Agility Robotics

Phase A — Understand the business

Company Overview

Agility Robotics builds Digit, a 5'9", ~140 lb bipedal (legged, human-form) robot designed for one job: moving totes and boxes around warehouses and factory floors — tote loading/unloading, stacking, de-stacking, palletizing, and bulk-tote handling in spaces built for humans. The company was founded in November 2015 as a spin-out of Oregon State University's Dynamic Robotics Laboratory; founders Jonathan Hurst, Damion Shelton, and Mikhail Jones (Hurst and Shelton met as PhD students at Carnegie Mellon).

The bet that distinguishes Agility from the rest of the humanoid field is commercial focus over capability theater. While Tesla, Figure, and 1X chase general-purpose autonomy and consumer-facing demos, Agility deliberately narrowed Digit to a near-term, dull, high-frequency logistics task and got it billing. As of 2026, Digit is the only bipedal robot generating revenue from paying commercial customers — it won The Robot Report's inaugural RBR50 "Robot of the Year," and has moved over 100,000 totes in live commercial operations.

Business model. Dual-track, with RaaS (Robots-as-a-Service) as the dominant motion: a recurring monthly fee bundling the robot, the work-cell hardware, and the software, benchmarked explicitly against fully-loaded human warehouse labor at $30/hour. The alternative is outright CapEx purchase ($250K sticker, ~$150K with integration support for direct buyers ) plus a software subscription. The connective tissue is Agility Arc, a cloud SaaS layer (launched 2024) that maps facilities, configures workflows, monitors uptime/throughput KPIs, integrates with WMS/WES/MES/PLC systems, and coordinates autonomous charging — letting a customer stand up a deployment "within days". Arc is the recurring-revenue and lock-in vehicle; the robot is the razor, Arc is the blade.

Customers (the actual logo list, not pilots): Amazon (tote recycling, testing since 2023, via the Amazon Industrial Innovation Fund which is also an investor), GXO Logistics (industry-first multi-year RaaS deal, June 2024, Spanx facility in Georgia), Schaeffler Group (German motion-tech firm — both customer and minority investor, committed to global plant deployment), Toyota (Canada manufacturing plant, RaaS, announced Feb 2026), Mercado Libre (San Antonio TX facility, 2026), and Ford (the original 2020 partner that took the first two production units). The customer set skews toward logistics and automotive manufacturing — repetitive material-handling in environments that already run cobots and conveyors.

Key payment terms. RaaS = recurring monthly, cancellable, OpEx-classified — which is the whole sales pitch (converts a $250K capex decision into a "hire a worker" operating-expense decision). This is recurring and sticky once Arc is integrated into a customer's WMS, but it is not take-or-pay and there is no disclosed multi-year minimum volume commitment publicly — concentration risk is real (see Lens 13).

Supply Chain

Mapping upstream → Agility → end customer, with named stakeholders [grounded in robotics KB supply-chain.md + web]:

Upstream inputs:

• Actuators / motors (joint torque): the cost-and-performance chokepoint of the entire humanoid industry. Western builders source from Harmonic Drive (Japan, strain-wave gearing), Nabtesco, and Maxon; Chinese builders (Unitree) vertically integrate to crush cost [KB: supply-chain.md]. Agility's actuator strategy is a mix of bought and in-house; its OSU heritage is in legged dynamics, so the leg actuation is a genuine in-house competency.
• Compute (onboard inference): NVIDIA Jetson / Thor-class modules — the on-device concentration point shared by the whole field [KB: supply-chain.md]. NVIDIA is also an investor (NVentures).
• Sensors: LiDAR, IMUs, cameras; Digit v4 carries integrated LiDAR. Tactile sensing remains research-stage industry-wide — a named bottleneck.
• VLA / foundation policy: the KB positioning matrix flags Agility's control-model approach as "Partner" rather than fully in-house [KB: positioning.md] — i.e. Agility is strongest at the embodiment + deployment layer and leans on the broader robot-learning ecosystem for the most general manipulation policies. This is a strategic exposure as foundation-policy players (Physical Intelligence, Skild, NVIDIA GR00T) become the brains of the stack.

Midstream (Agility): final assembly at RoboFab, Salem, Oregon — a 70,000 sq ft facility, the world's first purpose-built humanoid factory, ramping from ~8 units/shift in year one toward a 10,000 units/year peak; 500+ workers at full capacity.

Downstream (customers): Amazon, GXO, Schaeffler, Toyota, Mercado Libre, Ford — warehouse and factory operators.

Chokepoints / single-source dependencies:

1. High-torque compact actuators — partially substitutable, but Western builders pay a premium while Chinese vertical integration drags the price floor down [KB: bottlenecks.md].
2. On-device compute — NVIDIA concentration; same dependency as every peer.
3. Fleet-scale teleop/operating data — low substitutability; this is the moat candidate (see Lens 3). Agility's edge here is that its fleet is actually running in production (100k+ totes), so it accrues real operating hours, not demo hours.

Competitive Advantages (moats)

Per the KB positioning matrix, Agility's stated moat is "First commercial humanoid deployed (Digit at Amazon/GXO)" and its stated vulnerability is "Slower form-factor generalization" [KB: positioning.md]. Both are correct and they are two sides of the same coin.

Durable moats:

• Real-deployment data + operational learning. The single most defensible asset. Teleop/operating data does not transfer between companies — it's locked in per builder [KB: positioning.md, "switching costs"]. Agility is the only one accumulating production hours on a paid fleet doing a repeatable task. 100,000 totes is a data moat the demo-stage competitors structurally cannot have yet.
• Legged-locomotion IP & pedigree. A decade of OSU/CMU dynamic-walking research (ATRIAS → Cassie → Digit) is genuine, hard-won process IP. Bipedal balance under load is non-trivial; this is real engineering depth, not a wrapper.
• Arc software switching costs. Once Arc is wired into a customer's WMS/WES and the work-cell is commissioned, ripping it out for a competitor means re-integration and re-training. This is the classic enterprise-software moat layered onto hardware.
• Manufacturing readiness. RoboFab is a real, operating factory while most rivals are still hand-building units. Being first to manufacturing scale is a moat if (and only if) demand materializes.

Bargaining power. Weak-to-moderate over customers — its marquee buyers (Amazon, Toyota, GXO) are giants who can fund or buy alternatives, and Amazon is simultaneously an investor and developing its own warehouse robotics (it acqui-hired Covariant's manipulation team) [KB: positioning.md]. Over suppliers: weak on actuators and compute (NVIDIA holds the on-device compute card), stronger on the parts it makes in-house.

The moat's soft spot: "first commercial" is a timing moat, not a structural one. It compounds only if Agility converts the data and head-start into either (a) decisively better unit economics or (b) faster generalization to new tasks before Figure's capital and Unitree's price advantage close the gap.

Segments

No segment financials exist — Agility is private and our figures is empty. Reconstructing the de facto business segmentation from public data:

Geography: primarily North America (Oregon HQ, US/Canada deployments) with a European thread via Schaeffler (Germany). No Asia-Pacific commercial presence disclosed — which matters, because APAC is where the cheap Chinese competition is strongest.

Trend: the mix is shifting from pure-logistics pilots toward multi-customer manufacturing deployments through 2026 (Toyota, Mercado Libre added on top of GXO/Amazon/Schaeffler) — i.e. early evidence of horizontal expansion across the material-handling use-case, which is exactly the test of whether the "first-mover" timing moat converts into a durable franchise. Revenue/EBITDA by segment: n/a — private, not disclosed.

Phase B — Measure performance

+private overlay: Lens 5 → Funding & valuation trajectory; Lens 7 → Cap table & secondary marks; plus a Traction & unit-economics read. Lens 6 → founder/exec interviews; Lens 8 → funding/product events.

Funding & Valuation Trajectory (+private swap)

Total raised: ~$641M to date.

Key facts and one conflict to flag:

• The Series C is the defining event: ~$400M at ~$1.75B pre-money / ~$2.1–2.2B post. Sources disagree on whether it closed in March 2025 (Sacra/TSG) or September 2025 (Contrary) and on the exact post-money ($2.1B vs $2.2B) — I surface this rather than pick one. The $1.75B pre-money figure is consistent across sources and is the reliable anchor.
• Burn signal: the April 2024 layoff of 150 people (20% of staff), framed as cutting roles "not central to core product development and commercialization". A 20% cut 18 months before a $400M raise tells you the company hit a cash-discipline moment and re-focused hard on the commercial wedge — a positive signal about discipline, a negative one about how capital-hungry humanoid R&D is. Headcount is ~294.

Cap Table & Secondary Marks (+private swap)

Syndicate quality — the IPO-proximity read. The cap table is strong but not blue-chip-crossover-grade:

• Strategic / corporate investors dominate the signal: Amazon (Industrial Innovation Fund — also the #1 customer), NVIDIA (NVentures — also the compute supplier), Sony Innovation Fund, Schaeffler (customer + investor). This is a strategically endorsed cap table — the people who would use or supply the robots are funding it. That's a quality tell for product-market fit.
• SoftBank entered at Series C — a heavyweight late-stage name.
• Deep-tech VCs: DCVC and Playground Global from inception (per private-watch.json: "DCVC, Playground" are the named leads).
• The missing signal: I find no disclosed crossover-fund entry (no Fidelity / T. Rowe / Coatue / Wellington). Those are the funds whose presence usually telegraphs IPO-proximity, and their absence at a $2.1B mark is notable — it's consistent with this name being closer to a strategic-acquisition outcome than a near-term IPO (see Lens 11).

The single most important cap-table data point: SoftBank reportedly explored acquiring Agility outright at "at least $900M" in late 2024 — Agility declined and raised the Series C at $1.75B pre instead, and SoftBank then participated in that round. This is a double-edged signal: bullish that management held out and roughly doubled the implied value in ~12 months; bearish that the most plausible exit on the table was an acquisition under $1B, and a strategic acquirer's reservation price is a real-world valuation anchor far below the financing mark.

Secondary marks: n/a — not disclosed. No mutual-fund markups/markdowns or secondary-market prints found publicly; third-party private-market trackers (Premier Alts, TSG) cite the ~$2.1B primary mark, not independent secondaries.

Founder / Executive Signal (+private swap — interviews, not earnings calls)

What management is publicly focused on, from CEO commentary and press:

• Peggy Johnson (CEO) consistently frames the pitch in labor-economics terms — comparing Digit's hourly RaaS rate to a ~$30/hr fully-loaded warehouse worker, and pointing at the ROI-under-2-years payback. The messaging has shifted notably away from "look what our robot can do" toward "here is the unit economics of replacing a task." That is a commercialization-stage tone, not a research-stage one — and it's the right tone for the only player with paying customers.
• Recurring themes: safety certification (ISO functional-safety / human-collaboration clearance, ~18 months out), deployment speed (Arc → "live in days"), and the operating-ratio improvement (v4's ~2:1 operator-to-robot ratio → v5's targeted ~10:1). The operating-ratio metric is the tell to watch: it's the difference between "subsidized teleop demo" and "autonomous labor."
• What they've stopped saying: less about moonshot general-purpose autonomy, more about totes-per-hour and uptime. Healthy for a company that needs to bill, but it also concedes the narrative high ground (and the highest valuations) to Figure and Tesla.

Catalysts / Stock-Equivalent Events (+private swap — funding & product events)

For a private name, the "price-moving" events are funding rounds, marquee customer wins, and product milestones. The pattern over the last ~24 months:

• June 2024 — GXO industry-first RaaS deal: the legitimizing commercial event.
• April 2024 — 20% layoff: the discipline/burn event.
• Late 2024 — SoftBank acquisition approach (~$900M), declined.
• 2025 — ~$400M Series C at $1.75B pre: the scale-capital event.
• Oct 2025 — Digit v4 in market; v5 announced (50 lb payload, 10:1 ratio).
• Feb 2026 — Toyota Canada deployment; Mercado Libre commercial agreement; "100,000 totes moved" milestone; RBR50 Robot of the Year.

What the pattern reveals: the market (private capital + customers) reacts to proof of real commercial work and named blue-chip logos, and discounts demo virtuosity. Agility's value inflects on deployment counts and operating ratio, not on dexterity demos — the opposite of how Figure's valuation moves.

Traction & Unit Economics (+private add)

The economic crux: if the RaaS rate holds near $30/hr while operating cost falls from ~$11 toward ~$3, the gross-margin structure is genuinely attractive (60% → ~90%). The bear's entire case is that the $30 rate will not hold once $16K Chinese hardware and cheaper Western rivals arrive — i.e. the spread compresses from the price side, not the cost side (Lens 13).

Phase C — Judge people & books

Management

Peggy Johnson (CEO, since March 2024). The defining hire. Ex-Magic Leap CEO (where she pivoted a flailing consumer-AR company to enterprise and recapitalized it) and before that 6 years as EVP of Business Development at Microsoft, reporting to Satya Nadella. Named to TIME's 100 Most Influential in AI 2025. Track record read: she is a commercialization-and-partnerships operator parachuted in precisely when Agility pivoted from R&D to revenue — exactly the right archetype for this stage, and a notable signal that the board prioritized go-to-market and capital-raising over founder-led product vision. The Magic Leap recap experience is directly relevant to steering a capital-hungry hardware company.

Jonathan Hurst (co-founder, Chief Robot Officer / Chief Technology heritage). The scientific soul — OSU professor, the ATRIAS → Cassie → Digit lineage of dynamic legged locomotion is his life's work. Skin in the game and irreplaceable domain credibility live here. The "Chief Robot Officer" role was formalized to keep him on the deep-tech vision while professional managers run the business.

Damion Shelton (co-founder). Founding CEO 2015→March 2024, stepped down to Chairman / public-policy focus when Johnson arrived — a clean, non-acrimonious founder-to-professional-CEO transition (he stayed).

Pras Velagapudi (CTO). Decades in robotics planning/control; ex-VP Eng & Chief Architect of Mobile Robotics at Berkshire Grey before joining Agility (2023). Real warehouse-robotics depth.

Red flag / continuity watch — executive churn. Melonee Wise — a genuine robotics luminary (Fetch Robotics founder/CEO, 2022 Engelberger Award winner) — joined as CTO in 2023, moved to CPO in the May-2024 reshuffle, and departed at the end of August 2025. Losing a figure of Wise's stature inside ~2 years, amid the broader 2024 leadership reshuffle and the 20% layoff, is the clearest organizational-instability signal in the file. (Note: a separate later web result placing Wise at KUKA is a downstream move and does not change the Agility read.) The C-suite has turned over substantially in 24 months — normal for a scaling startup, but worth weighting.

Capital-allocation history: disciplined under pressure — the 2024 layoff + refocus, then declining a sub-$1B acquisition to raise at ~2x, is a defensible sequence. No value-destructive M&A or related-party patterns surfaced. Founder-vs-professional: now firmly professional-manager-led (Johnson) with founders in vision/chair/policy roles — appropriate for the commercialization stage, with the usual cost that the original product zealotry is one step removed from the CEO chair.

Forensic Red Flags

No audited financials exist — Agility is private, files nothing with the SEC, and our figures is empty. So classic forensic-accounting analysis (revenue-recognition aggressiveness, receivables vs revenue, SBC flattering non-GAAP, goodwill) is not possible — there is no income statement, balance sheet, or cash-flow statement to interrogate. State plainly: n/a — private, unaudited.

What can be flagged as analytical risk-of-overstatement:

• "Units sold" vs "units deployed and billing." ~100 units "sold" is not the same as ~100 robots autonomously earning RaaS revenue. With a current ~2:1 operator-to-robot ratio, a meaningful share of "deployment" is human-supervised. The honest metric is autonomous operating hours, which is undisclosed — treat headline deployment counts with caution.
• RaaS revenue quality. Recurring and cancellable; no disclosed take-or-pay minimums → revenue durability is unproven across a downturn.
• Valuation vs strategic reservation price. The $2.1B financing mark sits ~2.3x above SoftBank's ~$900M acquisition interest — a gap the bear will press.

Regulatory findings (required sub-section). Per the pre-fetched regulatory/regulatory-findings.md (2026-06-17): Agility Robotics has no CIK and is not an SEC filer, so no SEC Litigation Releases or AAERs are possible or found (total SEC findings: 0). Non-SEC web search ("Agility Robotics" (FTC OR DOJ OR FDA OR CFPB OR consent decree OR settlement OR fine OR penalty) enforcement) surfaced no material enforcement actions, fines, or consent decrees. The one regulatory theme (not an action) is functional-safety certification: there is no uniform safety standard for bipedal robots operating near humans, and Digit is pursuing ISO functional-safety / human-collaboration clearance, expected ~mid-to-late 2026 — a forward compliance dependency, not a violation. Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER → not applicable, no CIK), web search, and the absence of any public enforcement record, as of 2026-06-18. Unaudited per public sources.

Phase D — Project & stress-test

IPO-Readiness & Path-to-Tradeable (+private swap)

From research/private-watch.json: stage: late · ipo_readiness: 3/5 ("late-stage," on a scale where 4 = pre-IPO/secondary-active and 5 = S-1 filed) · catalyst: "Digit warehouse deployments (GXO)" · lead investors DCVC, Playground.

Assessment: a genuine late-stage company, but NOT IPO-imminent — and arguably more likely to exit via acquisition than IPO. The reasoning:

• No crossover-fund presence (no Fidelity/T. Rowe/Coatue) — the usual IPO-runway tell is absent at the $2.1B mark.
• SoftBank already tabled an acquisition at ~$900M and remains on the cap table — a strategic acquirer is circling, and SoftBank's robotics ambitions make Agility a logical bolt-on.
• Revenue is almost certainly sub-scale for a public listing — ~100 units and undisclosed (likely low-tens-of-millions) ARR is far below IPO threshold; the company needs the RoboFab ramp to land first.

Milestones that unlock an S-1 (the be-early checklist):

1. ISO safety certification (~mid/late-2026) — removes the human-collaboration constraint and unlocks dense deployments. The single most important near-term catalyst.
2. Operating ratio → ~10:1 (v5) — the proof that the fleet is autonomous labor, not subsidized teleop.
3. RoboFab volume ramp toward thousands of units/yr with the $30 RaaS rate intact — proof the unit economics survive scale.
4. A second flywheel customer at GXO/Amazon scale committing to volume.

Estimated tradeable window: 2028+ for an IPO under base assumptions; an acquisition could happen any time (a strategic — SoftBank, an automotive OEM, or a logistics giant — is the more probable liquidity event).

Write-back: this dossier should update the private-watch.json entry's dossier field (currently null) to point here, marking Agility dossier-warm. (Per wave rules, the file edit is left to the central post-wave step — flagged here for the operator.)

Bull vs Bear

Bull case. Agility owns the one thing money can't instantly buy in this race: a robot that paying blue-chip customers actually run in production. 100,000 totes moved, RBR50 Robot of the Year, a real factory (RoboFab) while rivals hand-build, and a recurring-revenue Arc layer that compounds switching costs. The unit economics are seductive — a $30/hr RaaS rate against an ~$11/hr cost heading toward ~$3/hr implies a margin structure that goes from good to extraordinary as volume scales. Management is now a commercialization machine (Johnson) sitting on a decade of irreplaceable legged-locomotion IP (Hurst). If humanoid logistics is a real market — and Morgan Stanley's $5T-by-2050 framing says the category is enormous — Agility is the incumbent operator with the data flywheel and the manufacturing head-start. The secular tailwind (warehouse labor shortages, $30/hr loaded labor, OpEx-friendly RaaS) is exactly the wind at its back. Earnings surprise to the upside = ISO certification lands early + a giant signs a 4-figure-unit order.

Bear case (2–3 things that permanently impair).

1. Squeezed from both ends. Above it, Figure raised at a $39B valuation (OpenAI-led $675M round) and Apptronik at $5B ($520M, Google-led) — rivals with 10–20x Agility's capital and the AI-narrative premium that wins talent and the best foundation-policy partnerships. Below it, Unitree ships a $16,000 G1 and filed a $610M Shanghai IPO with 335% YoY revenue growth — a Chinese cost structure that could make a $250K Digit look absurd for price-sensitive tasks. Agility is the mid-market squeeze play.
2. The $30/hr rate is the whole thesis, and it's the most vulnerable variable. Margin compression comes from the price side: as cheaper hardware floods in, the RaaS rate must fall, and the seductive 90% margin math evaporates. The bull models cost falling; the bear models price falling faster.
3. "First commercial" is a timing moat that decays. ~100 units is a pilot-scale franchise, not a fortress. If Figure's capital converts to deployments or Amazon (investor and customer and in-house robot developer) brings warehouse humanoids in-house, the head-start evaporates.

Pre-mortem (it's late-2027, the thesis broke — what happened?). ISO certification slipped past 2027, keeping deployments human-supervised and capping the operating ratio, so RaaS never reached autonomous-labor economics. Meanwhile a cheaper rival (Chinese hardware + a good-enough Western VLA) undercut the $30 rate at Agility's logistics customers, Amazon quietly favored its own/Covariant-derived system, and Agility — out-capitalized 20:1 by Figure — couldn't fund the next hardware generation and was acquired by a strategic at a flat-to-down round (the SoftBank ~$900M anchor reasserts itself).

Are the multiples too high? The $2.1B mark looks rich against ~100 units and undisclosed (likely <$50M) revenue — it's an option-value price on category leadership, sitting >2x above the only real-world acquisition bid ($900M). Defensible only if you believe Agility converts first-mover into franchise.

Contrarian view (what the market refuses to see). The crowd prices Figure and Tesla as the humanoid winners on narrative and capital. The contrarian read: the warehouse is the only proven humanoid market, and Agility is the only one actually winning it — boring, billing, and de-risked. If the market re-rates revenue over demos, Agility is the most under-valued name in the field. The symmetric risk: the warehouse may also be the market the cheapest hardware wins, in which case Agility's premium Western build is the wrong horse.

Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration is acute and conflicted. The flagship relationships are Amazon and GXO. Amazon is investor + customer + a company building its own warehouse robotics (Covariant acqui-hire) — the instant Amazon decides humanoids are core, Agility is disintermediated by its own backer. Lose Amazon or GXO and the "commercial leader" story has little left.
• The moat may be thinner than bulls think. "First to deploy" and "100k totes" are operational facts, not structural barriers. The teleop-data moat only matters if the task needs lots of data — but warehouse tote-moving is narrow and repetitive, the easiest thing to commoditize. The data advantage is widest exactly where it's least valuable.
• The most dangerous competitor bulls underestimate: Unitree. Not Figure (different game) — Unitree. A $16K robot, 335% revenue growth, an imminent Shanghai IPO, and a manufacturing cost structure Western builders cannot match. For price-sensitive material handling, a fleet of cheap, good-enough robots with a third-party brain may beat one premium Digit on total cost of ownership. Agility's premium build is a liability in a price war.
• Worst capital-allocation / governance signal: none egregious, but the C-suite churn (Wise out after ~2 yrs, full leadership reshuffle, 20% layoff) and the $2.1B mark vs $900M acquisition reality are the two things a short would hammer — a richly-marked private with organizational turbulence and a strategic acquirer's lowball already on record.
• What must hold for $2.1B: ISO cert on time, operating ratio to 10:1, RaaS rate defended near $30, RoboFab ramp to thousands of units, and no Amazon defection. If growth/deployments disappoint by 20–30%, the valuation has no revenue floor to catch it — it round-trips toward the strategic reservation price ($900M–$1.2B).
• The single scenario that permanently impairs: a credible, certified, cheaper competitor (Chinese hardware + commodity VLA) wins Agility's logistics customers on TCO before Agility reaches autonomous-labor economics — collapsing both the price and the data moat at once. Plausibility: moderate and rising, given Unitree's trajectory.

Management Questions (ordered by information value)

1. Of the ~100 Digits "sold," how many are running autonomously (operating ratio ≥5:1) versus human-supervised — and what is fleet-wide autonomous uptime? (The single number that separates real labor from subsidized demo.)
2. What RaaS rate are GXO/Amazon/Toyota actually paying today, and what is your assumed rate at 10,000-unit scale? (Tests whether the $30 economics survive competition.)
3. When does ISO functional-safety / human-collaboration certification land — committed date — and what is the deployment plan that's gated behind it?
4. What is current ARR and the net revenue retention on your earliest (GXO/Amazon) cohorts?
5. How do you defend the $30/hr rate against a $16K Unitree-class robot running a third-party policy at a fraction of the TCO?
6. Amazon is investor, customer, and a robotics developer. What contractual or strategic protection do you have against being insourced?
7. You're capitalized at a fraction of Figure ($39B) and Apptronik ($5B). What is your specific plan to fund the next two hardware generations without being out-spent into irrelevance?
8. RoboFab is built for 10,000 units/yr. What is the actual 2026 and 2027 production run-rate, and what is the binding constraint — demand or supply?
9. Your control stack relies on partners for the most general manipulation policies. What happens to your moat as foundation-policy players (Physical Intelligence, NVIDIA GR00T) commoditize the brain?
10. You declined SoftBank's ~$900M acquisition to raise at $1.75B. What changed in the business to justify ~2x, and is an acquisition still on the table?
11. What is the path to gross-margin-positive and cash-flow-positive unit economics at the fleet level, and when?
12. Why has the C-suite turned over so heavily in 24 months (incl. Melonee Wise's departure), and how do you de-risk further key-person loss?
13. What does Digit do in 2027 that it can't do in 2026 — which new tasks unlock which new customer verticals beyond tote-handling?
14. What is customer concentration today (top-3 % of revenue), and what is the plan to diversify below the GXO/Amazon dependency?
15. What single competitor or technology keeps you up at night, and why?