AI 低空经济与商业航天全景图 · 天上的生意 · 2026-07 · 双轨Low-Altitude & Commercial Space × AI · The Business of the Sky · Jul 2026 · Two tracks

发射是入场券,通信是收银台 Launch is the ticket; connectivity is the till

天上的生意走两条轨。航天轨:$/kg 曲线已经跑完最陡的一段——复用把成本打到约 600–1500 美元/kg(估),报价却反涨到 7400 万美元/次,垄断定价取代了技术进步,下一轮下降取决于第二个可回收玩家,不取决于算法。低空轨:卡点已经从「权限」转移到「需求」——门开了,没人进:全国起降点没有一份利用率统计,低空保险渗透率不足 0.1%。AI 的真实位置一句话说尽:它彻底接管了「看」(遥感解译)和「算」(星座避碰),完全没有接管「飞」(安全关键飞控获批数为零)和「管」(没有任何 AI 被授权做真实空管决策) The sky runs on two tracks. Space: the $/kg curve has run its steepest stretch — reuse cut costs to ~$600–1,500/kg (est.) while the list price rose to $74M a flight; monopoly pricing replaced technical progress, and the next decline waits on a second reusable player, not on algorithms. Low altitude: the choke point moved from permission to demand — the gates opened and no one walked in: no country publishes vertiport utilisation, and China’s low-altitude insurance penetration sits under 0.1%. AI’s true position in one line: it fully runs the «seeing» (EO interpretation) and the «computing» (constellation avoidance) — and none of the «flying» (zero certified safety-critical AI) or the «controlling» (no AI authorised to make a real ATC decision).

本图相信:真实的商业闭环存在——Starlink 约 114 亿美元营收、EBITDA 率约 63%(估);ICEYE 已盈利(backlog 15 亿欧元);极飞农业无人机净利率约 17%;美团无人机单均成本 2.3 元、五年年降 40–50%;Starlink 每星每年 40+ 次自主避碰机动——没有 AI 物理上无法运营(B)。This map believes: real loops exist — Starlink at ~$11.4B revenue and ~63% EBITDA (est.); ICEYE profitable on a €1.5B backlog; XAG’s agri-drones at ~17% net margin; Meituan’s per-drop cost at ¥2.3, falling 40–50% yearly; and Starlink’s 40+ avoidance manoeuvres per satellite per year — physically impossible without AI (B).
本图不相信:「万亿低空经济 3–5 年兑现」——约 40% 是关联带动产值,成熟民用无人机市场仅约 1500 亿元;欧洲投入 30 亿欧元造 eVTOL 原型,付费乘客为零;而全世界最著名的「AI 落地案例」(火箭自主着陆)用的是 G-FOLD 凸优化——里面没有 AI(A/B)。It does not believe the «trillion-yuan low-altitude economy on a 3–5 year clock» — ~40% is induced output, the mature drone market is ~¥150B; Europe spent €3B on eVTOL prototypes and carried zero paying passengers; and the world’s most famous «AI success» — rocket landing — runs G-FOLD convex optimisation: there is no AI inside (A/B).

主脊:双轨七站(火箭与发射→星座与卫星互联网→遥感与在轨 AI→eVTOL 审定→物流与巡检→空域与 AI 空管→保险责任与祛魅),③⑤标红;三条结构带:收入结构诚实层、承诺滑坡表、中美对照;判断层:复用次数×年频次 > 单次运力;场景付费意愿 > 技术可行性;保险与责任规则 > 适航证本身。姊妹:地面运输→transport,汽车→car,硬件→hardware,责任→liability,生存→survival The spine: seven stations on two tracks (rockets & launch → constellations → EO & on-orbit AI → eVTOL certification → logistics & inspection → airspace & AI ATC → insurance & debunking), ③⑤ in red; three bands: the honest revenue mix, the slippage table, the China–US ledger; judgment: reuse × cadence > single-flight payload; willingness to pay > feasibility; insurance & liability rules > the certificate itself. Sisters: transport, car, hardware, liability, survival.

635 : 0
SpaceX vs 中国的入轨级火箭成功着陆次数(截至 2026-07);单枚最高复飞 36 次 vs 0 次——差距约 10 年,且复用曲线更难追(A/B)SpaceX vs China in orbital-class booster landings (36 reuses on one booster vs zero) — a ~10-year gap, and the reuse curve is the harder one (A/B)
$114 亿
Starlink 2025 年营收(估):超过全球遥感行业(约 50 亿美元)的两倍——天上一块钱里六七毛来自「给地面卖网络」(估/弱)Starlink’s 2025 revenue (est.) — over twice the whole EO industry (~$5B); six or seven dimes of every sky dollar sell networks to the ground (est.)
30 亿欧
欧洲已投入建造 eVTOL 原型机的资金——搭载过的付费乘客数:零;Lilium 两次破产,Volocopter 破产被收购(A/B)What Europe spent building eVTOL prototypes — paying passengers carried: zero; Lilium bankrupt twice, Volocopter sold from insolvency (A/B)
0.1%
中国低空保险渗透率(全市场保费约 10 亿元)——保费是最诚实的产业探针:保险公司不会为 PPT 上的产值收保费(B)China’s low-altitude insurance penetration (~¥1B premiums) — the honest probe: insurers don’t underwrite slideware output (B)
口径裁判:SpaceX「IPO 1.8 万亿美元估值」两家媒体互证但无 SEC 一手文件;Joby 用五阶段、Archer 用四阶段——「谁进度快」的横向对比大多无效;UTM 市场规模两份报告差 30 倍=该赛道尚无可测量市场;电力巡检「全球 234 亿元 < 中国 890 亿元」同机构口径自相矛盾;低空经济 2025 年规模「1.5 万亿 vs 8591.7 亿」两个权威口径差近一倍——本图并列多口径,不采信单一数字。Refereeing the numbers: SpaceX’s «$1.8T IPO» is cross-reported but has no SEC filing; Joby counts five stages, Archer four — most «who’s ahead» comparisons are invalid; UTM market sizes differ 30× between reports (i.e., no measurable market yet); one agency prints «global ¥23.4B < China ¥89B» for inspection drones; and 2025’s low-altitude economy is either ¥1.5T or ¥859B depending on the authority — this map lists bases side by side and endorses none.
中心装置 · 一次升空的成本 vs 一次升空的权限The central device · The cost of a lift-off vs the permission for one
两条轨道的卡点在 2026 年都发生了位移:航天轨的瓶颈从技术移到竞争,低空轨的瓶颈从权限移到需求。Both tracks’ choke points moved in 2026: space from technology to competition, low altitude from permission to demand.
航天轨:成本革命完成,垄断定价接管Space: the cost revolution done, monopoly pricing in charge
复用把 SpaceX 的成本打到约 600–1500 美元/kg(估),报价却在 2026 年上调到 7400 万美元/次——报价与成本的裂口就是护城河,也是曲线停止下降的原因。2025 年全球 325 次发射,SpaceX 一家 165 次(占入轨质量 80%+)。转折信号:Blue Origin 2026-04 首次复用同一枚 New Glenn 助推器——复用不再是一家的专利,这是 $/kg 重新下降的必要前提。中国:2025 年 92 次发射创纪录、商业占 54%,但两次入轨级回收尝试全部失败——正站在 SpaceX 2015 年的位置上。Reuse cut SpaceX’s cost to ~$600–1,500/kg (est.) while the 2026 list price rose to $74M — the price-cost gap is the moat, and the reason the curve stalled. Of 325 global launches in 2025, SpaceX flew 165 (80%+ of upmass). The turn: Blue Origin re-flew a New Glenn booster in April 2026 — reuse is no longer one company’s patent, the precondition for $/kg to fall again. China: a record 92 launches (54% commercial), but both orbital-class recovery attempts failed — standing where SpaceX stood in 2015.
低空轨:门开了,客人没来Low altitude: gates open, no guests
权限侧确实在开:G/W 类空域立框、四川适飞面积超 80%、「扫码飞」把报备压到 2 分钟、美国行政令逼出 BVLOS 规则。但需求侧:深圳目标 1200+ 起降点而全国找不到一份利用率统计;低空保险保费仅约 10 亿元;北京仍全域禁飞——省际差异是「80% vs 0」的量级。2026 年的新死法:有飞行器、有航线、有起降点,依然没客户。Permissions are opening: G/W airspace classes, Sichuan above 80% flyable, two-minute QR clearances, a US executive order forcing BVLOS rules. Demand is not: Shenzhen targets 1,200+ vertiports while no utilisation figure exists anywhere in the country; premiums total ~¥1B; Beijing bans all flight — an 80%-vs-zero provincial spread. The 2026 way to die: aircraft, routes, pads — and still no customers.
判词Verdict航天轨在等第二个可回收玩家,低空轨在等第一批付费客户。AI 已经彻底接管「看」与「算」——遥感解译与星座避碰没有 AI 物理上无法运营;但「飞」与「管」获批数为零,而这两样才是规模化的瓶颈,也恰是 AI 最难攻克的部分Space waits for its second reusable player; low altitude waits for its first paying customers. AI fully owns the seeing and the computing — EO interpretation and collision avoidance cannot run without it; the flying and the controlling stand at zero approvals, and those two are both the bottleneck of scale and AI’s hardest ground.
Reading the Map

从这张图带走的五条规律Five patterns to take away

立场声明:本图为低空经济与商业航天的结构分析,不构成任何投资建议;军工与国防侧归 survival 图,本图只处理商业化部分;所有数字保留原始口径并标注可信度(官/财报/机/媒/弱/估),市场规模数字普遍不可靠,读者不应将任何一个当作精确值;市场地图 ≠ 推荐背书;仅供研究与批判性参考。Stance: a structural read of the low-altitude and commercial-space trades — not investment advice; defence sits in the survival map; figures keep original bases with confidence tags, and market-size numbers are broadly unreliable — treat none as precise; a market map is not an endorsement; for research and critical reference only.