Robots digest — 2026-10-20
Top line: The lead stays the long-read on China's grip on the robotics industry, whose freshest layer is the raw material at the bottom of it — rare-earth magnets, the dependency already blamed for Tesla's Optimus delays. Behind it, expert verdicts on GPT-6 "Astra" as a robot brain put a hard date on general-purpose robots: not before early 2028. The day's commercial arrivals: Agility Robotics converts its deployment-over-demos pitch into a signed agreement with its first named agricultural-machinery customer, AGCO, with a 2027 pilot at the Hesston, Kansas facility; Gecko Robotics lands a $300 million U.S. Navy contract; Germany and NVIDIA put physical AI and industrial robotics at the center of a national-level push; and Schaeffler opens a plant at Taicang, China where humanoid robots build the joint modules for other humanoid robots. The Agility thread thickens late in the day: The Information reports — relayed by Agility's own account — that Nyobolt has won a contract to supply batteries for robots operating in Walmart distribution centers, at a "thousands of humanoids" scale no filing yet supports. Around them: Exotec commissions two Skypod systems in Japan with Mitsubishi Shokuhin; NAPA Auto Parts' flow-center build-out with Hy-Tek Intralogistics and Hai Robotics gets its first substantive detail; FANUC's "AI Welding" is reported to generate weld paths from blueprints; and Reservoir opens an agricultural-robotics incubator for Washington growers. Earlier in the day, China's robotics correction reached its endpoint for one name — the ¥22.3-billion (≈US$3.1-billion) humanoid unicorn collapsed completely, with Unitree named a rare survivor — alongside two capability datapoints, a KAIST humanoid pushing 35 kg (77 lbs) while walking on onboard cameras alone, and a Figure 03 walking 560 km (348 miles) on a treadmill in an ongoing endurance test. The day closes with an evening layer published around 8:47 PM ET: Unitree closes October 20 at 526.03 yuan (≈US$74) a share, having added roughly ¥40 billion (≈US$5.6 billion) of market value in three sessions at a 388x P/E — the rare survivor repricing itself as the market's most expensive robotics name; a translated Tesla-watcher piece claims Tesla has already started building a factory to produce 10 million Optimus units a year, landing the night before the company's October 21 earnings call; a keynote panel at RoboBusiness 2026, Rodney Brooks among the named voices, quantifies the reality check — Chinese humanoid hands failing after about 200 hours against factory demands of 30,000 hours and a 99.999% success rate; and Amazon puts €10 billion (≈US$11.6 billion) behind European logistics and robotics, with a $1 billion global upskilling pledge the coverage frames as the tell. A follow-up also fills out the day's funding item: Eka Robotics is MIT professor Pulkit Agrawal's startup, and its $150 million Series A was led by a16z and F-Prime at $1.45 billion. Shorter lines round out the day: Boston Dynamics' CFO put a price floor on the company's robot dog — upwards of $300,000; chip supplier Allegro MicroSystems unveiled 48V power and precision sensing for humanoid robots, with a Sarcomere Dynamics dexterity collaboration; and Agility CEO Peggy Johnson named safety as the gate for humanoids — Digit stays inside fenced areas. The remainder (restates of the AGCO agreement and Agility's own safety, legs, and hiring posts, recirculated OpenAI humanoid items including a stale confirmation of a 2027 demo, a KAIST re-look, Faraday Future's FF Nova pricing repeated, a RoboBusiness rebrand, a Locus Robotics Gen Z returns survey, a 1X pre-IPO guide, a Forbes neurotech column, an Avenda Health software expansion, a first-person column on robot power supplies, a robot-massage consumer piece, TRR50 scheduling) is restate, evergreen, commentary, survey, or off-beat, dropped.
1. China's grip on the robotics industry runs through rare-earth magnets
What happened: A long-read published early on October 20, 2026 (MSN syndication) argues that China holds the global robotics industry in its grip, anchoring the claim in a Goldman Sachs projection of roughly 6.5 million humanoid robots built per year by 2035 — and in the fact that every one of them runs on rare-earth magnets, a component where China sits at the front of the supply chain. The piece extends the magazine's running theme of Chinese dominance in humanoid manufacturing rather than reporting a single new event.
What's genuinely new: The Goldman forecast itself is not fresh — the 6.5-million-by-2035 figure circulated when the bank's "Physical AI" report landed at the end of August 2026 — but the article's framing of the rare-earth magnet as the bottleneck component is the new layer. It dovetails with what the record already carries: Beijing's magnet export restrictions were blamed by Elon Musk on October 13, 2026 for setting Optimus production back by months, and US makers warned in early October that actuator and rare-earth gaps could triple domestic costs. The dependency is no longer a hypothetical — it has already inflicted a measured delay on a headline Western humanoid program. As a development of the long-running China-industrial-shift storyline (China's manufacturing machine repositioning into robotics, on record since September 9), the article adds the material dependency to the established scale story.
Watch: Whether the magnet bottleneck produces policy: dedicated funding for magnet production, export-license relief, or the waiver machinery already surrounding the FCC robot import ban. Also whether any US humanoid maker discloses a magnet-supply fix the way it discloses battery or actuator sourcing.
Links: MSN — Robot wars: how China has the robotics industry in its grip
2. Expert verdict on GPT-6 Astra as a robot brain: great on benchmarks, can't fold laundry — general-purpose robots not until early 2028
What happened: Also published early on October 20, 2026, a writeup of commentary by robotics researcher Xu Huazhe following OpenAI's release of GPT-6 "Astra": three experts delivered a split verdict, crediting the model with breaking robot benchmarks — while noting it is merely incremental on coding and agent tasks — and pointing out that benchmark performance does not translate into basic manipulation. The takeaway headline conclusion: Astra still cannot fold laundry at all, and general-purpose robots are unlikely to arrive until early 2028. This is commentary and projection, not a capability announcement; nothing new shipped.
What's genuinely new: A concrete, dated timeline from a named voice in the humanoid reality-check storyline — the same skeptical strand that has carried DeepMind's "GPT-2 stage" caution from Keerthana Gopalakrishnan, the IFR's Jane Heffner, and Mark Cuban's dismissals. The "GPT-2 stage" framing said reliability, not raw capability, is the binding constraint; this piece supplies the same argument from the benchmark direction: model scores rising while household-grade manipulation sits at zero. A timeline of early 2028 is more specific than the thread has carried before.
Watch: Whether other experts align on early 2028 or push the date further; whether the next frontier-model release gets tested for the laundry-grade manipulation gap rather than benchmark gains; and how this verdict is received against vendor timelines already on file, several of which aim earlier.
Links: BigGo Finance — Xu Huazhe: Astra breaks robot benchmarks, but can't fold laundry at all
3. China's robotics correction lands: the ¥22.3B humanoid unicorn collapses, Unitree survives
What happened: A 36kr feature published October 20, 2026 (~4:45 AM ET) reports that "the first batch of robotics unicorns are starting to line up for bankruptcy," with the centerpiece the complete collapse of a humanoid robot unicorn valued at ¥22.3 billion (≈US$3.1 billion) at its peak — and Unitree Robotics standing out as a rare surviving name in the field.
What's genuinely new: This is the endpoint of a correction thread with a paper trail. On October 16 the same outlet reported that this ¥22 billion (≈US$3.1 billion) "star" robotics project's secondary shares remained unsold even at a 50% discount; on October 19 Unitree's own tape framed the bubble risk explicitly — a 629% IPO surge followed by a >50% decline, coverage asking whether "the robots are real" at "the hundred-times price." The new piece closes the loop for one unicorn (from unsold stakes to complete collapse), widens the frame to an entire first cohort of robotics unicorns facing bankruptcy, and installs Unitree as the survivor — the same company whose Shanghai STAR debut slid 53% from mid-August highs. The survivor/outlier framing is the notable turn: the correction is no longer only about one volatile listing, but about which model can live through it. The evening then adds the counterweight: a market check published October 20 (~8:47 PM ET) finds Unitree closed October 20 at 526.03 yuan (≈US$74) a share after rising more than 30% in cumulative deviation across the October 16–20 sessions — roughly ¥40 billion (≈US$5.6 billion) of added market value — at a 388x price-to-earnings ratio the coverage frames as raising commercialization questions. The survivor named by the correction coverage ends the same day as the beat's most aggressively priced stock: the answer to "where the price settles" so far is an ever-steeper one, on a multiple that assumes a commercialization story no financial filing yet supports.
Watch: Named bankruptcy filings from the first unicorn cohort; whether Unitree's survivor status holds against its own −53% slide and the Shanghai IPO-curb backdrop; whether the 388x multiple survives commercialization questions being priced in; and how Western listings in the same repricing ledger (Agility's pending S-4, PlusAI's SPAC close) travel against this tape.
Links: 36kr — The first batch of robotics unicorns are starting to line up for bankruptcy · BigGo Finance — Unitree adds nearly ¥40 billion in market value over three sessions as 388x P/E raises commercialization questions
4. KAIST Humanoid v0.9: pushes 35 kg (77 lbs) while walking, no external cameras
What happened: Footage syndicated via MSN and published October 20, 2026 (~4:46 AM ET) shows KAIST's Humanoid v0.9 pushing objects up to 35 kg (77 lbs) and lifting 20 kg (44 lbs) while walking, with navigation and task execution running on the robot's onboard cameras alone — no external camera arrays or lab capture rig in the loop.
What's genuinely new: A fresh demonstration, not a continuation of an existing thread: the item supplies specific technical benchmarks for a new model — named payload figures for pushing and lifting under walking conditions, and the claim that perception is fully onboard. Payload-while-walking plus camera-only autonomy is a stronger combination than most demo clips carry, and it arrives alongside the broader reality-check conversation about whether household-grade manipulation is anywhere close.
Watch: Independent replication of the payload figures outside lab conditions; behavior on uneven or outdoor terrain; whether v0.9 graduates to a named deployment or partner program.
Links: MSN — KAIST humanoid pushes objects up to 77 lbs while walking, no external cameras
5. Figure 03 endurance test: 348 miles (560 km) walked in nine days
What happened: Heise reported on October 20, 2026 (~4:45 AM ET) that a Figure 03 humanoid robot walked 560 km (348 miles) on a treadmill in an endurance test; a RobotsBeat writeup (~1:05 PM ET) attributes the figure to CEO Brett Adcock and adds that the run logged 1.2 million steps over nine days in an ongoing hardware stress test. Neither report discloses error or failure data.
What's genuinely new: A durability metric for Figure's third-generation humanoid, and a debut for the company in the record. The single headline number is a company-run test, relayed by its own CEO with no error or failure data — the same announce-versus-evidence gap the reality-check thread keeps flagging: 560 km of treadmill walking says nothing about folding laundry, but it is a concrete claim about joint and actuator longevity that can be checked later. The "ongoing" framing means the number will keep moving.
Watch: Disclosure of error rates or mean-time-between-failures from the test; the final total and its duration; any third-party verification; what the endurance figure implies for multi-hour commercial shifts if Figure's deployments progress.
Links: heise — It keeps going and going … Figure 03 robot covers 560 km · RobotsBeat — Figure 03 walks 348 miles in nine-day treadmill stress test
6. Agility × AGCO: a commercial agreement for a 2027 humanoid pilot at the Hesston, Kansas plant
What happened: Two captures on October 20, 2026 — Agility Robotics on X (~8:16 AM ET) and AGCO's press release via Yahoo Finance (~8:46 AM ET) — report a signed commercial agreement between AGCO (NYSE: AGCO), the agricultural-machinery maker, and Agility Robotics: a humanoid pilot at AGCO's Hesston, Kansas facility, expected to begin in 2027, starting with parts sorting and stacking and placing totes for storage. Agility's own framing of the vertical expansion: "Automotive, logistics, and now agricultural machinery."
What's genuinely new: A development of Agility's ongoing record, and the first named agricultural-machinery customer on it — a signed commercial agreement with a named site, named starting tasks, and a date, arriving mid-SPAC with Churchill Capital Corp. XI (close date set November 2, 2026). It converts the deployment-over-demos thesis Agility has campaigned on since August into more named-customer paper. The 2027 start remains promise-grade — announced, not begun — and the filed S-4 trajectory (~800 Digit 5 units deployed in 2027) now has a real customer name to be measured against.
Watch: pilot start confirmation and task list from the floor; Hesston task-success data once running; whether the S-4's ~800-in-2027 figure counts AGCO units.
Links: X — Agility Robotics: commercial agreement with AGCO · Yahoo Finance — AGCO and Agility Robotics Advance Smart Manufacturing with Humanoid Robot Pilot in Hesston, Kansas, Facility
7. Gecko Robotics secures a $300 million U.S. Navy contract
What happened: The Robot Report, October 20, 2026 (~6:25 AM ET): Gecko Robotics has secured a $300 million contract with the U.S. Navy. Gecko claims the award signals the Pentagon's confidence that it is "a core military capability that is delivering results."
What's genuinely new: A debut for Gecko on the robots beat with its largest named dollar figure to date. The company has one prior appearance on the record — September 29, on the AI side — as the first named industrial-infrastructure deployment for NVIDIA's Open Agent Safety Platform. A defense-scale contract turns that vendor adjacency into a procurement datapoint. The capture carries no scope detail: duration, ship classes covered, or whether $300 million is a ceiling or an awarded amount.
Watch: contract scope and duration in Navy documents; independent confirmation of the $300 million figure; whether Gecko's Open Agent Safety Platform adoption figures into the naval deployment.
Links: The Robot Report — Gecko Robotics earns $300M contract with the U.S. Navy
8. Germany and NVIDIA put physical AI and industrial robotics at the center of a national push
What happened: RobotsBeat, October 20, 2026 (~8:38 AM ET): Germany and NVIDIA have placed physical AI and robotics at the center of an industrial push — described as a national-level strategic partnership targeting industrial robotics — with Siemens named as testing the applications in live production and a planned Agile Robots investment named alongside.
What's genuinely new: The country-level frame. The record already carries Germany's component-supplier angle on the humanoid build-out ("China builds them, but Germany could supply the joints and hands," October 1) and individual European deployments, but no previous national-level NVIDIA partnership over industrial robotics. Report-grade in this capture: a single trade source, with no government document, funding figure, or program timeline in hand.
Watch: the partnership's actual text and any funding figures; Siemens confirmation of live-production testing; the Agile Robots investment's terms.
Links: RobotsBeat — Germany and NVIDIA Put Physical AI and Robotics at Center of Industrial Push
9. Schaeffler opens a factory in Taicang, China where humanoids build the parts for humanoids
What happened: Heise, October 20, 2026 (~8:48 AM ET): Schaeffler has opened a factory where humanoid robots manufacture actuators — joint modules — for humanoid robots, with the robots also being trained on site. RobotsBeat's writeup (~1:04 PM ET) supplies the address and scale: Schaeffler's Taicang plant, six humanoid robots working across five stations, with digital twins linking production and training.
What's genuinely new: The robots-build-robots loop under a named Western supplier — but located in China, not in Germany. The record's earlier version of the loop was also China-side: UBTECH's Liuzhou humanoid-production factory, opened September 16, 2026. A German engineering group siting the loop at its Chinese plant fits the China industrial-shift storyline's onshore-manufacturing story, and it corrects the day's earlier "German factory" reading. Output volumes and named humanoid customers are absent from both captures.
Watch: which humanoid maker consumes Schaeffler's joint modules; output volumes; verification of the on-site training and digital-twin claims.
Links: heise — Schaeffler: Humanoid robots build joint modules for humanoid robots · RobotsBeat — Schaeffler's Taicang factory uses six humanoids across five stations to assemble robot joint modules
10. Boston Dynamics CFO to Fortune: the robot dog can cost upwards of $300,000
What happened: Boston Dynamics CFO Amanda McMaster, quoted in Fortune coverage the company retweeted on October 15, 2026 (the tweet surfaced ~12:35 PM ET), said its robot dog can cost upwards of $300,000.
What's genuinely new: A development of the Boston Dynamics storyline — the thread has carried Spot deployments (Kirin Brewery's Chitose Plant, Coca-Cola Europacific Partners scaling Spot globally), platform updates, Atlas production targets from Hyundai, and an IPO pushed toward 2029–2030 on the grounds that Atlas is "yet to deploy at scale," but never a price floor from company leadership. The figure is a quip-grade conference number: scope, configuration, and whether it includes a service bundle are all absent from the capture.
Watch: whether the $300,000-plus figure gets context (unit price versus a configured fleet); how it reads against the price points already on file — Agility's filed ~$200k equivalent and Unitree G1's ~$16k headline; any Spot pricing in future public documents ahead of a potential listing.
Links: X — Boston Dynamics RTs Fortune: robot dog upwards of $300,000
11. Eka Robotics raises $150 million at a $1.45 billion valuation for force-aware robot AI
What happened: RobotsBeat, October 20, 2026 (~11:46 AM ET): Eka Robotics raised $150 million for AI that combines vision and force feedback — targeting fast, adaptable robot manipulation for industrial applications — at a $1.45 billion valuation; a same-day follow-up published ~8:46 PM ET fills out the round: it is a Series A led by a16z and F-Prime, and the startup is led by MIT professor Pulkit Agrawal, training systems described as physics-first.
What's genuinely new: A debut: the record carries no prior Eka storyline, now developed same-day with the round's named people and investors. The raise's substance lines up with the manipulation-reliability conversation the record keeps flagging — force feedback as a route to the laundry-grade manipulation that benchmark-topping models still lack, the exact gap the Xu Huazhe verdict on GPT-6 "Astra" named earlier on October 20. Still funding-grade only: use of proceeds, round structure, and any named customer or pilot are all absent.
Watch: a named industrial deployment or customer; whether the force-feedback pitch gets manipulation metrics attached rather than demo footage; the round's board and additional investors.
Links: RobotsBeat — Eka Robotics Raises $150M for Force-Aware Robot AI at $1.45B Valuation · Wowtale — Physics-First Robotics Startup Eka Robotics Raises $150M Series A
12. Allegro MicroSystems unveils 48V power and precision sensing for humanoid robots, Sarcomere Dynamics collaboration along for dexterity
What happened: Globe Newswire, syndicated via the Manila Times, October 20, 2026 (~12:47 PM ET): chip supplier Allegro MicroSystems unveiled a portfolio of 48V power and precision-sensing solutions aimed at physical AI in humanoid robotics — pitched at the critical size, weight, and safety challenges in humanoid robots — with a collaboration with Sarcomere Dynamics said to enable near-human dexterity.
What's genuinely new: A development of the humanoid component supply chain storyline opened October 15 (China's robotic-hand suppliers gaining importance as shipments meet the harder test of reliable manipulation, with Japan strong in joint components — both watched for named suppliers): the named-supplier lane now includes a Western silicon supplier pitching power and sensing hardware, plus a dexterity partner. It is an announced portfolio, not a shipped design win — the capture carries no volumes, no named humanoid customer, and the "near-human dexterity" wording is vendor claim, not demonstrated metric.
Watch: which humanoid maker consumes the portfolio; the Sarcomere Dynamics hand's demonstrated dexterity against the manipulation test; whether Alphabet's humanoid bill of materials (or any other) names Allegro.
13. Agility CEO Peggy Johnson: safety is the gate — Digit stays inside fenced areas
What happened: An MSN-carried Fortt Knox video conversation with Agility Robotics CEO Peggy Johnson, October 20, 2026 (~12:47 PM ET), has the CEO calling safety the gate for humanoid robots — with the specific operational detail that Digit units stay inside fenced areas.
What's genuinely new: A development of Agility's safety campaign strand (the FORT deployment-safety announcement of October 1, inside the broader reality-check framing Agility has campaigned on since August): the CEO herself now supplies the fence line as the boundary of the current safety case. "Stays inside fenced areas" is a candid constraint rather than a capability claim, and it lands one day after the AGCO commercial agreement and under three weeks before the SPAC close dated November 2, 2026.
Watch: whether unfenced or fence-free operation gets a roadmap or a date; how the constraint reads against the S-4's ~800-Digit-5-units-in-2027 trajectory and customer expectations at sites like Hesston.
Links: MSN — Peggy Johnson, Agility Robotics CEO: A Fortt Knox Conversation
14. The Information: Nyobolt signed to power "thousands" of Agility humanoids at Walmart distribution centers
What happened: On X, October 20, 2026 (~2:00 PM ET), Agility Robotics retweeted a The Information report: Nyobolt, valued at $936 million in May, has secured a deal to power thousands of Agility Robotics humanoids after winning a contract to supply batteries for robots operating in Walmart distribution centers.
What's genuinely new: A development of the Nyobolt strand on Agility's record — the vendor-quantified battery claims of October 15 (90 minutes of work per 9-minute charge, a designed 10:1 run-to-charge ratio) and the first-half-2027 early-access window of October 16, both announced rather than shipped — now gains its first named deployment site and first scale figure. It is report-grade: a single outlet, relayed through Agility's own account, with no confirmation from Walmart or Nyobolt and no delivery date. The "thousands" figure also stands far above the filed S-4 trajectory the record measures Agility against (~800 Digit 5 units deployed in 2027), and lands inside the window before the November 2, 2026 SPAC close.
Watch: confirmation from Walmart, Nyobolt, or a filing; whether "thousands" reaches the paperwork; first Digit 5 units on Nyobolt packs at a Walmart site against the H1 2027 early-access window.
15. Exotec commissions two Skypod systems in Japan with Mitsubishi Shokuhin
What happened: Robotics & Automation News, October 20, 2026 (~4:54 PM ET): warehouse robotics company Exotec announced the commissioning of two new Skypod systems in Japan with Mitsubishi Shokuhin.
What's genuinely new: A debut — Exotec appears on the record for the first time. It lands in the warehouse-automation lane the record has been tracking company by company (NAPA's Hy-Tek × Hai Robotics build-out; LX Pantos's Incheon test lab; Amazon's million-robot fleet), and it is a completed-commissioning event with a named Japanese corporate client rather than a promise. Throughput, site counts, and timelines are absent from the capture.
Watch: throughput and site data from the two installations; whether Mitsubishi Shokuhin expands the deployment; Exotec's next named customer.
16. NAPA Auto Parts revs fulfillment efficiency with automation
What happened: Supply Chain Dive, October 20, 2026 (~4:54 PM ET): NAPA Auto Parts is working with Hy-Tek Intralogistics and Hai Robotics to maximize warehouse storage density and improve order fulfillment efficiency; a Yahoo Finance syndication followed the same day (~4:55 PM ET).
What's genuinely new: A development of the Hy-Tek × Hai Robotics selection for NAPA recorded October 15 — the first substantive detail since the partnership announcement, supplying the goals behind the "next-generation flow center transformation": storage density and order fulfillment efficiency in automotive-parts distribution. Flow-center locations, system counts, and a timeline remain absent.
Watch: flow-center locations and counts, Hai Robotics system scale, project timeline, whether the collaboration generalizes beyond NAPA.
Links: Supply Chain Dive — Napa Auto Parts revs up fulfillment efficiency with automation · Yahoo Finance — Napa Auto Parts revs fulfillment efficiency with automation
17. FANUC's "AI Welding": generating weld paths from blueprints
What happened: A Japanese-language writeup on note.com published October 20, 2026 (~4:53 PM ET) reports that FANUC's "AI Welding" aims to generate welding conditions and paths directly from blueprints — replacing the expert process of manually setting conditions and teaching the robot its movements.
What's genuinely new: A debut for FANUC, the industrial-robot incumbent, on the record. It lands beside the AI-welding capability Hyundai entered on October 18 (AI robot welding that autonomously locates and welds roughly 1,700 points per vehicle, unveiled at its "Forest Tech Day"). Single-source and product-claim grade: no product release, deployment, or customer appears in the capture.
Watch: FANUC primary documentation or a named deployment; independent testing of the blueprint-to-weld-path claim; whether other welding players follow.
18. Reservoir launches an agricultural-robotics incubator for Washington growers
What happened: The Robot Report, October 20, 2026 (~3:40 PM ET): Reservoir has launched a new agricultural robotics incubator, aiming to put agricultural robotics startups to work in orchards, vineyards and berry fields alongside researchers and growers.
What's genuinely new: A debut; no Reservoir thread exists on the record. It pairs with the day's agricultural-robotics signal higher up the value chain — AGCO signing Agility for a 2027 humanoid pilot at its Hesston, Kansas plant — but sits at the startup-incubation layer, with no funding figures, named startups, or trial dates in capture.
Watch: named startups and first field trials; funding behind the incubator; whether grower partners convert to deployments.
Links: The Robot Report — Reservoir launches new agricultural robotics incubator for Washington growers
19. Reality check at RoboBusiness: Chinese humanoid hands fail at 200 hours — factories demand 30,000 hours and a 99.999% success rate
What happened: Coverage published October 20, 2026 (~8:46 PM ET) of a RoboBusiness 2026 keynote panel — Rodney Brooks among the named roboticists — warning against excessive expectations for humanoid robots. Its centerpiece is a quantified reliability gap in the humanoid component chain: Chinese-built humanoid hands failing after roughly 200 hours of operation, against what factories demand — 30,000 hours and a 99.999% success rate.
What's genuinely new: A development of the humanoid reality-check storyline — Brooks is already the thread's standing skeptic voice, with a "10+ years to reliable complex-environment operation" verdict on record — and it extends the day's earlier benchmark verdict (GPT-6 "Astra" breaks robot benchmarks but cannot fold laundry) into the hardware: model scores are one bottleneck, but the hands on the shelves reportedly die after hours, not months, of production work. The named supplier lane opened October 15 — Chinese robotic-hand suppliers facing the harder test of reliable manipulation — now has its first failure figure and its first customer-side demand figure, both single-report grade.
Watch: the panel's full cast and any recording; independent confirmation of the 200-hour failure figure and which hand makers it covers; how the 30,000-hour demand reads against Figure's 560-km treadmill endurance claim in this day's record.
20. Claim: Tesla has started building a factory to produce 10 million Optimus units per year
What happened: A translated note.com piece published October 20, 2026 (~8:47 PM ET) claims Tesla has started building a factory intended to produce 10 million Optimus units per year, framing the number as a brutal scale check on the robot program's earlier "millions" and 2027-launch framings. The timing is the night before the company's October 21 Q3 earnings call.
What's genuinely new: A development of the Optimus production record — the dedicated factory with structural steel scheduled by October, the on-record "several hundred Optimus a week" output, and Musk's ">1,000 robots a week" claim with hand troubles. A 10-million-per-year target is an order of magnitude above all of it, and the piece supplies no Tesla source, no site, no timeline, and no filing behind either the target or the claim that construction has started. Against the same record's measured constraints — the China magnet-export restrictions blamed on October 13 for setting Optimus production back by months — this is an announce-grade claim, with the distinction sharp and unaltered: claimed, not shipped.
Watch: the October 21 earnings call for an on-record production figure; a named site and timeline behind the claimed factory; whether the 10-million target appears in any Tesla document.
Links: note.com — Tesla has started building a factory to produce 10 million Optimus units per year
21. Amazon: €10 billion (≈US$11.6 billion) for European logistics and robotics, with a $1 billion upskilling pledge framed as the tell
What happened: AInvest, October 20, 2026 (~8:47 PM ET): Amazon announced €10 billion (≈US$11.6 billion) for European logistics expansion and robotics, alongside a $1 billion global upskilling pledge. The coverage's critical frame sizes the two numbers against each other — the workforce pledge described as "a rounding error" beside the hardware spend, and a "big tell" about where the money is going.
What's genuinely new: A development of the Amazon warehouse-robotics record — over 1 million robots deployed against a ~1.5-million worker force, opened October 17 — which gains its first dated investment figure: an amount of money behind the fleet's expansion rather than a headcount datapoint, plus a named workforce countermeasure attached to it. All figures are announce-grade: no robotics spending breakdown, no European sites, no robot counts the €10 billion covers, no named upskilling programs.
Watch: the European sites and robot counts the €10 billion funds; whether the $1 billion upskilling pledge converts to named programs; robotics-per-worker and injury/safety data as the deployment continues.
Links: AInvest — Amazon's $1 billion upskilling pledge is a rounding error with a big tell
Dropped
- How Autonomous Mobile Robots (AMRs) Improve Warehouse Operations: an evergreen explainer of how AMRs differ from AGVs and cut walking/heavy lifting in warehouses — general background with no dated news, no new deployment, no new product. No news event to file; dropped as non-news.
- Agility Robotics selects Nyobolt as its ultra-fast-charging battery technology partner for its new humanoid (Robotics & Automation News, October 20): a restate of the partnership announced and quantified on October 14–16, when Agility itself put the numbers on record — 90 minutes of work per 9-minute charge, a 10:1 run-to-charge ratio, up to 20 hours of work in a 24-hour day, early access in the first half of 2027. Nothing new beyond the named pairing; dropped as a restate.
- OpenAI plans to develop AI-powered humanoid robots (Gulf Today, October 20): restates OpenAI's widely reported humanoid-robotics ambitions — bringing AI into the physical world — with no new date, product, or program detail; dropped as a restate.
- Elon Musk says Optimus and Cybercab will account for most of Tesla's value (News.bitcoin.com, October 20): a restate of Musk's standing value prediction, landing the day before Tesla's October 21 Q3 earnings call, with no new metric, date, or production figure; dropped as a restate.
- Elon Musk reveals plans for a $119 billion AI chip factory in Texas to power Tesla robots and SpaceX (MSN, October 20): the $119 billion total and the Texas location are both already on the public record — May 2026 filings put Terafab phase 1 at $55 billion en route to a $119 billion total, and the Grimes County, Texas site was confirmed August 6; no new tool order, milestone, or date for Optimus-related chip supply; dropped as a restate.
- Icetana plugs AI security into SoftBank's robot rollout (MSN, October 20): a small vendor integration — Icetana's AI surveillance software extending beyond CCTV onto SoftBank Robotics deployments, connecting incident detection with response — with no scale, figures, deployments, or dates in capture; dropped as minor.
- YYForce acquires 20 wearable data collection systems for training humanoid housekeepers (Yahoo Finance / Globe Newswire, October 20): a single-source press release (NASDAQ: YFOR) about wearable data capture for housekeeping tasks, with a Novotel hotel named as a capture location; no robot deployment, customer contract, or robot model named; dropped as announcement-grade PR.
- A guide to Day 1 of RoboBusiness 2026 (The Robot Report, October 20): a schedule for the first day of the Santa Clara, Calif., conference; agenda listing, not news; dropped as non-news.
- OpenAI is building its own humanoid robots, with Sam Altman targeting a first demonstration in 2027 (ynet, October 20): recirculated coverage of OpenAI's humanoid plans, flagged as a stale update with no new detail; dropped as a restate of the item above.
- OpenAI Confirms Humanoid Robot Development—Demo Expected in 2027 (note.com, October 20): further recirculation of Altman's humanoid-robot ambitions and the 2027-demo timeline, flagged in the window itself as a stale update with no new detail; dropped as a restate of the items above. note.com — OpenAI Confirms Humanoid Robot Development—Demo Expected in 2027
- Why Your Robot's Power Supply is the Real Bottleneck (Robotics & Automation News, October 20): a first-person engineering column — an eleven-week warehouse bug traced to a voltage sag, general power-design lessons — with no dated announcement, product, deployment, or named customer; dropped as commentary. Robotics & Automation News — Why Your Robot's Power Supply is the Real Bottleneck
- Inside My Robot Massage: Can Machines Replace Human Hands? (CNET, October 20): a first-person consumer tryout of a massage robot with no dated launch, product release, or announcement; dropped as off-beat. CNET — Inside My Robot Massage: Can Machines Replace Human Hands?
- Gen Z Shoppers Are Four Times More Likely Than Baby Boomers to Buy With Plans to Return (Locus Robotics research, via TMCnet and Chain Drug Review, October 20): a company research release on generational bracketing behavior — 61% of Gen Z and 52% of Millennials have bracketed an apparel purchase versus 37% of Gen X and 15% of Baby Boomers — with no robot, deployment, customer, or automation figure attached; dropped as PR-grade research.
- Investing in 1X | How to Buy Pre-IPO Shares (Securities.io, October 20): an evergreen explainer on buying pre-IPO shares of 1X, the humanoid and embodied-AI developer, with no dated news event, product, or filing; dropped as non-news.
- Humanoid Robot Pilot by AGCO and Agility Robotics Planned for Kansas Manufacturing Plant (The AI Insider, October 20): restates the AGCO–Agility commercial agreement carried in item 6 — same pilot, same Kansas site, same 2027 start; dropped as a restate.
- Smart Hands, Human Touch: Neurotech Tightens Grip On Robotics Market (Forbes, October 20): a commentary column on neurotech prosthetics as robotics' first serious foray, citing Phantom Neuro and BIOS clinical implantations — no dated announcement, funding, or deployment in capture; dropped as commentary.
- Avenda Health expands Unfold AI platform to support intraoperative prostate cancer care (The Robot Report, October 20): a software expansion — patient-specific 3D cancer maps during robotic prostate procedures — with no robot hardware event, deployment, or customer figure; dropped as medical-AI adjacent.
- 2027 TRR50 Robotics Innovation Awards open nominations (The Robot Report, October 20): awards-process scheduling ahead of the Robotics Summit & Expo on June 2–3, 2027 in Boston; dropped as non-news.
- Agility RTs Chris Paxton on Digit 5's leg tradeoff — backwards legs good for speed, humanlike legs good for moving heavy loads from floor to high shelf (X, October 20): recaps the leg-redesign explanation already on the record from September 17 and the October 19 IEEE Spectrum founder essay; dropped as a restate.
- Agility's same-day X housekeeping — the Fortt Knox tweet promoting the CEO conversation carried in item 13, and a cable-harness supply-manager job posting (October 20): amplification and a hiring notice; dropped as non-news.
- Three further AGCO–Agility recaps (letsdatascience and a Yahoo Finance "AGCO, Agility Robotics Sign Deal" version, October 20): same pilot, same Kansas site, same 2027 start; dropped as restates.
- KAIST Humanoid v0.9 re-look (Gadget Review, October 20): the same 35 kg (77 lb) push / 20 kg (44 lb) lift figures as item 4; dropped as a restate.
- OpenAI humanoid recirculations (Ubergizmo, bdnews24, October 20): Altman teasing a 2027 demonstration, no new detail; dropped as restates.
- Faraday Future's FF Nova developer-platform writeup (RobotsBeat, October 20): the $2,490 price was on record October 15–16; the only additions are an optional $390 developer package and limited-invitation-release wording; dropped as announcement-grade pricing detail.
- RoboBusiness rebrands as Robotics Summit & Expo West (The Robot Report, October 20): an event-branding announcement with strategic rationale; dropped as non-news.