Figure AI

Phase A — Understand the business

Lens 1 · Company Overview

Figure AI (founded 2022, Sunnyvale CA) builds general-purpose humanoid robots intended to do physical labor in manufacturing, logistics, warehousing, retail, and eventually the home . The product line has iterated fast: **Figure 01** (2022–23 prototype) → **Figure 02** (the workhorse deployed at BMW) → **Figure 03**, the third generation introduced **October 2025**, 5'8", ~61 kg, 20 kg payload, ~5h runtime on wireless charge, 9% lighter than F.02 with 2x faster actuators and home-grade sensing .

The intelligence layer is Helix, an in-house vision-language-action (VLA) model (more below). The manufacturing arm is BotQ, a high-volume facility unveiled March 2025 with a stated first-line capacity of 12,000 robots/year ``.

Business model: Robotics-as-a-Service (RaaS / HaaS) for enterprise first, consumer later (targeting ~2030). Reported pricing is ~$1,000/robot/month covering hardware, software updates, maintenance and support — a recurring, capex-light model for the customer . Separately, the consumer/enterprise *build* cost target is **under $20,000/unit** at volume .

Customers: BMW Manufacturing (Spartanburg, SC — the flagship) and a reported second major customer, widely identified as UPS (logistics), where discussions were ongoing as of 2025 . First revenue: **December 2024 / January 2025** .

Key payment-term caveat (provenance-critical): the scope of the BMW relationship is disputed (see Lens 13). Figure characterizes a multi-robot "fleet"; BMW, in an April 2025 Fortune inquiry, described a single robot during non-production hours ``. Treat all "fleet / commercial-scale" framing as company-sourced and contested.

Lens 2 · Supply Chain

Figure's thesis is vertical integration — it designs and builds its own motors, actuators, sensors, battery systems and compute integration in-house, and (post-Feb 2025) its own AI stack ``. Named stakeholders along the chain:

• Upstream compute / silicon: NVIDIA (GPUs for training + strategic investor); Intel (Intel Capital investor, x86/edge relevance); Qualcomm (Qualcomm Ventures investor — edge inference silicon is a plausible on-robot compute path). These are both suppliers-of-relevance and cap-table members — a recurring Figure pattern ``.
• In-house (the differentiator): actuators, end-effectors, perception stack, battery, and the Helix model — Figure deliberately owns the hard, expensive layers rather than buying subsystems.
• Manufacturing: BotQ, self-operated, with a hybrid workforce (Figure robots help assemble Figure robots) — a vertically-integrated factory rather than a contract-manufacturer (CM) relationship ``.
• Downstream / end customers: BMW (automotive body shop — sheet-metal handling), reported UPS (logistics), and the eventual home consumer.
• Strategic distribution / channel investors: Salesforce, T-Mobile Ventures, LG Technology Ventures, Macquarie, Brookfield — several look like balance-sheet / infrastructure / channel relationships more than pure financial bets ``.

Chokepoints: (1) rare-earth magnets / precision actuators — the universal humanoid bottleneck; not Figure-specific but real. (2) Training data — embodied-AI data is scarce; Figure's "Helix Lab" + fleet data-collection is explicitly meant to attack this ``. (3) Single-customer concentration — BMW is the only proven deployment; UPS is unconfirmed. This is the dominant structural risk (Lens 13).

Lens 3 · Competitive Advantages (moats)

1. Vertical integration + in-house VLA (Helix). Figure's sharpest claim is that owning both the robot and the brain is the only way to "solve embodied AI at scale" — Adcock's stated rationale for firing OpenAI . Helix is reportedly the first VLA to output high-rate continuous control of the full humanoid upper body and to run two robots collaboratively . If real and durable, this is a genuine technical moat; if commoditized by open robot-foundation-models (Physical Intelligence's π-zero, Skild, NVIDIA's GR00T), it erodes.
2. Capital & syndicate. At ~$1.9B raised and a $39B mark, Figure is the best-capitalized pure-play humanoid outside Tesla ``. Capital is a moat in a hardware-iteration race (6-month cycles are expensive).
3. First-mover deployment narrative. A real (if contested-in-scope) 10–11 month BMW production deployment is more than most rivals can show ``.
4. Founder distribution / hype machine. Adcock's ability to raise at 15x in 18 months is itself an asset (and a risk — see bear).

Bargaining power: Today, weak over customers (BMW/UPS are blue-chips Figure needs more than they need it) and moderate over suppliers (strategic investors are aligned). Power flips only when there's a multi-customer order book.

Lens 4 · Segments

n/a — private, not disclosed. No segment P&L exists. Qualitatively, 100% of proven commercial revenue is automotive (BMW), with logistics (UPS) as a reported pipeline and home/consumer as a 2026–2030 ambition. Revenue began Dec 2024–Jan 2025; no breakout is public ``. The "segment" story is really a single-deployment story — which is itself the most important fact in this dossier.

---

Phase B — Measure performance

Lens 5 · Funding & valuation trajectory (+private — swaps "Earnings Result")

The valuation curve is the performance metric for a pre-revenue-scale private. Round history [all web, unaudited]:

Sources: . **Total raised ≈ $1.9B** .

Reads:

• The 15x step-up in 18 months is among the steepest in the private market — it prices category leadership and a scaling fleet, not current cash flows. ``: at ~$1.9B raised vs a $39B post, Figure has sold roughly ~5% cumulative of the company to outside capital across all rounds — i.e., founder/employee ownership is unusually high for a late-stage name, which preserves Adcock's control but also concentrates key-man risk (Lens 9/13).
• Use of proceeds (Series C): scale the robot fleet, build training infrastructure, accelerate data collection `` — i.e., the money is going into the thing that isn't proven yet (multi-site scale), which is the right place but also the unproven place.
• Burn signal: n/a — not disclosed, but a >$1B raise to fund 6-month hardware cycles + a 12k-unit factory + a frontier AI lab implies a very high burn; the Series C is explicitly framed as multi-year runway that removes near-term IPO pressure ``.

Lens 6 · Founder interviews / sentiment trend (carried — replaces "Earnings Calls")

No earnings calls exist. The sentiment signal comes from Adcock's public channel (X) and interviews (Shawn Ryan Show, etc.):

• Feb 2025 — the inflection. Adcock: "Today, I made the decision to leave our Collaboration Agreement with OpenAI. Figure made a major breakthrough on fully end-to-end robot AI, built entirely in-house" and, later, "we were running circles around OpenAI… way better at this" ``. Tone: maximally confident, combative, control-oriented.
• Apr 2025 — defensive/litigious. In response to Fortune's BMW reporting, Adcock pivoted to threatening defamation litigation rather than disclosing contract specifics ``. Tone shift: from offense to defense; chose escalation over transparency.
• Jun 2025 — evasive onstage. At a tech conference Adcock skipped a live demo and sidestepped questions on the BMW commercial relationship ``. For an analyst, declining a live demo during a fundraise is a yellow flag.
• Oct 2025 onward — execution cadence. Figure 03 launch + BotQ ramp messaging returned to product-momentum framing ``.

Net: the founder narrative oscillates between visionary confidence and defensiveness-under-scrutiny. The things he stopped saying — specifics on BMW's contractual scope and unit volumes — are exactly the things a skeptic most wants.

Lens 7 · Cap-table & secondary marks (+private — swaps "Comps")

Syndicate quality (the IPO-proximity tell):

• Tier-1 strategics — strong: NVIDIA, Microsoft, Intel, Qualcomm, Salesforce, LG, T-Mobile. A deep bench of strategic corporates — unusually strong, signals real industrial interest and channel optionality ``.
• Marquee financial names — present but infra-flavored: Brookfield and Macquarie (infrastructure / real-asset managers), Parkway (lead, both B and C), Bezos Expeditions, ARK ``.
• The notable absence (analyst flag): no marquee public-markets crossover (Fidelity / T. Rowe / Wellington / Coatue / Tiger) is named in the Series C syndicate ``. Crossover entry is the classic "IPO within ~18 months" signal; its absence here suggests the round was driven by strategics + private capital, consistent with an IPO that is not imminent (2027–28 window, Lens 11). This is the single most useful cap-table read for IPO-timing.

Secondary marks / mutual-fund markups: n/a — private, not disclosed in primary sources. Secondary-market trackers (Sacra, Hudson Point, IPO Club) quote interest in Figure shares but no audited mark; treat any secondary price as indicative only ``.

**Peer "comp" by valuation-per-deployment (the honest cross-check), :** Figure carries the **highest valuation of any pure-play humanoid** ($39B) on **one proven customer**. Apptronik (~$431M raised, Mercedes + GXO pilots) and Agility (~$178M raised, GXO/Digit warehouse deployments) trade at small fractions of Figure's mark with arguably *comparable or broader* pilot footprints . Figure's premium is a bet on Helix + BotQ + Adcock, not on a deployment lead that is 20x larger.

Lens 8 · Funding & product catalysts (carried — replaces "Stock-Price Catalysts")

Events that re-rated the private mark or the narrative [all web]:

• Feb 2024 — Series B / $2.6B / OpenAI partnership announced → category-defining round [CNBC].
• Feb 2025 — OpenAI split + Helix reveal → the defining strategic pivot; re-cast Figure as a full-stack player [Adcock X / Maginative].
• Mar 2025 — BotQ unveiled (12k/yr) → manufacturing credibility [figure.ai].
• Apr–Jun 2025 — Fortune BMW dispute + onstage demo-skip → the credibility drawdown; the bear case crystallized here [Fortune / TechCrunch].
• Sep 2025 — Series C / $39B → the market re-rated through the controversy; capital validated the story [TechCrunch].
• Oct 2025 — Figure 03 launch + production-ramp messaging [figure.ai].
• Nov 2025 — BMW Spartanburg 10–11 month deployment completed: 30,000+ X3s, 90,000+ parts, 1,250 operational hours, >99% placement accuracy, 84s cycle time [BMW press / Repairer Driven News / figure.ai, Nov 2025].
• Feb 2026 — BMW announces Plant Leipzig (Germany) expansion — first European humanoid deployment [BMW press, 2026-02-27]. This is the most recent positive datapoint and a partial rebuttal to the "it's only a pilot" bear.

Pattern: Figure's marks move on funding + product launches + the BMW relationship, and the BMW signal cuts both ways — it is simultaneously the strongest bull proof point and the locus of the sharpest bear attack.

---

Phase C — Judge people & books

Lens 9 · Management (founder archetype)

Brett Adcock (b. 1986) — founder/CEO. The single most important variable in this name.

• Track record (quantified, genuinely strong): Built Vettery (talent marketplace, 2012) → sold to Adecco for $110M (2018). Founded Archer Aviation (2018, eVTOL) despite no aviation background → raised $1B+, took it public at ~$2.7B, flight-tested 5 aircraft generations, signed a $1.5B United Airlines deal ``. He has taken a hard-tech company public before — directly relevant to Figure's eventual S-1.
• Skin in the game: Self-funded Figure's ~$100M seed; founder/employee ownership is unusually high given how little equity was sold (`` from Lens 5). High alignment, high control.
• Capital-allocation instinct: Aggressive, vertical-integration-maximalist (build everything in-house, including firing your AI partner to do it yourself). Bold; capital-intensive; high-variance.
• Red flags: (1) Promotional intensity — the BMW "fleet" framing vs BMW's own description (Lens 13) suggests a founder who markets ahead of the verifiable facts. (2) Litigation-as-PR — threatening to sue Fortune rather than disclosing specifics. (3) Demo-skip under scrutiny (TechCrunch, Jun 2025). These are style risks, not accounting risks — but for a name whose entire valuation rests on believing forward claims, founder credibility is the audit.
• Archetype: Classic high-velocity serial founder-CEO with a real exit pedigree and a documented willingness to oversell. The Archer comparison is the tell: he did deliver a public hard-tech company — but Archer's post-IPO commercialization has been slower than the pre-IPO promise, a pattern worth holding in mind for Figure.

Lens 10 · Forensic / governance red flags (re-pointed — no financials to audit)

With no audited statements, "forensic" here means claim-integrity, governance, and disclosure risk:

• Claim vs. independent confirmation gap (the central red flag): the BMW deployment's scope is the one place where Figure's public claims met independent reporting — and they diverged (Adcock's "fleet" / "end-to-end operations" vs BMW's "single robot, non-production hours" per Fortune's Apr 2025 inquiry) . BMW's later official numbers (30k cars, 1,250 hours) describe **one robot over ~10 months**, which is *consistent with the skeptical read*, not the "fleet" read .
• Revenue recognition: n/a — private. But note RaaS at ~$1k/robot/month implies that even a generous read of deployed units produces trivial revenue relative to a $39B mark (Lens 11 arithmetic).
• Related-party texture: many suppliers (NVIDIA, Intel, Qualcomm) are also investors — common in deep-tech, not inherently problematic, but it means some commercial validation is from aligned parties.
• Governance: founder-controlled, very little external equity sold → limited outside-board check on a promotional founder. A pre-IPO governance build-out (independent directors, audited financials, a CFO with public-company experience) would be a maturation signal to watch for.

Regulatory / legal findings: No SEC filings exist (private). Web search surfaced no material government enforcement (FTC/DOJ/SEC) against Figure. The only live legal item is Figure-as-plaintiff — Adcock's threatened defamation action against Fortune `` — which, if filed, is a reputational/distraction risk, not a liability. No material regulatory or legal findings against the company as of 2026-06-15, per web search (no SEC EDGAR record — private); unaudited.

---

Phase D — Project & stress-test

Lens 11 · IPO-readiness & path-to-tradeable (+private — swaps "Forward EPS Projection") — THE BE-EARLY LENS

Anchor (private-watch.json): stage late, ipo_readiness 4/5 (pre-IPO / secondary-active), catalyst = "humanoid commercial pilots (BMW); mega-round." Both halves of that catalyst have now occurred (BMW deployment completed Nov 2025; $39B mega-round closed Sep 2025) — yet readiness stays at 4, not 5, because the scale proof and the S-1 mechanics aren't there.

Distance to a tradeable event — ``: ~18–36 months (a 2027–2028 listing window). Reasoning:

• No S-1, no underwriters, no price range, no date as of mid-2026 ``.
• Adcock has explicitly deprioritized an IPO — "revenue first," and the Series C gives multi-year runway, removing the need to list ``. A founder who controls the company and doesn't need cash will list late and on his own terms.
• No public-markets crossover on the cap table (Lens 7) — the usual ~12–18-month-pre-IPO signal is absent, arguing against a 2026–early-2027 listing.

What would trigger an S-1 (the milestones to watch — this is the be-early checklist):

1. Multi-customer order book — a second confirmed paying customer at scale (UPS conversion from "considering" to "deployed fleet") breaks single-customer concentration and is probably the gating commercial milestone.
2. BotQ at real volume — visible production toward the 12,000-units/year line (and units actually deployed, not just buildable).
3. A revenue figure worth printing — RaaS revenue that's material enough to underwrite a public valuation. `` reality check: at ~$1,000/robot/month, 1,000 deployed robots = ~$12M ARR; 10,000 robots = ~$120M ARR. Even the optimistic case is ~$120M ARR against a $39B mark = ~325x ARR — a multiple that requires the market to underwrite the 2030 "millions of robots" TAM, not current economics. The IPO becomes feasible when the growth slope of that ARR is steep enough to defend the number.
4. Governance build-out — public-company CFO, audited financials, independent board.
5. A crossover round — a Fidelity/T. Rowe/Coatue mark would be the clearest "S-1 within ~18 months" tell.

Write-back: updating research/private-watch.json → set figure-ai.dossier to this file's path and keep readiness at 4 (mega-round + BMW pilot both banked, but scale + S-1 mechanics absent → not yet 5). (Manual edit recommended; the local dossier is the deliverable per task scope.)

Lens 12 · Bull vs Bear

Bull case. Figure is the best-capitalized, most vertically-integrated pure-play in the single largest robotics TAM imaginable (general physical labor). It owns the full stack — robot and Helix VLA — which, if embodied-AI rewards integration the way Adcock argues, is the winning architecture. It has a real, completed, blue-chip industrial deployment (BMW, now expanding to Europe), a founder who has taken hard-tech public before, a 12k/yr factory standing up, and a strategic syndicate (NVIDIA/Microsoft/Intel/Qualcomm) that doubles as channel and compute. If humanoids reach the "$50k-to-match-human-labor" threshold this decade, Figure is the Western incumbent and $39B is an entry price, not an exit price.

Bear case (2–3 permanent-impairment risks).

1. The valuation prices a fleet that is, on the public record, a pilot. $39B / one proven customer / ~$120M ARR even in a generous case = a story stock where any slip in the scale narrative re-rates hard.
2. Helix could be commoditized. Robot-foundation-models from Physical Intelligence (π-zero), Skild, NVIDIA (GR00T), and Tesla's data flywheel could erode the in-house-AI moat — and firing OpenAI removed a hedge.
3. Tesla. Optimus targets ~$20k and rides Tesla's manufacturing scale + data; if Tesla makes humanoids a commodity, Figure's premium evaporates ``.

Pre-mortem (18 months out, thesis broke): UPS never converts to scale; BotQ ships far fewer than 12k units; a Tesla/Chinese (Unitree) cost-leader undercuts the RaaS economics; Adcock's credibility takes another hit (a second Fortune-style exposé); the next round is a flat or down mega-round, the secondary market marks Figure 30–40% lower, and the "highest-valued humanoid" becomes the cautionary tale of the cycle.

Contrarian view (what the market refuses to see): The bull consensus treats $39B as cheap optionality on a trillion-dollar labor market. The contrarian read is that the BMW "completion" actually confirmed the skeptics — one robot, ~10 months, non-production-grade scope — and the market re-rated up anyway, on narrative and capital momentum. The thing nobody is pricing: Figure's biggest risk isn't a competitor, it's a credibility event from its own founder during the pre-IPO window.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration is ~total. One proven customer (BMW), one reported prospect (UPS, unconfirmed). If BMW doesn't expand to a true fleet and UPS stalls, the commercial story is a single deployment dressed as a platform. ``.
• The scope dispute is the smoking gun. BMW told Fortune: one robot, non-production hours. Figure marketed: fleet, end-to-end operations. BMW's own Nov 2025 numbers (one robot, 1,250 hours over ~10 months) corroborate the skeptic, not the founder. A company whose CEO threatens to sue a reporter rather than show the contract, and skips a live demo at his own fundraise (TechCrunch, Jun 2025), is exhibiting the classic pattern of a name where the deck outruns the deployment.
• The moat may be thinner than bulls think. Helix is impressive in demos; demos aren't a moat. Open robot-foundation-models are converging fast, and the hardest part — reliable, generalized real-world manipulation at low cost — is unsolved by everyone, including Figure.
• Most dangerous competitor bulls underestimate: not Tesla (priced in) but Chinese cost-leaders (Unitree) and the open-model players (Physical Intelligence/Skild) that could collapse both the hardware price and the AI differentiation simultaneously.
• Worst capital-allocation / governance moves: firing OpenAI removed an AI hedge; minimal external equity sold means minimal independent board oversight of a promotional founder.
• Assumptions that must hold for $39B: (a) Helix stays meaningfully ahead; (b) BotQ actually ships ~12k units; (c) a second large customer converts; (d) RaaS unit economics work at <$20k build cost; (e) no founder-credibility blowup. If commercial scale disappoints by 20–30%, there is no earnings floor — it's a venture mark that re-rates to the next-best private comp (Apptronik/Agility scale), implying substantial downside to the secondary mark.
• Single scenario that permanently impairs: a Tesla Optimus (or Chinese) credible ramp at ~$20k while Figure is still pre-scale — Figure becomes the expensive, sub-scale Western alternative, and the down-round cascade begins. Plausibility: moderate and rising.

Lens 14 · Management Questions (15, ordered by information value)

1. BMW, precisely: how many Figure robots are in production-hours service today, under what contract term, and what is the committed expansion path to Leipzig — in units and dollars?
2. Who is the second paying customer, is it under contract (not LOI), and at what deployed-unit count? (UPS — confirm or deny.)
3. What is current deployed-unit count and ARR, and what is the unit count you underwrite the $39B valuation against?
4. At a <$20k target build cost and ~$1k/month RaaS, what is the gross margin per robot-month at scale, and what utilization/reliability is required to clear it?
5. How many units has BotQ actually produced and shipped (vs. nameplate 12k/yr capacity), and what is the current yield?
6. Post-OpenAI, how do you sustain a Helix lead against open robot-foundation-models (π-zero, GR00T, Skild) — what is proprietary and non-replicable?
7. What is your answer to Tesla Optimus at ~$20k — where do you win when a vertically-integrated automaker undercuts you on price and data?
8. What is your monthly burn and cash runway, and what milestones must hit before the next raise — and would you accept a flat/down round?
9. Why skip the live demo in June 2025, and will you commit to independent, third-party-verified deployment audits going forward?
10. On the Fortune dispute — will you release the BMW deployment terms (redacted) to settle the "fleet vs. single robot" question definitively?
11. What is your path and timeline to an S-1, and what triggers it — a revenue threshold, a customer count, or a market window?
12. When do you bring on a public-company CFO and independent board, and audited financials?
13. What is the real-world reliability (MTBF, intervention rate) of Figure 03 in an 8-hour shift, unstaged?
14. Consumer/home by ~2030 — what are the safety, liability, and unit-economics gates, and how is that not a decade of capex away?
15. What is the one assumption in your plan that, if wrong, breaks the company — and how are you de-risking it?

---