Aerospace Autonomous

Differential insight

China's "sleeping data" problem is the actual market: 50M+ annual logistics drone flights + 47K eVTOL test flights exist at scale NOW, but zero entities are aggregating them into a cross-platform AI pilot cognition layer, the exact gap a founder with CAAC relationships + pilot credentials + capital market access can uniquely close before Western competitors can even enter the jurisdiction.

---

Consensus → Δ

Consensus: eVTOL OEMs (EHang, AutoFlight, Volant) are building the autonomous aviation future.

Δ: Every OEM is building model-specific flight control, structurally identical to pre-Mobileye automotive. The cognition/perception layer is being reinvented nine times simultaneously. The "Mobileye of low-altitude" white space is confirmed and unoccupied. Volant's $2B post-money with zero revenue/TC is the bubble signal that the OEM layer is overvalued relative to the software stack. [CONF: HIGH, confirmed across all nine surveyed OEMs with zero exceptions]

Consensus: Full autonomy requires waiting for regulatory frameworks to mature (FAA 2028+, CAAC unclear).

Δ: CAAC already certified pilotless manned flight (EH216-S, 31 months), issued the world's first OC for autonomous passenger ops (April 2025), and published AC-21-AA-2025-XX eVTOL framework (Dec 2025), China is 3–5 years ahead operationally. The "operate-to-certify" pathway (Xwing model applied to China: embed within SF Express/Meituan to generate CAAC-acceptable operational data) is faster than clean-sheet TC pursuit. [CONF: HIGH, CAAC track record documented; operate-to-certify CAAC applicability UNKNOWN]

Consensus: The pilot shortage drives autonomous aviation demand.

Δ: The real demand signal is the 1M-person operator/maintenance talent gap (NDRC) + Shenzhen's ¥12B vertiport buildout + six BVLOS pilot zones, not pilot shortage per se. The structural constraint is dispatch/fleet management cognition at scale (1,200 platforms by 2026 in Shenzhen alone), which is an AI ops problem, not a hardware problem. Whoever owns the fleet-intelligence layer owns the recurring revenue. [CONF: HIGH, infrastructure commitments documented; fleet AI revenue model unproven]

---

Why-now

Three simultaneous unlocks converging in 2025–2026: (1) CAAC issued first-ever OC for commercial autonomous passenger ops (April 2025) + published dedicated eVTOL certification AC (Dec 2025), the regulatory permission structure just went live; (2) Meituan crossed 10,000 daily orders on drone delivery (Dec 2024) creating a real-world dense-airspace dataset that didn't exist 18 months ago; (3) All nine major OEMs are in active TC pursuit simultaneously, creating a captive customer pool for a platform-agnostic cognition layer precisely when they need it most and before they've locked in proprietary stacks. The window closes when the first OEM achieves commercial scale and forecloses the platform opportunity, estimated 18–30 months before Volant/AutoFlight reach fleet size where switching costs become prohibitive.

---

Binding constraint

Regulation, specifically the absence of a modular software certification pathway in China (analogous to DO-178C) that would allow a cross-platform AI pilot OS to be certified independently of airframe TC. Currently CAAC co-certifies software with the aircraft type, making platform-agnostic autonomy certification structurally impossible under existing rules. CCAR-92 rulemaking is in progress but no public timeline. This is not a technology constraint (XPeng VLA 2.0 proves the AI architecture exists) nor a capital constraint (China has 17+ funded rounds in 2024 alone), it is purely a regulatory architecture problem. Your edge: founder with CAAC/AVIC senior relationships is the only profile that can a) shape the CCAR-92 rulemaking to include a modular software certification pathway, and b) negotiate data-sharing frameworks with Meituan/SF Express/EHang that no pure-tech startup can access.

---

Wedge

Don't build an OEM. Don't pursue passenger TC first. Wedge: become the autonomous dispatch + fleet cognition OS for China's cargo drone operators (SF Express 丰翼 + Meituan tier), embedded as an operational partner under existing CAAC logistics operating certificates. This generates: (a) CAAC-acceptable operational data at scale (Xwing model, China-localized); (b) recurring SaaS revenue from operators managing 100+ aircraft who currently use human dispatch; (c) a cross-platform training corpus owned by your entity, not the OEM. Once CCAR-92 modular certification pathway exists (your founders help write it), retrofit the cognition layer into manned eVTOLs as a certified avionics module. This is the two-phase moat: data asset first, certified platform second.

---

72-hour MVP spec

Target: SF Express 丰翼 or Meituan drone ops team, Shenzhen.

Build: A cross-route flight anomaly detection + dispatch decision dashboard ingesting real telemetry from 10 existing drone routes via API/SDK. Feature set: (1) real-time route deviation flagging with AI-suggested correction; (2) weather-risk scoring per corridor; (3) automated post-flight log structuring for CAAC data submission. Stack: Python + existing CAAC-format telemetry parser + GPT-4o for natural language ops summary. Goal: demonstrate you can sit between the operator's dispatch system and CAAC reporting requirements, making you structurally necessary, not optional. Ask: a 30-day data-sharing pilot agreement, not a contract. Metric: reduce dispatcher cognitive load on 10 routes measurably (time-per-decision). This MVP requires no new aircraft, no TC, and no capital beyond a laptop, it proves the data aggregation thesis before writing a line of autonomy code.

---

Fundability

VC power-law case (credible but narrow): The "Mobileye of low-altitude China" framing works for Tier-1 China funds (Sequoia CN, HillHouse, Shunwei) and strategic CVCs (CATL Ventures, Geely Capital, COMAC's investment arm) IF you can credibly demonstrate: (a) CAAC data-sharing framework is achievable (your edge), (b) cross-platform training corpus creates a winner-takes-most dynamic. The comp is Mobileye at $50B exit, plausible if you control the cognition layer across China's projected 1M+ low-altitude vehicles by 2035. Series A target: $15–20M for operate-to-certify data accumulation phase. Series B: $80–120M post-CCAR-92 modular cert pathway confirmation.

Honest caveat: This is also an excellent cash business without VC, fleet dispatch SaaS for cargo operators is profitable at 200+ aircraft under management with zero autonomy certification required. If CCAR-92 modular pathway takes 7+ years, the SaaS business sustains the team while the long-term autonomy moat compounds. Tell both stories; raise from strategics who benefit from the platform outcome.

Capital markets / liquidity: Merlin's SPAC at $800M on $8.5M revenue (94x) sets the public market comp. A China-domiciled entity faces Hong Kong/A-share listing path; STAR Market (科创板) has accommodated deep-tech pre-revenue companies. The more realistic near-term liquidity event is strategic acquisition by an OEM (AutoFlight, EHang, or COMAC) once the data corpus is proven, 3–4 year M&A exit more likely than IPO in the 5-year window.

---

Biggest UNKNOWN

Whether CAAC's CCAR-92 rulemaking will include a modular software certification pathway (analogous to DO-178C) that allows a cross-platform AI pilot OS to be type-certified independently of a specific airframe. Everything else is solvable, the technology exists (XPeng VLA 2.0), the data exists (Meituan/SF), the capital exists, the market exists. If CAAC creates this pathway, the "Mobileye of low-altitude" play becomes the highest-conviction venture bet in the sector. If CAAC maintains aircraft-coupled software certification for the next decade, the business is still viable as fleet-dispatch SaaS but the power-law outcome requires an OEM acquisition rather than independent platform scale. Your CAAC relationships are the only way to get a credible answer to this question before any competitor can.