Shield AI
autonomous aircraft + AI mission-autonomy software (Hivemind) · dual hardware (V-BAT, X-BAT) + software-licensing model
Priced primary rounds only; Series C (2020/21, $90M) and Series D (Aug 2021, $210M, ">$1B" but no clean post-money disclosed) omitted per honesty gate. All USD.
The thesis on this name
State of Physical AI
Autonomy-software-first defense name — Hivemind (GPS/comms-denied autonomy pilot) + V-BAT surveillance drone, now extended via the Aechelon sim acquisition. $12.7B (Series G, Mar-26, >2x prior round); ~$540M 2026 rev (+80%). Smaller and earlier than Anduril but with a credible CCA/collaborative-combat-aircraft autonomy angle that could compound fast if Hivemind becomes the de-facto third-party autonomy layer.
Earnings, margins, COGS & capex
Revenue ~$267M (2024) → ~$300M (FYE Mar 2025) → $540M+ projected 2026, ex-Aechelon. Hivemind software ~30% of revenue (FYE Mar 2025); rest is V-BAT/ISR hardware + contractor-operated services. Backlog reportedly up 100%+ YoY on US Navy/Air Force task orders. Big ceiling vehicles: $46B Eglin EWACC prime position (10-yr IDIQ ceiling), $950M USAF JADC2 IDIQ ceiling, $198M USCG 5-yr ISR. NOT yet profitable; 2025 profitability target missed. Operating margin / EBITDA / FCF / capex all undisclosed — no audited financials public until/unless an S-1 is filed.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~80¢ is cost of goods and ~20¢ operating expense, leaving ~0¢ of operating profit.
Revenue trend
Margins
COGS structure
Not disclosed (private company).
Capex
Not disclosed.
Growth drivers
- Hivemind already flies on a rival's aircraft (Anduril YFQ-44A) — proving the platform-agnostic thesis and creating switching-cost lock-in as the autonomy reference layer
- USAF CCA production contract (Jun 2026) converts demos into a program of record at the exact moment the DoD is committing to collaborative combat aircraft at scale
- Revenue compounding ~80%/yr off ~$300M toward a stated $1B 2028 target, with backlog up 100%+ YoY
- Crossover capital (Advent, JPMorgan, Blackstone) + an IPO-proven CEO/CAO = a company being explicitly built for a ~2027 public listing
Bull & bear
Shield AI is the purest pre-IPO play on the autonomy LAYER of modern airpower — software that outlives any single airframe — with a real USAF production win, a credible $1B-by-2028 ramp, and an operator who has done the 0-to-IPO playbook before.
- Hivemind already flies on a rival's aircraft (Anduril YFQ-44A) — proving the platform-agnostic thesis and creating switching-cost lock-in as the autonomy reference layer
- USAF CCA production contract (Jun 2026) converts demos into a program of record at the exact moment the DoD is committing to collaborative combat aircraft at scale
- Revenue compounding ~80%/yr off ~$300M toward a stated $1B 2028 target, with backlog up 100%+ YoY
- Crossover capital (Advent, JPMorgan, Blackstone) + an IPO-proven CEO/CAO = a company being explicitly built for a ~2027 public listing
- Combat validation in Ukraine and 7+ allied government customers give it a moat of real-world, contested-environment flight hours competitors can't replicate quickly
A $12.7B mark on a still-unprofitable, services-heavy revenue base that has already missed its own targets twice — priced for an Anduril-like trajectory while fighting Anduril head-on and betting big on unflown hardware (X-BAT).
- Missed 2024 AND 2025 profitability/revenue targets — execution credibility gap vs the $1B-by-2028 narrative
- Anduril is bigger, better-funded, and competes on BOTH software (Lattice) and hardware — Shield AI risks being out-scaled
- ~$12.7B on ~$300M trailing revenue is ~25-40x sales; the multiple assumes flawless ramp and a clean IPO window
- X-BAT (~$27M/unit) is a capital-heavy hardware program that won't be mission-capable until ~2028 — a multi-year cash drain before payoff
- Revenue quality: large IDIQ 'ceilings' ($46B Eglin, $950M JADC2) are not booked revenue; actual recognized revenue is far smaller and procurement-cycle-dependent
- Lethal-autonomy regulatory/ethical risk is a tail that could cap the addressable market
What it is worth
Last priced round (Series G, Mar 26 2026) anchored at $12.7B, cross-checked vs revenue multiple and the Anduril comp. ~$300M trailing revenue → ~40x sales; ~$540M 2026E → ~24x forward sales. Reverse read: the mark requires ~70-100%/yr growth to ~$1B by 2028 AND a credible path to defense-software margins — i.e. it is priced for an Anduril-like trajectory and a clean 2027 IPO.
$6-8B re-rate
if profitability keeps slipping, X-BAT delays drain cash, Anduril out-scales it, or the defense-tech IPO window closes — back toward the pre-Series-G level
~$12.7B last-round mark holds into a 2027 listing as revenue roughly doubles to ~$540M+ and the CCA program de-risks
$20-25B+ at/around IPO
if X-BAT flies on schedule, CCA scales to multi-year production, and revenue compounds to ~$1B with improving margins (secondary/IPO marks already reaching toward ~$24B in optimistic scenarios)
SWOT
Strengths
- Hivemind is the first/only AI pilot fielded in real combat (since ~2018) and proven in GPS/comms-denied ops (Ukraine, Middle East)
- Platform-agnostic autonomy — flies on legacy F-16s, V-BAT, Kratos/GA targets, and notably rival Anduril's YFQ-44A under the USAF CCA program
- USAF CCA production contract (Jun 2026) validates a software-as-standalone-capability procurement model under the A-GRA reference architecture
- Heron Systems acquisition gave it the AlphaDogfight RL pedigree (beat an F-16 pilot in DARPA trials)
- IPO-proven leadership — CEO Gary Steele took Proofpoint 0→IPO and led Splunk's $28B sale to Cisco; CAO hired for IPO expertise
Weaknesses
- Unprofitable and capital-intensive; missed 2024 and 2025 financial targets
- Heavy dependence on slow-moving US/allied government procurement and appropriations cycles
- X-BAT is pre-flight (first VTOL flight by end-2026, mission-capable ~2028) — a large unproven hardware bet at ~$27M/unit
- Revenue still concentrated in services/ISR and a handful of large ceiling vehicles rather than recurring software ARR
- No public audited financials — investors underwrite on press-release figures and secondary marks
Opportunities
- USAF CCA + Navy autonomous-wingman programs scaling into multi-year production
- Hivemind Enterprise (developer platform) licensing to allied air forces (Singapore, South Korea, Japan, Korea Aerospace Industries)
- Export demand surge from Ukraine combat validation and allied rearmament
- Aechelon acquisition adds tactical simulation/synthetic-environment for autonomy training
- Software-decoupled autonomy could become the de-facto standard layer across multiple primes' airframes
Threats
- Anduril (Lattice OS, ~$30B+ valuation) is the dominant, better-capitalized rival across both software and hardware
- Defense primes (Lockheed, Northrop, RTX, GA) building or buying competing autonomy in-house
- Ethical/regulatory pressure on lethal autonomous weapons ('Stop Killer Robots') could narrow the market
- Budget/appropriations risk and program-of-record cancellations
- Talent competition for AI/ML engineers vs big tech and other defense startups
Moats, dependencies & bottlenecks
Moats
fielded AI pilot (Hivemind) with real contested-environment flight hours since ~2018
Platform-agnostic / airframe-decoupled autonomy adopted into the USAF A-GRA reference architecture — flies even on competitors' aircraft
Deep DoD + allied programs-of-record relationships and security clearances that take years to replicate
Vertical integration (Hivemind + V-BAT manufacturing + ViDAR/Sentient sensors + Aechelon simulation)
Dependencies
appropriations, and procurement timelines
NVIDIA / edge-compute GPUs and avionics-grade silicon for onboard inference
General Atomics, Boeing, KAI) whose platforms host Hivemind
Advantages
- First-mover, combat-fielded autonomy vs lab/demo-stage rivals
- Software-first model with higher-margin licensing optionality vs pure airframe makers
- IPO-operator leadership (Steele) and a board stacked with public-market crossover investors
- Allied-nation footprint (Ukraine, Japan, Singapore, S. Korea, Israel, Romania, Brazil) diversifying beyond US DoD
Weaknesses
- Unprofitable; repeated target misses
- Smaller and less-capitalized than Anduril
- Revenue still services/ISR-weighted, not recurring software ARR
- Heavy forward capital needs for X-BAT
Bottlenecks
- Government procurement and certification cycle length (months-to-years per program)
- X-BAT flight-test and production maturation (first flight end-2026, mission-capable ~2028)
- Path to profitability — scaling fixed cost base ahead of recognized revenue
- Scarce cleared AI/ML + flight-software engineering talent
- Manufacturing throughput for V-BAT/X-BAT airframes
Top signals & trends
Top signals
Converts autonomy from a demo into a standalone-procured program of record; flies on Anduril's YFQ-44A — validates platform-agnostic thesis
Crossover, public-market-oriented capital structure typically precedes an IPO within a few years
Execution credibility gap against the $1B-by-2028 plan; not profitable in 2025 as once guided
Adds synthetic-environment/training stack for autonomy; tuck-in to accelerate Hivemind dev
Major hardware milestone; success de-risks the airframe bet, slippage extends the cash drain
Not company-confirmed; supported by IPO-operator CEO Steele + IPO-experienced CAO Marlise Ricci
Trends
Structural multi-year demand pool Shield AI is positioned at the autonomy layer of
Favors a Hivemind-style independent autonomy vendor over airframe lock-in
Export tailwind across Europe and Indo-Pacific
Peer listings could pull-forward Shield AI's window and set comps
Policy tail risk that could constrain addressable market or export approvals
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Edge GPUs/compute for onboard autonomy inference
EW/comms payloads (DiSCO); also a Series F investor
Series F investor / allied aerospace supplier — non-US, named for analysis only
Navy, Army, SOCOM, USMC) Primary customer — CCA production, JADC2, EWACC, ISR services
$198M 5-yr contractor-operated ISR services
Japan, Singapore, South Korea, Israel, Romania, Brazil) V-BAT and Hivemind Enterprise exports; combat-proven in Ukraine
Primary rival — Lattice OS + autonomous hardware; bigger and better-funded. Coexists on CCA: Hivemind flies on Anduril's YFQ-44A. Both IPO candidates.
Low-cost uncrewed jets/targets (XQ-58 Valkyrie, Firejet) — overlaps Shield AI's drone category; also an autonomy host partner
Top small-drone/loitering-munition supplier to US military; less autonomous than Hivemind but a direct UAS competitor
Defense AI/software — partner on some programs but increasingly competitive in autonomy/decision software
MQ-20/MQ-9 maker — both an airframe partner (Hivemind on MQ-20 Avenger) and a CCA competitor
Primes building or acquiring in-house autonomy; both potential competitors and acquirers/customers