026-10-23 (~8:47 PM ET, shattered.io): the claim layer gains a valuation framing — Musk says Optimus could drive 80% of Tesla's value, with the Cybercab targeted under $30,000, in a piece re-running the 20,000-a-week ramp framing (shattered.io). No transcript, filing, or named venue attached; the 20,000-weekly production figure stays headline-grade until corroborated, and the promised Optimus demo video remains absent. Watch: the 80%-of-value claim surfacing in a named interview or filing, the demo video, external deliveries holding for late 2027.
- 2026-10-22 (~12:45 AM ET cluster): the mass-production confirmation plus three new dated commitments — Tesla says initial mass production of Optimus has begun at Fremont, with coverage specifying the robot line is replacing the Model S and Model X production lines at the plant (finance.biggo.com/news); Musk says reaching 20,000 robots a week will take at least about another year (driveteslacanada.ca/news); he targets external supply by late 2027 (shattered.io); and Tesla says a video demonstration is "maybe a month or two" out (marketwatch.com/livecoverage), while MarketWatch frames the earnings run as Musk's chance to move "the entire" robotics sector. Development of the thread: the October 21 earnings-day record already had production-quality Optimus units off Fremont's high-volume lines; this window adds the first named production-line casualty, the first quantified ramp horizon, the video-demo window, and an external-supply date. Announced is not shipped — no external customer named, no video released, and the 20,000-a-week figure joins the thread's history of unit-count promises outrun by reality (hands, magnets, the flat-and-long S-curve).
- 2026-10-22 (~8:47–8:51 AM ET): the training layer acquires a name and a fleet size — Musk compared teaching Optimus to raising a toddler: the robot learns by watching people do tasks and practicing them itself, imitation learning rather than teleoperation rigs (MSN syndication, msn.com/en-us), and coverage reports Tesla needs 10,000–20,000 Optimus robots practising in an "Optimus Academy" because the program lacks training data for general tasks (The Next Web, thenextweb.com/news). Development of the October 7–09 video-learning claim: the training-method wager now has a named program — and the 10k–20k fleet-size requirement is itself a unit-count claim landing while the external record on this thread shows several hundred a week with breaking hands. Announced, not shipped: the Academy fleet is a promise with no footage shown. Watch: whether the Academy fleet materializes on camera against the 20,000-a-week claim, and whether any Academy footage shows autonomous rather than teleoperated practice.
- 2026-10-22 (~11:56 AM ET and ~12:56 PM ET, late-window cluster): the earnings-call blocker layer and the skeptic close-out — Musk put external Optimus deliveries "around the end of 2027" on the Q3 2026 call and named the hands, the robot's durability, and its AI as the problems still needing work before then, with the trade read that XPeng and Hyundai's Boston Dynamics "look at least as far along" (electrek.co/2026); The Verge argues being "in production" doesn't make Optimus actually ready — useful, reliable, and durable enough still being worked out (theverge.com/tech); and investor coverage positions Texas Instruments as a chipmaker that "can get paid on every one" of the Robots off the Fremont line — a supply-chain thesis, not a named contract (finance.yahoo.com/technology). Development of the overnight mass-production and training-layer records; the standing distinction holds — announced production, not shipped product. Watch: the demo video and whether it shows hands/durability progress, the first external customer, and whether the 20,000-a-week promise survives 2027.
- 2026-10-22 (~4:49–4:50 PM ET, late-window pair): two developments to the earnings-day layer — (1) the financial tier gains dollars: Tesla's Q3 2026 profit was $787 million, down from $1.37 billion a year earlier (−43%), with spending on robots and self-driving cars framed as what pushed the bill up (decrypt.co/381624); (2) the factory tier gains its first named-executive statement — CFO Vaibhav Taneja says construction has started on a dedicated Optimus factory in Austin that will make far more robots than Fremont (teslanorth.com/2026), a CFO-confirmed site beside the October 1 Giga-Texas building-progress record and the October 21 Fremont-vs-Texas contradictory headlines. Announced construction and a reported loss, not shipped product: unit counts, first external customer, and the Gen 3 public unveiling stay open. Watch: whether the Austin site gets its own ramp date or capacity figure.
- 2026-10-21 (early ET, CTOL Digital, citing Tesla's October 19 annual report): the filing tier — Tesla's annual report dated October 19, 2026 states Optimus V3 is "readied" and first production lines are being actively built, with a July company update describing factories prepared ahead of a commercial-delivery test (ctol.digital/news). Development of the thread: the record's prior V3-line evidence came from watcher outlets, unnamed sources, leaks (October 14 3D models), and Musk commentary; an annual-report filing is the first official-document confirmation. "Readied" is not shipped — unit counts, external customer, and the Gen 3 public unveiling all remain open, and the filing lands the same day as the October 21 Q3 earnings call the thread's October 14 record flagged as the venue for Optimus numbers.
- 2026-10-21 (~4:47 PM ET, TechCrunch): the earnings-cycle development — coverage reports Tesla's profit fell 43% with revenue flat and expenses climbing on the accelerated push into Optimus robots, AI, and autonomy, with Optimus robot production in California headlining the piece (techcrunch.com/2026). The thread's first named financial drag, landing the same day as the October 19 annual-report filing that confirms V3 first-line construction; "readied"/announced is not shipped — unit counts, first external customer, and the Gen 3 public unveiling all stay open.
- 2026-10-21 (~8:45–8:50 PM ET, earnings-day cluster on Tesla's Q3 2026 call): the production tier moves — coverage reports the first production-quality Optimus bots coming off Tesla's high-volume lines at the Fremont factory (teslanorth.com/2026), with Tesla's own qualitative status that early production has proven the process "suitable for a commercial product" (businessinsider.com); on the call, Musk says Tesla aims to begin humanoid robot sales next year — 2027 (msn.com/en-us); and one outlet in the same Wednesday-evening cluster contradicts the location, headlining high-volume Optimus production started at Gigafactory Texas (basenor.com/blogs). Financial layer already logged (profit −43%, revenue flat, expenses up); companion call coverage notes an after-hours beat with strong FSD revenue before shares fell. All produced-or-claimed, nothing sold or deployed with a customer: unit counts, first external customer, the Gen 3 public unveiling, and Fremont-vs-Texas reconciliation all stay open.
- 2026-10-18 (~4:48 AM ET, MSN syndication): the thread's supply-chain layer gets its own headline piece — Musk reiterates Optimus as Tesla's most important product, and syndicated coverage identifies third-party suppliers/technology partners positioned to profit from Optimus production, framed as "paid before Tesla"; the supplier names are not recoverable from capture. Development of the supplier-strain record (Sept 25 The Information "suppliers can't keep up"; Sept 18 Chinese part audits; Oct 13 magnet-restriction delay) — adds an upstream who-profits framing, no new dated claim. Watch: named suppliers vs the audited Chinese parts base, supplier lead times as ramp corroboration.
- 2026-10-14 (~12:52 PM ET, MSN aggregation): next-gen Optimus 3D models surfaced before any planned reveal — the physical-design layer visible through an accidental leak while the on-file record shows Gen 3 in limited production at Fremont since late July 2026, no public unveiling yet, and the Oct 13 magnet-restriction delay framing. Development of the production/hype-gap thread; a leaked design artifact is not the unveiling and not shipped hardware. Watch: an official reveal, whether the leak matches production hardware, unit counts. msn.com/en-us
- 2026-10-14 (~5:04 PM ET, Teslarati): Tesla opened the Q&A for its Q3 earnings call set for October 21, framed as "its biggest earnings call yet" with Cybercab, Robotaxi, Optimus and regulators looming — a dated venue for the market-context layer filed October 7–09 (Optimus/FSD/robotaxi outlook mattering more than the print). It lands one beat after the same day's next-gen Optimus 3D-model leak. Watch: what Optimus unit counts or production claims surface on the October 21 call against the several-hundred-a-week/hands-breaking record. teslarati.com
- 2026-10-13 (hands layer): Musk in an interview calls designing a human hand for robots "shockingly difficult" and says it is a problem Tesla is working on (MSN-syndicated, October 13) — the company's own leader conceding the dexterity bottleneck the trade-press skeptic voice named on October 10. Talk about a component under development; nothing shipped, no unit counts. Watch: whether the next-generation hand design gets disclosed with mass-production intent, as the prior-generation hand claim pattern went. msn.com/en-us
- 2026-10-13 (~8:47 PM ET, Drive Tesla Canada): Musk says China's magnet export restrictions have set Optimus production back by months — the ramp record's first named external cause and first quantified delay, landing on the several-hundred-a-week/hands-breaking reality-check layer and consistent with the XPeng-reps-briefed-delay color already on file; the article simultaneously carries his continued public-sales targeting. Report-grade single syndication; the unit-count, external-customer, and shipped-not-announced watches all stand, and the delay is logged to predictions as slippage against the 2027 commercial-sales window. driveteslacanada.ca/news
- 2026-10-10→11 (claims layer, three items): (1) Musk interview Oct 10: a sharp line between "real AI" and "polished" robotics (basenor, ~8:45 PM ET Oct 10) — framing arrives while his own showcase agent stays demo-grade. (2) Digital Optimus claims clearing half of Diablo and competing in Counter-Strike on pixel-only input (~12:47 PM ET Oct 10) — announced, not shipped; the ledgered September 2026 "user experience ready" window lapsed with no shipped product, so this is capability-claim color, not a resolution. (3) The video-imitation-learning claim (recorded 10-07→09) gets a named mirror bet: Figure AI's "Project Go-Big" — two camps, same training-method wager. Watch stands: shipped-not-announced on Digital Optimus, a demo behind the video-learning claim, Gen 3 Q4 output. cryptobriefing.com
- 2026-10-07→09 (training-method claim + prophecy layer): (1) "Optimus learned to do chores just by watching humans on video" (Tesla: vacuuming, sorting parts — "from human video footage" instead of teleoperation rigs): the training-method pivot claim — vendor-claimed, no deployment shown; the thread's teleop-vs-autonomous labeling watch applies directly. (2) Patent: soft, curved touch sensors — "a skin that can feel." (3) The prophecy layer: ex-Stability AI CEO Emad Mostaque — "Tesla will sell more robots than cars within two years," ~$400B from 10M robots (noted; ledgered). (4) Market context: Q3 coverage says Optimus/FSD/robotaxi outlook matters more than the earnings print (Goldman), stock dipped 1.4% on the capex-gap "Optimus proof" ask; one consumer piece keeps Optimus public buyers "by end of 2027." Watch stands (monthly output, hand-durability, external customer) plus: the video-learning claim needs a demo, not a sentence.
- 2026-10-03 (ramp claim + color; number already covered): ">1,000 Optimus robots a week" production claim with hand troubles keeping most in testing ("The production scaling challenge is very substantial") — the thread's ramp-vs-reality gap in one sentence. Color: Halloween teaser (Optimus hand rising from a graveyard, hidden vehicle) — Teslarati.
- 2026-10-01 (ramp infrastructure): the dedicated Optimus factory at Giga Texas has gone "from raw earth to a nearly complete structural steel skeleton" by October (Teslarati-grade; feed story 5995) — the physical-layer corroboration inside the ramp picture The Information drew Sep 25 (hands break, suppliers behind, no generalization). Watch stands, plus Macrohard's Sep UX deadline now lapsed — check the ledger.
September 2026 · 5 updates
- 2026-09-25 the reality-check layer arrives from the same reporting lineage — The Information (via electrek, pub ~12:51 PM ET): Tesla now builds several hundred Optimus a week, but the hands break, suppliers can't keep up with component demand, and the robots can't generalize across tasks; a companion rundown adds The Information's V3 production-milestone detail. First concrete picture of what breaks at rate — the strongest external counterweight to the Sep 6 "production hit" ledger resolution. Same cluster: Yangtze Delta supplier tours (Sep 24 — the supply-side corroboration) and Musk at the G20: robotics "could add trillions," 1 billion robots in 10 years. The gap between the G20 claim and The Information's report is the story.
- 2026-09-19 deployment-infra artifact — "Optimus chargers" reference spotted in the Tesla app (Sep 18, wholemars; blog-grade writeup). Charging infrastructure in consumer app code is hard-to- fake ramp corroboration; two independent datapoints in two days in the same window.
- 2026-09-18 first Optimus item of its kind on record: syndicated "sources" reporting Tesla officials/engineers auditing Chinese part suppliers ahead of an Optimus rollout "later this year" (~4:50 PM ET). Two flags: unnamed-source syndication (not filing-grade), and "release later this year" is a fresh date-claim layer on top of the production record (Gen 3 internal production since late July; consumer sales targeted 2027). Consistent with a real ramp; verify to the originating outlet before weight.
- 2026-09-18 first Optimus item of its kind on record: syndicated "sources" reporting Tesla officials/engineers auditing Chinese part suppliers ahead of an Optimus rollout "later this year" (~4:50 PM ET). Two flags: unnamed-source syndication (not filing-grade), and "release later this year" is a fresh date-claim layer on top of the production record (Gen 3 internal production since late July; consumer sales targeted 2027). Consistent with a real ramp; verify to the originating outlet before weight.
- 2026-09-06 coverage came from web-searched Tesla-watcher outlets (RobotsBeat, Tesorb, All Things Elon, Teslarati, electrek, Business Insider) — see ledger for the dated claims. The hype-gap frame connects directly to the "humanoid reality-check war" thread: Tesla is that war's hype pole.
Background
- The thread: Tesla's humanoid crossed from demo-reel to (internal, low-rate) production — Gen 3 limited production at Fremont on the dismantled Model S/X line (late July 2026), ~1,000 units deployed inside Tesla facilities doing narrow repetitive tasks (cell sorting, kitting, inspection) that Musk himself calls "primarily for learning and data collection," not productive work (Jan 2026). Consumer $20–30k sales targeted 2027; 1M units/yr capacity is a plan, not output; Musk warns the ramp is "the hardest to scale in manufacturing that we've ever made at Tesla" with a "flat and long" S-curve, and concedes nobody has ever solved the general-purpose part ("like no one's ever achieved this,"). Adjacent live claim: Digital Optimus / Macrohard (joint xAI-Tesla computer-use agent, announced Mar 11 2026) has a September 2026 UX-readiness deadline that is NOW — no public demo or shipped product as of Sept 6, and the xAI-side effort reportedly stalled (leaders departed, 600-contractor data op paused, rebuilt with two Cursor hires).
- Watch for: Gen 3 monthly output by Q4 (hundreds vs dozens), first EXTERNAL customer (Figure already bills BMW $25/hr; Tesla has none), Macrohard ship-or-slip this month (ledger), the repeatedly-delayed Gen 3 public unveiling, teleop-vs-autonomous labeling on any new demo.
- Watch stands: monthly output by Q4 (hundreds vs dozens), first EXTERNAL customer, the repeatedly-delayed Gen 3 public unveiling, teleop-vs-autonomous labeling — plus now: hand-durability fix cadence and whether supplier scaling shows in lead times.