Carbon Robotics

Phase A — Understand the business

Lens 1 · Company Overview

The business in one line: Carbon Robotics builds AI-powered farm robots — the flagship is the LaserWeeder, a tractor-pulled (and now autonomous) implement that uses computer vision to tell crop from weed and then kills the weed with high-powered lasers, replacing herbicides and hand-weeding crews in high-value specialty row crops.

How it actually works: An array of high-resolution cameras images the bed in real time; onboard AI classifies each plant as crop or weed; targeting lasers fire at the weed's growth center, disrupting cellular structure so it doesn't regrow. The G2 platform runs 36 high-resolution cameras and 24 onboard NVIDIA GPUs to "analyze thousands of plant images per second," eliminating up to ~600,000 weeds/hour (≈5,000 weeds/minute). The classifier is a proprietary "Large Plant Model" (LPM) the company says is trained on 150 million labeled plants across crops, soils, climates and growth stages, announced Feb 2026.

Founded: 2018, Seattle WA, by Paul Mikesell (still CEO/founder).

Products (the lineup as of mid-2026):

• LaserWeeder G2 — the core capital product. Modular line (G2 200 / 400 / 600) covering ~30–300 hectares of working width, launched Feb 2025; ~2× the kill throughput and ~25% lighter than the G1.
• Carbon AutoTractor (a.k.a. ATK, Autonomous Tractor Kit) — a retrofit autonomy kit (launched Mar 2025) that makes existing John Deere 6R/8R tractors driverless within a 24-hour install; does weeding and tilling, plowing, cultivating, discing, mowing, etc. Sold on a per-hour RaaS-style model, ~$15–25/hr with Carbon operators supervising remotely over a satellite link.
• Carbon Ops Center — the farm-operations software layer / fleet dashboard.
• A third, still-"secretive" AI robot funded by the Oct 2025 "D-2" round, reusing the existing AI stack for a new application.

Main customers: specialty / vegetable & organic row-crop growers — lettuce, spinach, onions, carrots, broccoli, cauliflower, chard, cilantro, kale, tomatoes, sugar beet ("100+ crops"). ">100 growers" / ">200 LaserWeeders deployed on US farms," across 15 countries (North America, Europe, Australia, New Zealand).

Suppliers: NVIDIA is the critical compute supplier (also an investor via NVentures) — onboard GPUs on the LaserWeeder, Jetson Orin on the AutoTractor. Lasers, optics and high-res cameras are the other key bill-of-materials inputs (specific vendors not disclosed). Manufacturing is in-house in Richland, WA, with a new Netherlands plant opened to hedge tariffs and speed EU deployment.

Competitors: John Deere (See & Spray, via Blue River), Ecorobotix (CH), FarmWise (acquired by CNH), Verdant Robotics, Naïo (FR), Laudando & Associates (litigant) — see Lens 3.

Contract structure / payment terms: two distinct models, which matters a lot for the thesis —

• LaserWeeder = outright capital equipment sale ($1.2M G1; G2-600 ~$1.4M plus annual support; a 3rd-party review cites a smaller config "$500K" with 5-yr TCO 1.5–2.5× purchase price). Plus tiered seasonal service plans. Financing arranged via banking partners, not vendor paper.
• AutoTractor = per-hour recurring ($15–25/hr) — the first toe into recurring revenue.
• Direct sales only — no dealerships. Mikesell is explicit he won't sell through ag dealer networks, wanting the direct customer relationship.

Lens 2 · Supply Chain

Map upstream → Carbon → end customer, naming every stakeholder I can source:

Upstream inputs:

• Compute: NVIDIA — onboard GPUs (LaserWeeder) + Jetson Orin (AutoTractor). This is the single most strategic input and a single-source dependency at the AI-silicon layer; NVIDIA is also a shareholder (NVentures), which de-risks allocation but couples Carbon to NVIDIA's roadmap and pricing.
• Lasers / optics: high-power laser modules + beam-steering optics — vendor(s) undisclosed ``. This is the cost and reliability chokepoint of the whole machine and a plausible second single-source risk.
• Cameras / imaging: 36 high-res cameras per G2 unit — vendor undisclosed.
• Chassis / steel / hydraulics / harnessing: standard ag-implement supply, assembled in-house.

The company (midstream): In-house assembly, Richland WA + Netherlands plant (2025). Vertical integration of manufacturing was a deliberate margin move ("manufacturing moved in-house, improving margins" — Mikesell). Carbon also owns the data layer (LPM training on 150M plants) — proprietary, not bought.

Downstream / route to market: Direct sales force (US, EU, Australia) → grower (the farm operator buys or rents) → financing via third-party banks. No distributor/OEM/dealer intermediary by design. For AutoTractor, Carbon inserts itself as an ongoing remote-operations service provider (its own staff supervise the fleet).

Chokepoints / single-source dependencies:

1. NVIDIA silicon — concentrated, but investor-aligned.
2. Laser/optics module — the defining, likely-concentrated input; the L&A litigation is literally a fight over "lowest $/watt laser weeding" (Lens 10), i.e. the laser cost curve is the competitive battleground.
3. Skilled field-service network — a thin service footprint in remote ag regions is an industry-wide barrier; Carbon's "redundant cellular + remote monitoring" design is partly a mitigant.

(This lens is web-only; supply-chain.md for agtech is missing in the commercial layer.)

Lens 3 · Competitive Advantages (moats)

Where Carbon genuinely leads: It is repeatedly described as the commercial leader in laser weeding in North America — ">200 LaserWeeders deployed," first & only commercial LaserWeeder, >250k acres serviced in the 2024 season alone, and (cumulatively) >40 billion weeds eliminated across >500k acres. First-mover commercial scale in this exact modality is real.

The four candidate moats, judged:

1. Data / AI flywheel (strongest). Every deployed machine images millions of plants; that feeds the Large Plant Model (150M labeled plants), which improves discrimination accuracy, which sells more machines, which generate more data. This is a genuine compounding edge that a late entrant cannot buy — the closest thing Carbon has to a durable moat. ``.
2. IP / patents (real, and being litigated). Carbon holds laser-weeding patents (e.g. 12,108,752 and 12,127,547) and won a preliminary injunction enforcing them against Laudando & Associates. A court-tested patent that produces an injunction is a stronger-than-average moat signal for a startup.
3. Switching costs / installed base. A $1M+ capital machine plus a service relationship plus operator familiarity creates real lock-in once installed — but the base (>200 units) is small, so this is nascent.
4. Brand / category ownership. "LaserWeeder" is effectively the category name in North America. Useful but not defensible alone.

Bargaining power:

• Over customers: moderate-to-strong inside the niche — the ROI story (cut weed-control cost up to 80%; yield up 10–15%, sometimes 50% for organic spinach/lettuce at Triangle Farms) is compelling where hand-weeding labor is scarce and expensive. But the customer is price-sensitive and cyclical (farm capex), and the $1M+ ticket means a small buyer pool.
• Over suppliers: weak on NVIDIA (price-taker), weak-to-moderate on laser/optics (the whole industry is racing the $/watt curve).

The honest moat verdict: the data flywheel + court-tested IP are above-average for a Series-D hardware startup. But none of it protects against the one competitor that matters most — John Deere — which doesn't need Carbon's modality to win the broader "precision weed control" war (see Lens 13).

(Ground note: positioning.md / bottlenecks.md for agtech are missing; this is web-only.)

Lens 4 · Segments

segments.csv is empty (header-only) — there is no segment data. Carbon does not disclose revenue by product or geography. What can be reconstructed /:

• By product: Revenue is today overwhelmingly LaserWeeder capital sales (the only product with multi-year commercial history). AutoTractor (per-hour, launched Mar 2025) and the 3rd robot (unlaunched) are early-stage; recurring revenue is a small but strategically critical and growing slice. ``.
• By geography: US is the core market; 15 countries total incl. Europe (the new NL plant signals EU as the #2 growth vector), Australia, New Zealand. No country-level revenue split disclosed.
• By crop: specialty/vegetable & organic row crops — lettuce, spinach, onions, carrots, brassicas — i.e. high-value, hand-weed-intensive, herbicide-poor crops where the ROI is sharpest. Notably not broadacre commodity crops (corn/soy/wheat) — that is precisely the frontier Carbon has not yet credibly entered, and the gap John Deere See & Spray already fills at 5M acres (Lens 13).

Trend: accelerating top line (revenue crossed $100M for FY ending Jan 31 2026 ) with the mix broadening from pure-weeding capex toward an autonomy/recurring layer — the right strategic direction, but unquantified.

---

Phase B — Measure performance (+private overlay applied)

Lens 5 · Funding & valuation trajectory (swapped in for "Earnings Result")

No earnings to print — so the equivalent "result" is the financing & traction record. All ``, unaudited:

.

Traction "print":

• Revenue: >$100M annual, FY ending Jan 31 2026 — and, per the company, achieved without a concurrent equity raise (a positive efficiency signal). Fiscal year-end = Jan 31.
• >500k cumulative acres serviced; >40 billion weeds eliminated; 100+ crops; 15 countries; >200 LaserWeeders; >100 growers.
• Growth rate, gross margin, burn, profitability: not disclosed — n/a — private, not disclosed.

⚠ Funding-total conflict (surfaced, not resolved): Primary announcements sum to ~$177M ($157M post-Series-D + $20M D-2), and PremierAlts/Notice corroborate "$177.2M total funding". But Tracxn reports "$276M over 8 rounds from 11 investors". The gap (~$100M) is most likely debt/venture-debt facilities or unannounced extensions counted by Tracxn vs. announced equity — I cannot reconcile it from public sources. Do not treat $276M as equity raised. Flag for diligence.

Lens 6 · Founder interviews / sentiment trend (swapped in for "Earnings Calls")

No earnings calls. Reading Mikesell across 2024→2026 interviews (AgFunder, GeekWire, NVIDIA podcast, Fuse, Built In):

• Consistent throughline: "We are an AI robotics company, not an agtech company" — he explicitly coaches founders to position away from "agriculture" to keep VCs engaged ("Be careful how you talk about it"). This is a deliberate narrative-engineering posture — useful for fundraising, but a tell that the pure-ag TAM may not carry the valuation alone.
• The labor thesis hardens over time: by 2025 he frames "farmers' biggest challenge is labor and labor availability" as the central driver — which is why the 2025 pivot toward autonomy (AutoTractor) and a 3rd robot, not just weeding.
• Discipline signals: insists on direct sales (won't be "starry-eyed" about dealerships), demands ROI proof (1–3 yr payback) before claiming value, and leans on customer testimonials as the #1 growth lever. This is an operator who undersells rather than oversells — a positive vs. typical agtech hype.
• Maturity-of-org language (2026): the CFO hire is framed as "evolving our leadership to match our rapidly increasing maturity in the market" — public-company-prep vocabulary.
• What he stopped emphasizing: early-days "save the planet / sustainability-first" framing has receded behind labor + ROI + AI — a shift from mission to margin that reads as maturation.

Net tone: increasingly confident, disciplined, deliberately AI-coded, pre-IPO-posturing.

Lens 7 · Cap table & secondary marks (swapped in for "Comps")

Syndicate quality (the IPO-proximity tell):

• Tier-1 / strategic: BOND (Mary Meeker's firm — a crossover-quality growth lead; its entry at Series D is a meaningful late-stage signal), NVentures (NVIDIA) — strategic + supplier, Anthos Capital, Revolution (Steve Case), Sozo, Fuse, Ignition, Voyager.
• What's missing for a strong IPO-proximity read: no disclosed mutual-fund crossover (Fidelity / T. Rowe / Wellington) markup yet. BOND is growth-stage quality but not the same signal as a T. Rowe round. That keeps readiness at the 2–3/5 band, consistent with private-watch.json (2/5).

Valuation & secondary marks (all ``, dispersion is wide — treat as soft):

. The honest read: secondary marks suggest a ~$370M → ~$550M markup over ~18 months — directionally up on the back of the $100M revenue milestone, but the share-price estimates conflict by ~5× ($2.64 vs $12.14), so any single number is n/a — not reliably sourced. There is no clean public multiple; at ~$550M EV on ~$100M revenue that's **~5.5× trailing revenue **, reasonable for a profitable-ish hardware-scaler but rich if margins are thin.

Mechanism/peer set (no P/E to compare — these are the strategic comparables):

• John Deere (See & Spray / Blue River, acq. $305M 2017) — public, the 800-lb incumbent, 5M acres on See & Spray in 2025.
• CNH Industrial (acquired FarmWise 2025) — public incumbent #2.
• Ecorobotix (CH, private) — European leader, spray-based.
• Verdant Robotics, Naïo (FR), Laudando & Associates — private challengers.
• Public ag-automation read-throughs: AGCO, Trimble, Deere for multiples context.

Lens 8 · Stock-price catalysts → funding/product events (swapped)

No ticker, so the "what moves the name" question = which events re-rate the private mark / drive the narrative:

• Series D ($70M, BOND-led), Oct 2024 — the institutional-validation re-rate to ~$370M.
• LaserWeeder G2 launch, Feb 2025 — product credibility + TAM-widening (smaller/cheaper configs → more farm sizes).
• AutoTractor launch, Mar 2025 — the strategic pivot to autonomy + recurring revenue; arguably the most important narrative event because it's the bridge out of the weeding niche.
• $20M "D-2" + "secretive new robot," Oct 2025 — signals platform ambition beyond weeding.
• $100M revenue + CFO hire (Krysler), Mar 2026 — the pre-IPO milestone double-header; the single biggest re-rate catalyst to date and the basis for the secondary markup.
• Large Plant Model (150M plants), Feb 2026 — reinforces the data-moat narrative.
• CNBC Disruptor 50 (2025 and 2026) — recurring credibility stamp.
• L&A patent injunction win, 2024–25 — IP-moat validation.

Pattern: this name re-rates on (1) product-line expansion beyond weeding and (2) hard revenue/operating milestones, not on incremental funding alone. The market (secondary buyers) is buying the platform-ization story, not the weeder.

---

Phase C — Judge people & books

Lens 9 · Management

Paul Mikesell — Founder & CEO. The headline asset.

1. Track record (quantified, strong): co-founded Isilon Systems (2001), scale-out storage, IPO'd 2006 (later acquired by EMC for ~$2.25B) — a genuine, realized, large outcome. Then co-founded Clustrix (distributed SQL), acquired by MariaDB (2018). Then Director of Infrastructure Engineering at Uber, working on deep learning / computer vision — directly relevant to Carbon's stack.
2. Tenure & skin in the game: founder, in seat since 2018 (8 years); founder ownership not disclosed but presumptively significant for a Series-D founder-CEO ``.
3. Capital-allocation history at Carbon: disciplined — brought manufacturing in-house to lift margins, hit $100M revenue without a fresh raise, raised a small $20M extension rather than a bloated mega-round, refuses dealer channels to protect economics. This reads as a capital-efficient operator, unusual in hype-prone agtech.
4. Red flags: none material surfaced. The one yellow flag is the deliberate "we're not really agtech" narrative-engineering — savvy, but a reminder the ag-only numbers may not justify the mark, and that the IPO story leans on an AI re-label.
5. Archetype: repeat technical founder (storage/database/CV pedigree) applying a software/data playbook to hardware — the right archetype for a data-flywheel thesis, and a credible IPO CEO.

Kevan Krysler — CFO (appointed Mar 2026). The pre-IPO signature hire. ~30 yrs finance; most recently CFO of Everpure; senior finance exec at VMware (a public company) and partner at KPMG (audit/public-reporting credibility). Hiring a public-company/Big-Four CFO at the same moment you announce a $100M revenue milestone is the clearest single tell that an S-1 process is being readied. Board: Mood Rowghani (BOND) joined at Series D.

Net: an above-average management read — a proven exit-generating founder plus a public-markets-credible CFO. This is the strongest single pillar of the bull case.

Lens 10 · Forensic Red Flags

Important caveat: there are no audited financials, no 10-K, no segment disclosure — so the biggest forensic red flag is structural: everything is company-disclosed and unverifiable. The $100M revenue, the 80%-cost-saving and 10–50% yield claims, and "profitable-ish without a raise" all rest on management's word. Specifically un-sourceable: gross margin, burn, cash runway, deferred-revenue treatment of service plans, and how AutoTractor per-hour revenue is recognized. All n/a — private, not disclosed.

What public sources do surface:

• Revenue-quality question (capex sale + service + RaaS mix): with three different revenue models (one-time machine sale, multi-year service plans, per-hour AutoTractor), revenue-recognition and the durability/quality of the $100M are genuine open questions a diligence process must crack. A capital-equipment sales number is lumpier and lower-quality than recurring ARR. ``.
• TCO vs. headline ROI gap: a 3rd-party review pegs 5-yr total cost of ownership at 1.5–2.5× the purchase price — i.e. the all-in cost is materially above sticker, which can stress the "1–3 yr payback" claim for marginal buyers and create churn/repurchase risk.
• Customer concentration: unknown but plausibly meaningful given only ~100 growers / ~200 units — n/a — not disclosed, but a real risk for a thin installed base.

Regulatory findings (required sub-section). Read companies/carbon-robotics/regulatory/regulatory-findings.md (Stage-1 pre-fetch, dated 2026-06-18):

• SEC (EDGAR EFTS — LR + AAER): zero findings. Carbon has no CIK and is not an SEC filer, so no Litigation Releases or AAERs are possible.
• Item 3 Legal Proceedings: n/a — no 10-K exists (private).
• Litigation (web, the material item): Carbon is the plaintiff, not defendant. Carbon Autonomous Robotic Systems Inc. v. Laudando & Associates LLC (No. 2:24-cv-03012, E.D. Cal., filed Oct 2024), alleging infringement of patents 12,108,752 and 12,127,547 plus trade-secret misappropriation. Carbon won a preliminary injunction barring L&A from making/using/selling its machines in the US pending resolution; L&A responded by open-sourcing its "L&Aser" module on GitHub, and the case was stayed for L&A to obtain substitute counsel. This is moat-positive (offensive IP enforcement that worked), but carries counter-risks: an open-sourced low-$/watt design could commoditize the laser cost curve, and aggressive IP litigation invites counterclaims / invalidity challenges.
• Other agencies (FTC/DOJ/FDA/CFPB/laser-safety/recalls): web search returned no material enforcement, recall, fire, injury, or laser-safety action against Carbon Robotics, and no layoffs surfaced.

Conclusion: No material adverse regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER, all zero — no CIK), web search, and the absence of any 10-K, as of 2026-06-18. The only litigation is offensive and favorable to Carbon to date. The dominant true risk is information opacity, not known misconduct.

---

Phase D — Project & stress-test

Lens 11 · IPO-readiness & path-to-tradeable (swapped in for "Forward Projection")

Grounding (research/private-watch.json): stage growth, ipo_readiness 2/5, leads Anthos + BOND, catalyst "LaserWeeder deployment across acreage." Scale: 1=seed … 5=S-1/imminent.

My read — call it a 2.5–3/5 now, trending up. The case for moving it up from the recorded 2:

• ✅ $100M revenue (a real public-market-relevant scale threshold).
• ✅ Public-company CFO (Krysler, ex-VMware/KPMG) installed Mar 2026 — the canonical S-1-prep hire.
• ✅ Crossover-quality lead (BOND) + strategic (NVIDIA) on the cap table.
• ✅ Profitable-ish efficiency signal (hit $100M without a fresh raise).

What still caps it below 4/5 (why it isn't pre-IPO yet):

• ❌ No disclosed mutual-fund crossover round (no Fidelity/T. Rowe markup) — the strongest IPO-proximity tell is absent.
• ❌ TAM-credibility gap: at ~$100M revenue and a specialty-crop niche, the public-market story needs the AutoTractor RaaS + 3rd robot + a row-crop path to justify a growth multiple. The platform story is announced but not proven in the numbers.
• ❌ Capital-equipment revenue quality (lumpy hardware sales vs. recurring ARR) is a harder IPO narrative than SaaS.
• ❌ Macro: ag-capex is cyclical; a soft farm-income year could stall the order book right when an IPO window matters.

Milestones that unlock an S-1: (1) recurring/ARR mix from AutoTractor reaching a credible share of revenue; (2) a row-crop or broadacre proof-point that expands TAM beyond vegetables; (3) the "secretive 3rd robot" launching into a second vertical; (4) a crossover (Fidelity/T. Rowe-type) primary or large structured secondary at a marked-up valuation; (5) two clean audited years from the new CFO. Estimated tradeable window: H2 2027–2028 `` — *more conservative than the speculative "H2 2026/H1 2027" floated by a secondary commentator *.

Write-back: This dossier is the first for the name (private-watch.json shows dossier: null). Per the +private overlay, the carbon-robotics entry should have dossier set to this file's path — noting it here for the conversational write-back step; per strict wave boundaries this run does NOT edit private-watch.json or watchlist.json.

(No forecast.ts create — --watchlist unattended mode skips logging a Brier forecast, and there is no EPS line to commit for a private.)

Lens 12 · Bull vs Bear

Bull case (narrative). Carbon is the clear category leader in a real, growing problem — herbicide resistance + a collapsing supply of hand-weeding labor in high-value specialty crops, with zero good chemical options in many vegetables. It has the only at-scale commercial laser-weeding fleet (>200 units, >500k acres), a compounding data moat (150M-plant LPM that rivals can't buy), court-tested patents, a proven repeat-founder with a real IPO under his belt, and a fresh public-company CFO. Crucially it has begun the escape from the weeding niche: AutoTractor turns it into a recurring-revenue autonomy platform retrofitting the world's installed John Deere base at $15–25/hr, and a 3rd robot extends the AI stack into new verticals. It hit $100M revenue without burning a fresh round — capital-efficient growth that, if the recurring layer compounds, supports a clean IPO at a premium to today's ~$550M mark.

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

1. The incumbent eats the broad market without ever beating Carbon's modality. John Deere's See & Spray is already at 5M acres — Deere owns the dealer network, the financing, the broadacre customer, and an OTA-updating fleet. Carbon's laser approach may be better in vegetables and irrelevant in the 100×-larger corn/soy/wheat market that Deere already serves with spray. Carbon could "win" laser weeding and still be a niche vegetable-tool company — which does not justify a growth-IPO multiple.
2. The niche is small, cyclical, and price-capped. A $1M+ machine for ~100 growers, TCO 1.5–2.5× sticker, payback only where labor is dear, demand tied to volatile farm income. The serviceable buyer pool for a seven-figure weeder is inherently thin; growth could plateau faster than the platform story matures.
3. Cost-curve commoditization of the laser. The L&A fight is over "lowest $/watt" — and L&A open-sourced its design. If laser-weeding hardware commoditizes, Carbon's premium hardware margin compresses and the moat narrows to data + service.

Pre-mortem (18 months out, thesis broke — what happened?). Most likely: ag-capex softened, LaserWeeder unit sales stalled, AutoTractor recurring revenue ramped slower than hoped, the "3rd robot" slipped — so the platform re-rate never arrived, the IPO window closed, and the secondary mark drifted back toward the ~$370M Series-D level. A nastier version: a high-profile field-safety/reliability incident (lasers + autonomy + remote ops is a real liability surface) dents trust in the installed base.

Are the marks too high? ~$550M on ~$100M lumpy hardware revenue (~5.5×) is defensible but not cheap — it already capitalizes the platform optionality that hasn't shown up in the numbers. You're paying for the AutoTractor/3rd-robot future today.

Contrarian view (what the market refuses to see). Bulls treat Carbon as "the laser-weeding leader." The real question — and the only one that justifies the valuation — is whether the AutoTractor autonomy-RaaS layer becomes the main event, making Carbon a recurring-revenue farm-autonomy platform that happens to have started in weeding. If yes, today's mark is cheap. If the company stays a premium vegetable-weeder vendor, it's a fine business but a poor growth-IPO at this price. The weeder is the wedge, not the thesis.

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

Dismantling the bull case:

• Revenue concentration & quality: ~$100M from lumpy, ~$1M+ capital-equipment sales to ~100 growers — not durable ARR. What's the customer concentration? What's the repeat/replacement rate? A few large-grower cancellations in a weak farm-income year and the "growth" story inverts. n/a — not disclosed, which is itself the indictment.
• The moat is weaker than bulls think. The data flywheel only matters if classification accuracy is the binding constraint — but the binding constraint may be $/watt laser cost and machine price, where an open-sourced L&A design and a global $/watt race can erode the premium. Patents bought an injunction against one tiny rival; they don't stop Deere or Ecorobotix doing spray-based precision that sidesteps lasers entirely.
• The most dangerous competitor bulls underestimate: John Deere. Not because Deere will out-laser Carbon, but because Deere redefines the battlefield — "precision weed control" via See & Spray on 5M broadacre acres, sold through dealers Carbon refuses to use, financed by Deere's own balance sheet, updated OTA. Carbon's direct-sales-only insistence is a moat against scale: it caps reach to whatever its own salesforce can cover. CNH (FarmWise) is the same threat from the #2 incumbent.
• Capital-allocation / governance: few public red flags — but the "we're an AI company, not agtech" narrative-engineering is a yellow flag that the ag economics alone don't support the mark, and there's no audited verification of any claim.
• What must hold for ~$550M: that AutoTractor recurring revenue scales fast, the 3rd robot lands, and the niche keeps growing through a farm-capex cycle — three things at once, none yet proven.
• If growth disappoints 20–30%: a lumpy-hardware name that misses by 20–30% in a cyclical end-market typically re-rates hard; the secondary mark likely reverts toward the Series-D ~$370M (a ~30%+ haircut from ~$550M) ``.
• Single scenario that permanently impairs: Deere/CNH bundle "good-enough" precision weeding into the broadacre platform farmers already own and finance, capping Carbon at the vegetable niche permanently — a fine ~$100–200M-revenue specialty business, but a broken growth-IPO thesis. Plausibility: moderate-to-high — it's arguably the base case unless AutoTractor changes the game.

Lens 14 · Management Questions (ordered by information value)

1. Of the ~$100M FY26 revenue, what % is recurring (service plans + AutoTractor per-hour) vs. one-time machine sales, and where do you expect that mix in 3 years? (The single number that decides whether this is a platform or a hardware vendor — and the valuation.)
2. What is the row-crop / broadacre path? Can laser weeding ever be economic in corn/soy/wheat, or is the TAM structurally specialty-crop only? (Defines the ceiling.)
3. Gross margin on the LaserWeeder today, and the margin trajectory now that manufacturing is in-house?
4. Customer concentration — what share of revenue is the top 5 / top 10 growers, and what's the unit replacement/repeat rate as the early G1 fleet ages?
5. AutoTractor unit economics: contribution margin per machine-hour at $15–25/hr after remote-operations labor, and how many hours/units are live today?
6. Are you cash-flow positive? If not, what's the burn and the runway after the $20M D-2 — and do you need another primary before an IPO?
7. What exactly is the third "secretive" robot, what vertical, and when does it ship revenue?
8. How real is the John Deere See & Spray threat in your accounts — are you losing deals to "good-enough" precision spray, and what's your win rate against it?
9. Direct-sales-only caps reach to your own salesforce — at what scale does that break, and would you ever reconsider channel/OEM partnerships?
10. The L&A patents are being challenged and the design open-sourced — how durable is the IP moat, and what's your exposure to a $/watt cost-curve collapse?
11. Reconcile the funding figures: announced equity is ~$177M but third parties report ~$276M — how much venture debt is on the balance sheet?
12. What's the realistic IPO window, and which 2–3 milestones gate the S-1 (ARR mix? crossover round? audited years?)?
13. What is the field-reliability / uptime record across the fleet, and your safety-incident history (lasers + autonomy + remote ops)?
14. How exposed is the bill of materials to NVIDIA pricing/allocation and to tariff regimes — and is the Netherlands plant a sufficient hedge?
15. Where does the data moat actually win? Quantify the classification-accuracy advantage of the 150M-plant LPM vs. the best competing model.

---