Robotics
A genuine engineering bet (tendon-driven safe humanoid + in-home data flywheel) wrapped in a marketing fiction — NEO ships "autonomous" but runs on stranger-in-a-VR-headset teleop, and the company most likely to make it work, OpenAI, just became its competitor.
Research
The verdict
A genuine engineering bet (tendon-driven safe humanoid + in-home data flywheel) wrapped in a marketing fiction — NEO ships "autonomous" but runs on stranger-in-a-VR-headset teleop, and the company most likely to make it work, OpenAI, just became its competitor.
1X Technologies is a Norwegian-American humanoid-robotics company founded May 2014 in Moss, Norway as Halodi Robotics, rebranded to 1X in March 2023. Founder-CEO Bernt Børnich (also CTO) is a Norwegian roboticist; the company is now dual-sited with engineering/HQ shifting to Palo Alto, CA (announced July 2025) while retaining manufacturing roots in Moss. Headcount ~663 employees.
How it makes money — and the pivot. The original, only-real revenue line was industrial: leasing the "EVE" wheeled security/utility robot, anchored by an ADT Security Services lease deal that produced ~$7M revenue in 2022. The whole company has since pivoted to a consumer-first model around NEO, a bipedal home humanoid:
Contract structure / payment terms. Two revenue shapes: (1) hardware sale (one-time $20K), and (2) recurring subscription ($499/mo) — the latter is the strategically important one, because it keeps the robot (and its in-home camera feed) on 1X's books as a data-generating asset, not a sold appliance. The EQT deal is a volume framework, not a take-or-pay commitment — "up to 10,000" is a ceiling, not a floor.
The core mechanic to understand. NEO is sold as an AI home robot but today operates primarily via human teleoperation — a remote 1X operator in a VR headset ("Expert Mode") performs the chores while the onboard AI ("autonomous mode") handles a narrow set of tasks. The business is explicitly a "data flywheel": put robots in homes, have humans pilot them, harvest "continuous embodied experience" to train the policy toward genuine autonomy. The customer is, functionally, paying $20K to be a beta data-collection node.
Map: in-house actuators + motors → 1X assembly (Moss/Hayward) → consumer home or EQT enterprise site, with onboard AI trained on a teleop+sim data loop. Named stakeholders along the chain (grounded in kb/robotics/wiki/supply-chain.md + web):
| Node | Role | 1X's position | Named players |
|---|---|---|---|
| Actuators / motors | Joint torque | Vertically integrated — custom Revo1 cable-driven differential transmission, made in-house | 1X (vs. external Harmonic Drive / Nabtesco / Maxon used by others) |
| Compute (onboard) | Inference | Embedded GPU running models offline | NVIDIA (Jetson/embedded GPU; GR00T N1 integration demoed Mar 2025) |
| Sensors | Perception | Cameras/IMUs; tactile still weak sector-wide | sector: Hesai, Orbbec, tactile research-stage [kb/robotics/wiki/supply-chain.md] |
| Foundation policy | Control | In-house — Redwood VLA + 1X World Model | 1X proprietary (+ NVIDIA GR00T as complement) |
| Teleop data | Training fuel | Proprietary in-home fleet — the moat candidate | 1X operators (VR) + future customer homes |
| Manufacturing | Assembly/test | In-house, automotive-style DFMEA quality | 1X (Moss → consolidating Palo Alto) |
Chokepoints / single-source dependencies:
kb/robotics/wiki/supply-chain.md). This is the one node 1X does not control.kb/robotics/wiki/bottlenecks.md], and China's vertical integration is "dragging the price floor down aggressively."Real, defensible edges:
kb/robotics/wiki/positioning.md].Bargaining power: Weak today. As a pre-scale private burning toward a $1B raise, 1X needs capital, NVIDIA silicon, and customers more than they need 1X. Its only leverage is the OpenAI halo and the safety-design lead.
Moat durability — the hard question. The data-flywheel moat only exists if NEO sells in volume and the teleop-to-autonomy transfer actually works. Neither is proven. And the safety-hardware lead is a 12–24 month lead at best — copyable by a well-funded rival.
No audited segment financials exist (private). Reconstructed /:
| "Segment" | Status | Notes |
|---|---|---|
| Industrial (EVE leasing) | Legacy, ~$7M rev (2022) | The only historically real revenue; being deprioritized for NEO |
| Consumer (NEO sales/subscription) | Pre-revenue → first deliveries 2026 | $20K / $499-mo; no shipped-unit revenue yet |
| Enterprise (NEO via EQT) | Framework signed, not yet deployed | Up to 10,000 units 2026–2030; routes "home" robot to factories |
| Geography | Norway → U.S. (HQ + mfg shift); deliveries U.S. 2026, intl 2027 |
Trend: decelerating-to-zero on the one real (industrial) line, betting everything on a consumer ramp that hasn't started. Production roadmap (management's own, ``): thousands (2025) → tens of thousands (2026) → hundreds of thousands (2027) → millions (2028) — aggressive to the point of implausibility (see Lens 13).
(+private overlay: Phase B reframed — there is no earnings print. Lens 5 → funding & valuation trajectory; Lens 7 → cap table & secondary marks; plus a traction/unit-economics read.)
| Round | Date | Amount | Lead | Participants | Valuation |
|---|---|---|---|---|---|
| Seed/early (as Halodi) | pre-2023 | undisclosed | Norwegian firms | Nistad, Sandwater, Skagerak | n/a — not disclosed |
| Series A2 | Mar 2023 | $23.5M | OpenAI Startup Fund | Tiger Global | $210M post-money |
| Series B | Jan 2024 | $100M | EQT Ventures | Samsung NEXT, OpenAI Startup Fund, Tiger Global | not disclosed |
| Pending mega-round | announced Sep 2025 | up to $1B (target) | n/a | n/a | target $10B+ |
Børnich is unusually candid for a fundraising founder — a sentiment tell in itself. The trend across 2024→2026 interviews:
Tone shift: from visionary ("humanoids will change everything") to damage-control + reframe. The recurring phrase that appeared is "data flywheel"; the thing they'd rather not say out loud is "the robot can't actually do this yet." Sentiment: founder is technically honest but commercially over-promising — a dangerous combination for consumer trust.
Syndicate quality. Mixed-tier. The crown jewel is the OpenAI Startup Fund lead (its first-ever humanoid investment) — a strong IPO-proximity signal historically. Tiger Global (crossover) and Samsung NEXT (strategic) add credibility. But: no Sequoia/a16z/Founders Fund-tier lead, no Fidelity/T. Rowe crossover mark yet, and the syndicate skews European/strategic rather than the Silicon Valley elite backing Figure (Nvidia, Intel Capital, Salesforce, Qualcomm, Parkway). No secondary marks disclosed.
Humanoid peer table (private valuations — ``, last marks):
| Company | Latest valuation | Total raised | Status | Edge |
|---|---|---|---|---|
| Figure AI | $39B (Sep 2025) | ~$1.9B | private | Commercial-first (BMW, UPS); elite syndicate |
| Apptronik | $5–5.5B | ~$1.45B ($935M Series A) | private | Apollo, logistics/mfg; Google/Mercedes ties |
| Agility Robotics | n/a (est. ~$1B+) | $641M+ | private | Digit actually deployed at Amazon/GXO |
| 1X Technologies | $10B target (unconfirmed); ~$500M+ last confirmed | ~$136.5M | private | Consumer-first; safety hardware; OpenAI halo |
| Tesla Optimus | embedded in TSLA | — | public parent | $20K BOM target at automotive scale |
| Boston Dynamics | n/a (Hyundai sub) | — | subsidiary | Atlas R&D pedigree [kb positioning] |
The provenance conflict worth flagging: Humanoid Index lists 1X at "~$500M+, Sequoia among investors" — but Sacra/Contrary/TradedVC do not show Sequoia in the cap table, and the $500M figure is stale vs. the $10B target. I do not treat the $10B as a mark (round unconfirmed closed), and I flag the Sequoia attribution as unverified — likely a data-aggregator error. Where sources conflict, the bear case uses the conservative reading.
Events that have moved 1X's narrative valuation (the private analog of a >5% stock move):
Pattern: the market (private capital) reacts to proximity to OpenAI/NVIDIA and funding leapfrogs by Figure, far more than to NEO's actual capability. The recent shift is that capability scrutiny (does it work autonomously?) has finally started to bite the narrative.
No audited financials exist (private, no SEC filings). A forensic accounting teardown is therefore impossible — which is itself the headline risk: every number in this dossier is company-sourced or press-reported and unaudited. Specific concerns:
Regulatory findings (required sub-section).
regulatory/regulatory-findings.md (0 findings, fetched 2026-06-17)."1X Technologies" (FTC OR DOJ OR FDA OR... ) enforcement) surfaces no formal enforcement action, consent decree, fine, or settlement as of 2026-06-17. The teleop launch drew heavy media criticism and privacy debate but no regulator has filed an action to date.+private overlay; no private-watch.json exists, so reconstructed /)Stage: Late-stage private, pre-IPO and not S-1-ready. Pre-revenue on its flagship, unaudited, and mid-raise on a target (not closed) $10B round.
Milestones that unlock an S-1 (and the realistic window):
Estimated IPO window: not before 2028–2029. Figure is far closer to tradeable (higher valuation, commercial deployments, elite syndicate, IPO chatter for 2026) — 1X is the earlier, cheaper, riskier humanoid bet.
Be-early write-back: No research/private-watch.json exists to update; the +private ledger entry should be created (stage: late-stage private / pre-S-1; ipo_readiness: low; catalyst: $1B round close + first audited NEO shipment revenue; dossier: this file). Flagging for the central post-wave step — I do not edit ledgers in --watchlist.
No Brier EPS forecast logged — pre-revenue private, no EPS line, and forecast.ts create is skipped in the breadth loop per SKILL.
Bull case. 1X is the only serious Western bet on the single largest robotics TAM — the home — with the one thing that market actually requires: a robot safe enough to share a kitchen with (66 lb, soft, tendon-driven, yields on contact, fridge-quiet). The humanoid TAM is enormous and early — Goldman ~$38B by 2035 (250k+ units by 2030, ~70% CAGR), Morgan Stanley $5T by 2050 / 1B units. If even the teleop data flywheel works as designed, 1X compounds an unstructured-environment dataset no factory-bound rival can match, and the $20K/$499-mo model gets it into homes years before competitors target consumers. Vertical integration (Revo1, in-house AI) controls its cost curve; an A-list ex-Tesla/SpaceX/BMW/Google-Brain team is built precisely to crack the manufacturing wall. At a ~$10B target it's a ~4x discount to Figure's $39B for the consumer-market optionality. The earnings/valuation surprise: a credible autonomy demo (NEO doing a real task with zero teleop) would re-rate it overnight.
Bear case (2–3 permanent-impairment risks).
Pre-mortem (18 months out, thesis broke): The $1B round closed smaller/down, or didn't close. NEO shipped a few thousand units to enthusiasts; teleop never meaningfully receded; a viral privacy incident (a leaked in-home feed) or a regulator's inquiry froze consumer demand. Meanwhile OpenAI + Figure soaked up the talent and capital. 1X becomes an acqui-hire — its motors and safety IP bought by a better-capitalized rival.
Are multiples too high? At a $10B target on ~$0 product revenue, this is a pure narrative/optionality mark — by definition "too high" on fundamentals, justifiable only on the home-TAM call and the OpenAI-halo (which is eroding). Contrarian view of what the market refuses to see: the consensus treats teleop as a temporary scaffold to autonomy; the market refuses to price the possibility that for the home, teleop is the product for the next 3–5 years — in which case 1X is a labor-arbitrage staffing company (remote operators piloting hardware) wearing an AI-robotics multiple.
Dismantling the bull case. The way 1X "makes money" structurally breaks the moment you notice it doesn't make money — the one real revenue line (~$7M EVE/ADT) is being abandoned for a consumer product that cannot perform a single chore without a human in a headset. Revenue is concentrated at zero on the thing they're valued for. The "data flywheel" is a bet that imitation-learning from teleop converges to autonomy — an unproven research conjecture the CEO himself undercuts ("AIs completely fail at most reasoning tasks"). The most dangerous competitor bulls underestimate isn't Figure — it's OpenAI, which just turned from patron to rival (June 2026) and owns the model layer 1X depends on; if OpenAI builds the brain, 1X is a contract motor-and-chassis shop. The worst "capital allocation" signal is vertical integration into manufacturing right as China commoditizes the BOM at 40–60% lower cost — they're building a high-cost moat against a falling price floor. For today's ~$10B target to hold, you must assume: NEO ships tens of thousands by 2026 (vs. first deliveries late 2026), teleop recedes fast, no privacy disaster, the $1B closes, and OpenAI doesn't out-build them. If the consumer ramp disappoints by 20–30%, there's no floor — there are no earnings, no audited revenue, and the comparable (a sold appliance vs. a piloted-by-strangers beta unit) collapses. The single scenario that permanently impairs the business: a publicized in-home privacy breach (a teleoperator records or leaks a customer's home), which would make "let 1X's robot into your house" radioactive — plausible at scale, and uninsurable against in reputation terms.
The only profitable humanoid maker and #1 by volume — but the float is a Shanghai-only embodied-AI bet whose revenue is 74% lab demos, not labor replacement, and whose Western TAM is being legislated to zero.
A profitable EV maker priced as a solved-autonomy robotics company — the car business is shrinking, the GAAP profit prop (reg credits) is going to zero, and the entire ~190x multiple now rents on robotaxi + Optimus execution that is real but years behind the price.
A genuinely differentiated case-handling robotics moat wrapped around an ungovernable accounting-and-concentration core — adverse ICFR opinion + active SEC whistleblower probe + >90% Walmart revenue means the multiple is pricing a clean compounder that the filings say does not yet exist.