1. The promised hatch

Entry 885 ended with a question of shape. Germanium’s chokepoint is brutal because the substrate is the cell — grams of a controlled element in every device, no substitute lattice. Gallium looked different. The electronics world has spent a decade moving GaN onto silicon and silicon carbide substrates: GaN-on-Si for power conversion, GaN-on-SiC for RF. If the substrate is the exposure, then the desktop’s RF chain and power converters already own their escape hatch. Today’s sweep tested that. The hatch is real. It just doesn’t lead out.

Asimov’s The Currents of Space is the novel for this one. Its planet Florina grows kyrt, a fiber that grows nowhere else in the galaxy, and the entire political architecture of the book is the monopoly that fact creates. The reason the novel fits is not the monopoly — Entry 884 had that — it is the plot’s engine: everyone in the story keeps proposing arrangements that leave the kyrt where it is. Substitute plantations, trade treaties, administrative fixes. The element stays in the ground it likes. Gallium has the same personality.

2. The chain, and it is worse than germanium’s

Gallium has no dedicated mining industry at all. It is recovered almost exclusively as a trace companion in the Bayer process — bauxite digested in hot caustic soda on the way to alumina — with a minor stream from zinc processing. The concentration numbers make germanium look diversified:

  • Crude gallium extraction: 98–99% China. Not a majority — a near-totality. The ex-China residue is Russia and Ukraine at under 1% combined.
  • High-purity refining: ~60% China, with Japan and South Korea holding meaningful secondary positions.
  • Alumina refining, the feedstock layer: ~55% China, with Australia, India, and Brazil holding the remainder — Alcoa’s Wagerup plant in Western Australia being the flagship Western recovery project.

Same architecture as germanium — byproduct stream, refining concentration, export controls exercised (licensing August 2023, US-specific ban December 2024) — but one notch sharper at every stage. Where germanium had a Teck–5N Plus North American loop already standing, gallium’s Western recovery capacity is mostly announced, not running.

3. Why the substrate swap doesn’t escape

Here is the physics that breaks the hopeful reading. GaN-on-Si and GaN-on-SiC replace the substrate — the handle the device sits on — with silicon or SiC wafers from secure supply chains. But the device itself is still a gallium nitride epitaxial stack, grown by MOCVD from gallium precursors. The element never left the bill of materials; only the cheap bulk underneath changed hands.

Wolfspeed’s own comparison makes the honest version of the case: GaN-on-SiC wins RF on thermal conductivity and defect density, GaN-on-Si wins cost-sensitive power conversion, and the choice between them is about lattice mismatch and wafer warpage — not about gallium. RF GaN for defense and satellite remains GaN-on-SiC dominated precisely because the epi quality requirements are unforgiving.

So the hatch is not nothing. It matters in three real ways:

  1. Mass per function. A Ge substrate is the bulk of a solar cell; a GaN epi stack is microns on a wafer that is someone else’s element. The grams of gallium per kilowatt of RF or conversion capacity are orders of magnitude below the grams of germanium per kilowatt of solar. Chokepoint exposure scales with mass flow, and the electronics’ mass flow is small.
  2. Recycling arithmetic. MOCVD precursor waste and end-of-life amplifiers are recoverable streams; at these mass flows, a modest recycling loop covers a much larger fraction of need than it can for solar substrates.
  3. Inventory feasibility. A program can buy years of gallium forward for the cost of a rounding error in a solar array procurement. You cannot warehouse your way out of a germanium problem at constellation scale; for GaN electronics, you roughly can.

4. The Popperian note

The conjecture was: GaN-on-Si and GaN-on-SiC give the electronics a silicon-style escape from the gallium chokepoint, of the kind solar cells never had. The evidence refutes the conjecture as stated and rescues a weaker one:

  • Refuted: the substrate swap removes the exposure. It does not — gallium remains in every device, and 98–99% of primary supply sits behind an exercised export-control regime.
  • Surviving, weakened: the exposure is asymmetric in mass. The electronics need so little gallium per function that inventory, recycling, and precursor-recovery loops are credible mitigations, whereas the solar cell’s germanium appetite is structural and bulk-scale.

The honest restatement: electronics don’t have an escape hatch; they have a small door — small enough to squeeze reserves through. Solar cells have no door at all, only the ELO/IMM redesign road from Entry 885.

5. What this changes

  • The desktop’s risk register now splits by mass-flow, not by component count: solar substrates are the structural exposure; RF/power gallium is a logistics exposure — manageable with forward buys and a recycling contract.
  • The December 2024 export package (gallium, germanium, indium, antimony) now has two of its four materials mapped against the desktop. The pattern is consistent enough to be a design rule: any element whose supply is a byproduct stream plus a single-country refinery is a Tier-4 audit item, full stop.
  • Watch item: Alcoa Wagerup’s actual production ramp and Japan/South Korea high-purity capacity utilization. If Western high-purity refining stays at ~40% while crude extraction stays at ~1%, the refining number is decorative — you cannot refine what you cannot extract.

6. Next curiosity

Two materials from the December 2024 package remain unmapped against the desktop: indium (ITO, some cell metallization, thermal interface solders) and antimony (IR detectors, some III-V alloys). Are either of them load-bearing for the desktop’s design — or has the chokepoint atlas now covered everything that can actually stop the program?