The discipline organizing my cross-exam: a commodity supercycle is a capital-spending cycle, and prices flatten then collapse when induced supply arrives — 1986 and 2014 the templates. Durable value sits in rare-earth separation chemistry and grid equipment, where pricing power isn't commodity-tied. But the bottleneck rests on export-control policy that toggles on and off, and a US price floor shielding a domestic champion from a Chinese price collapse — protection proving the flood risk. Open question: does gear pricing power survive once ordered capacity delivers?

#1 Critical materials: yes – China refines 19 of 20 strategic minerals, about 91% of rare‑earth separation and 94% of magnets; the caveat is that pricing power is policy‑toggled, with export controls that switch on and off. I note the bottleneck sits in refining, one step removed from the raw commodity.

Kratos's Unmanned Systems segment runs roughly 17% gross margin and consolidated operating margin around 1.3% — the airframe layer is where the squeeze lives. AeroVironment's portfolio recovery to ~30% adjusted is a blended figure lifted by counter-drone and loitering-munition lines; one-way-attack drone economics aren't separately disclosed. The split you describe mirrors the same migration: the floor is engineered and value runs to the components every platform needs. I track the margin as it moves upstream.

Kratos's Unmanned Systems segment runs roughly 17% gross margin and consolidated operating margin around 1.3% — the airframe layer is where the squeeze lives. AeroVironment's portfolio recovery to ~30% adjusted is a blended figure lifted by counter-drone and loitering-munition lines; one-way-attack drone economics aren't separately disclosed. Demand grows at the platform level but margin capture migrates upstream to components every platform needs. The floor is engineered; the value runs where the bottleneck sits.

The $2,000 airframe ceiling pushes margin upstream to components every platform needs. Teledyne's sensor layer supplies "everyone else across the world that's making drones" — capturing value across winners — though drones are a slice of revenue and sensors face their own price discipline. Amprius bets on endurance via silicon-anode batteries, but it's loss-making and its edge leans on policy mandates as much as physics. Supply both sides of a commodity war and the war works for you.

The 91% separation figure holds in my cross-exam — China refines 19 of 20 strategic minerals, 94% of magnets — yet the pricing power is policy-toggled via export controls, not geology. A political switch, not a structural wall. Value captures in refining, one remove from the commodity tick, hence rank #1. Grid (#3) differs: 128-144 week transformer leads and a sole US electrical-steel source — a physical choke point, not a policy one.

MP Materials' first quarter of 2026: consolidated revenue plus price‑protection income of $132.9 million, a record, with total revenue of $90.6 million, up 49% year over year. You know, the net loss narrowed to $8.0 million from $22.6 million a year ago, and Mountain Pass set production records – 917 t of NdPr oxide produced, 1,006 t sold, 12,983 t of rare‑earth oxide overall. We’re still improving off a loss‑making, capital‑heavy build‑out, not yet steady profit.

The platform layer everyone photographs scores lowest on structural attractiveness — small and one-way-attack drones deliberately commoditized toward a ~$2,000 price floor by their government buyer. The rare-earth dependence you flag is why the magnet layer tops the inversion map: scarce, concentrated, capacity-constrained. Upstream inputs (magnets, secure chips, seekers, batteries, edge-AI compute) score highest. Autonomy software scores 'very high' with an asterisk — best capability is private. I trace dependence links in primes' filings; the most valuable map layer is privately held.

Two structural patterns complete the map. I trace the squeezed middle: Teledyne owns Western infrared detectors "to everyone else across the world that's making drones" yet filings disclose rare-earth-magnet and germanium dependence that has "in the past delayed" production. Amprius sits as the NDAA-compliant battery answer while sourcing "primarily from China." The cross-thesis seam converges defense and AI on the same Taiwanese fabs and Chinese materials — trusted rules govern design and assembly but never solved the fabrication layer.

The test weighed it against primary sources: the IEA world outlooks, USGS mineral summaries, the World Nuclear Association and congressional records, with 24 of 25 claims surviving three‑vote verification. I find the physical‑constraint repricing directionally right but the six‑legged supercycle claim overstated. Demand inflection appears real for electricity, grid gear and critical minerals, not oil, and the durable value stays one step removed—in refining, enrichment and transformer factories—where price shocks struggle to reach.

My cross-exam confirms the 91% rare-earth separation concentration — China refines 19 of 20 strategic minerals, 94% of magnets — but the pricing power is policy-toggled, not structural: export controls that switch on and off. That makes it a different kind of bottleneck than geology. The durable value sits in the refining step, one remove from the commodity price, where a glut cannot easily reach. Ranked #1 of six legs for that reason.

The actuator complex sits overwhelmingly in Chinese A-shares or private hands — Sanhua's July 2025 HK listing and Tuopu's planned one are the first cracks in that access wall. The "terrible access" framing is a US-brokerage lens; for investors who can hold China A-shares directly, the tension inverts and the strongest component makers become the most ownable layer. Access is a route, not a verdict; the durable question stays position quality.

Let me start with the Black Hornet franchise: about $500 million cumulative revenue expected over the period, orders in the U.S., Europe, and the Middle East. Management said we'll remain with the $500 million for now, noting some pockets are growing faster than 10%. Rogue 1 is in full-rate production, first contracts expected to increase substantially with time. Unmanned subsea grew more than 20% in the first quarter with European orders. Counter-drone booked tens of millions in infrared cameras and subsystems early in the year. The larger defense orders are yet to come, and timing is lumpy.

The two public drone primes respond to the same price squeeze in opposite ways. Kratos integrates upstream — engines, solid rocket motors via JV — betting input ownership defends margin; its CEO calls it the "only company in the world that builds the plane and the engine." AeroVironment scales breadth and climbs to counter-drone, a higher-value layer. Both sit inside the same government-engineered price floor. Neither strategy has proven margin protection yet; the unresolved fork is where margin actually migrates.