Two weeks apart, two separate US policy actions closed off the low-cost path into American drone and robotics hardware.
July 28, the FCC added foreign-made "advanced robotic devices" — humanoids, quadrupeds — to its Covered List, blocking new authorizations.
August 13, the White House used Section 232 to impose tariffs of up to 100% on Chinese-made drones and their components.
Neither move is really about drones or robots. Both are about a rare-earth magnet and a lithium cell that the United States still can't make enough of, fast enough, without China. Here's the size of the gap, and who's racing to fill it.
1. The Scale of the Shortage
The policy stack, precisely:
FCC Covered List: Chinese UAS/components (added Dec 2025) + "advanced robotic devices" — humanoids, quadrupeds (added July 28, 2026). This is a forward-looking gate — it blocks new authorizations, not devices already in the field.
Section 232 (August 13, 2026): 100% tariff on any UAS over 25kg MTOW or any drone with thermal imaging regardless of size; 25% on smaller non-thermal drones and most listed components, with additional component tariffs stepping up February 9, 2027.
Allied-nation caps: 15% for Japan, Korea, Taiwan, the EU, Switzerland, and Liechtenstein; 10% for the UK — contingent on strict "substantially all content" origin rules.
A 180-day grace period applies only to companies already on the Blue UAS Cleared List, the Blue UAS Framework, or the FCC Conditional Approval List.
Section 232 has already proven more legally durable than the administration's earlier tariff architecture — IEEPA-based tariffs were struck down by the Supreme Court in February 2026 (Learning Resources v. Trump), but Section 232 authority hasn't faced a comparable challenge.
Sizing the gap:
Consumer drones: ~15–20 million units shipped globally per year, with China supplying roughly 99% of the category.
Industrial robots: global shipments of roughly 600,000–700,000 units per year — and reducer capacity is increasingly being pulled away to serve humanoid orders instead.
Humanoid robots: Tesla alone is targeting 50,000–100,000 Optimus units in 2026, with industry-wide estimates pointing toward roughly 500,000 units/year by 2027 across Optimus, Figure, 1X, Unitree, and others.
Rare-earth magnets: the input every one of these categories shares. Pentagon suppliers told Reuters this month that US rare-earth magnet supply sits at roughly 300 metric tons against approximately 48,000 metric tons of demand — a gap that federal officials are now acknowledging can't close before the January 1, 2027 DFARS deadline banning Chinese-origin magnets from defense procurement.
Magnet pricing has moved accordingly: spot rare-earth concentrate is up 131.79% over seven quarters as of Q2 2026.
Why this matters more than the tariff schedule suggests: the tariffs hit finished units. The actual chokepoint sits upstream of anything a tariff line item touches — magnets and battery cells, both still roughly 90–99% China-sourced even for drones and robots assembled entirely on US soil.
2. Top 20 Components to Watch
A quick-reference list for anyone auditing their own supply chain. Tags:
Global: capacity-constrained everywhere, China or not
China-ban: workable alternatives exist but are blocked by FCC/NDAA rules
Price/Lead-time: non-Chinese supply exists but costs 2–10x more
China-only: almost no non-Chinese production capacity exists.
Rare-earth permanent magnets (NdFeB) —Global— up 131.79% in seven quarters
Harmonic reducers —Global— the single tightest bottleneck in humanoid production
RV reducers —Global— Nabtesco alone holds ~60% share, and can't scale fast enough
Planetary roller screws —Global— four Western makers hold ~80% of a still-tiny market
Six-axis force/torque sensors —Global— 3–6 needed per humanoid, at the wrist, ankle, and waist
Lithium cells (18650/21700/pouch) —China-only— feeds everything from vacuums to drones
LiPo battery packs (drones) —China-only— ~99% Chinese supply
Frameless torque motors —Price/Lead-time— Western equivalents run 3–5x domestic cost
Hollow-cup motors (dexterous hands) —Price/Lead-time — Swiss/German alternatives 2–4x pricier
Gimbal camera modules —China-only— DJI's ecosystem left almost no non-Chinese product line
Image-transmission chips (OcuSync-class) —China-ban— the compliant substitute is downgraded military hardware at 20–50x the cost
Motor + ESC assemblies —China-only— 4–8 per airframe
RTK high-precision GPS modules —Price/Lead-time— Western alternatives run 2–4x domestic pricing
Long-endurance power (hydrogen fuel cell, solid-state) —Global— genuinely early-stage scarcity, not a China issue
Multispectral/agricultural sensors —Price/Lead-time— 2–3x Chinese equivalents
Blue UAS-certified flight controllers —Price/Lead-time— 10–25x commercial pricing, and now structurally locked in by federal procurement rules
Anti-jam/military-grade GNSS modules —Price/Lead-time— demand rising with counter-jamming requirements
Tactical datalink radios —Price/Lead-time— swarm tactics mean multiple units per platform
Small turbojet/EDF propulsion (interceptor drones) —Global— counter-UAS is a genuinely new market with thin global capacity
LDS lidar navigation modules —China-only— low/mid-tier navigation lidar is almost entirely Chinese-sourced
3. The New Pipe: Japan, Korea, Taiwan, Mexico
Japan — upstream materials and extreme-precision control. The highest-barrier tier of the chain: Harmonic Drive Systems (near-monopoly in harmonic reducers), Nabtesco (heavy-load reducers), Toray and Teijin (aerospace-grade carbon fiber), Sony (CMOS sensors, >40% global share), Nidec (brushless motors). The catch: Harmonic Drive's own output is part of the global capacity bottleneck — Japan isn't a China alternative here so much as the actual ceiling on supply.
Korea — integrated actuators and battery manufacturing at scale. Korea moved automotive supply-chain manufacturing — steering, drive-by-wire, braking — into robotic actuators, giving it a real cost and lead-time edge over Japan. HL Mando's steering-derived actuators already sit in North American humanoid supply chains. LG Energy Solution and Samsung SDI combine for more than 20% of global battery market share, with US-based capacity already built under IRA rules — a genuine head start on onshore compliance.
Taiwan — systems integration, EMS, and semiconductor capacity.Taiwan turns components into shippable systems, fast, at volume. Foxconn is the world's largest electronics manufacturing services provider (revenue north of $200B) with existing US and Mexico footprint. TSMC holds more than 90% share of advanced-node wafer fabrication — the AI-chip chokepoint for both drones and robots. The tradeoff: Taiwan is the integrator, not the source, of the hardest-to-replace core technology, and it carries the highest geopolitical tail risk of the three.
Mexico — the nearshoring wildcard. A genuinely different mechanism from the Japan/Korea/Taiwan story: USMCA-adjacent final assembly and EMS/PCBA work rather than core-component origination. This is the least-verified part of the picture right now — we'll have a dedicated look at Mexico's role in a future issue once we've run it through trade-flow data.
4. US Onshoring and the Venture Money Behind It
The capital is moving faster than the factories.
Robotics and defense-tech funding just had a record year — and it's only August.
Global robotics startup funding hit $18.8 billion so far in 2026, already ahead of the full-year 2025 total of $15 billion and well past the prior peak of $14.1 billion set in 2021 (Crunchbase).
Defense tech funding blew past its own full-year record in the first five months of 2026 alone — more than $14.6 billion, versus $9.6 billion for all of 2025. Three megarounds drove roughly $8.75 billion of that: Anduril's $5 billion Series H, Shield AI's $2 billion raise, and Saronic's $1.75 billion Series D (Crunchbase).
For scale on how fast this has moved: defense tech funding was $1.6 billion in 2020.
The manufacturers are putting real capital behind capacity, not just announcements.
Skydio committed $3.5 billion over five years to expand US manufacturing, R&D, and supplier development — including a new domestic facility five times the size of its current footprint and more than $1 billion directed to US suppliers. The company has shipped more than 60,000 drones to 3,800+ customers and is opening production space to co-locate suppliers on-site.
Red Cat Holdings expanded total manufacturing capacity to 254,000 square feet across its BlueOps, FlightWave, and Teal divisions as of year-end 2025, and reported 646% year-over-year quarterly revenue growth on the back of the Army's Short Range Reconnaissance contract — a deal it won after unseating Skydio.
On the input side: MP Materials' Fort Worth magnet facility is scaling from 1,000 to roughly 3,000 tons/year with a DoD-backed path to 10,000 tons/year by 2028, and its planned 10X campus in Texas — a $1.25 billion investment — targets another 10,000 tons/year on top of that.
The demand signal is federal, and it's enormous. The Pentagon's Drone Dominance Program has committed $1 billion to purchasing more than 200,000 drones by 2027, scaling to 340,000 by 2028. That's the order book US onshoring is racing to build capacity against — and, per the magnet numbers above, currently can't fully supply.
5. The Software and Embodied AI Shift
Hardware-agnostic autonomy stacks — Skydio Autonomy, Athena AI — are the hedge against any single frame or vendor getting tariffed or banned out from under a fleet operator.
PX4 and ArduPilot adoption is accelerating as builders scrub proprietary Chinese firmware from their stacks entirely — which lines up directly with the FCC's forward-looking authorization gate. New models now need a demonstrably clean software provenance story just to get approved, not just cheaper hardware.
6. Outlook and Strategic Playbook
Where this is headed over the next 12–36 months:
The magnet gap is the real constraint on every timeline in this issue. Industry executives told Reuters this month that full replacement of Chinese rare-earth supply won't happen by the January 2027 deadline, and the administration is reportedly weighing whether to extend access to some Chinese materials rather than let defense procurement stall outright.
US capacity is coming, just later than the 2027 defense deadline implies: MP Materials at 10,000 tons/year by 2028, Niron Magnetics' rare-earth-free iron nitride plant online in 2027 at 1,500 tons/year, VAC's South Carolina plant targeting ~1,500 tons/year. None of these individually close a 48,000-ton gap — collectively, they start to.
The broader drone market is still expanding underneath all of this: from roughly $73 billion in 2024 to a forecast $160 billion-plus by 2030, at a projected ~14% CAGR. The supply squeeze is a rough two-to-three-year passage, not a ceiling on the category.
Procurement playbook for operators auditing their own exposure:
Map every component against the four-category shortage tags in Section 2 — a China-only battery pack and a Global-bottleneck harmonic reducer require completely different mitigation strategies.
Check Blue UAS Cleared List, Blue UAS Framework, and FCC Conditional Approval List status now — that's the only path to the 180-day tariff grace window.
Build origin documentation ahead of need. Component-laundering crackdowns mean informal assurances from a distributor won't hold up under the new origin-tracing rules.
Plan budget around the February 9, 2027 step-up on Annex III components, not just the September 3, 2026 effective date — a second cost shock is already scheduled.
7. Coming Soon
We're publishing an detail report on supply chain companie and startups in this space — separate publication, separate lens, more soon.
Physical AI Signal Report is written by David Cao, published through Physical AI Builders. F50 Capital is a supporter of this newsletter.