AXTI research

Position in the Physical Layer of AI thesis

AXTI is the compound-semi substrate pivot. The Physical Layer thesis says indium phosphide (InP) and related compound semiconductor substrates are the gating constraint for the next generation of optical interconnects (1.6T+ datacenter optics). AXT manufactures those wafers. ALMU is the anchor that BUILDS the InP photonic ICs; AXTI is the upstream wafer supplier.

Quinn's book opened AXTI on 2026-04-27. The 2026-05-05 snapshot had AXTI at +28.72%. Sized at 2% of book.

Recent catalysts (last 30 to 60 days)

$100M InP backlog disclosed (Q4 print). Management put a number to the indium phosphide order book for the first time. The number is large relative to total revenue (~30 percent of trailing twelve months).
Beijing fab restart. Production capacity ramping back after 2024 export-control overhang cleared.
Photonics customer roster. Lumentum, Coherent, and (per inferred 10-Q language) ALMU are AXTI's named customers in the InP segment.
Q1 2026 print mixed. Revenue beat, gross margin light. The margin compression is one-quarter-only per management commentary, on product mix.

The thesis (what has to be true)

1. InP wafers stay the dominant substrate for 1.6T+ datacenter optics. Silicon photonics is the lower-cost alternative; if hyperscalers accept silicon for 1.6T, AXTI's premium compresses. 2. The Beijing fab restart hits production targets. Capacity-constrained today; the fab is the load-bearing variable. 3. Photonics customer mix grows. Today AXTI is ~30 percent InP / 70 percent gallium-arsenide. The mix shift to higher InP is the margin story. 4. Export-control regime doesn't tighten. AXTI's Chinese manufacturing footprint is the structural risk.

Kill vectors (what would break the thesis)

Export controls tighten. If US-China trade policy restricts compound semi exports, AXTI's customer base in non-US markets contracts.
Beijing fab capacity miss. Production ramp slower than guidance pushes revenue beyond financing window.
Silicon photonics displacement. If silicon photonics qualifies for 1.6T before InP scales, AXTI's growth thesis breaks.
Customer in-house substrate. Coherent or Lumentum could vertically integrate substrate production. Capital-intensive but real risk.

Layer context

Test & Materials per current metadata; arguably also Photonics. AXTI is dual-layer: it makes the materials (Test & Materials) that go into the photonic ICs (Photonics). Sister names: ALMU (the InP PIC builder, anchor), LITE (Lumentum, the datacenter-optics integrator). AXTI is the cleanest substrate-supplier exposure.

Position discipline

Pivot sleeve rules.
Trim trigger: +100 percent from cost.
4-quarter thesis-fail rule applies.

Moat 6. Asym 7. Moat is moderate because compound semiconductor wafer production is capital-intensive but reproducible. Asym is high because the InP-1.6T story is concrete and AXTI is the only public pure-play.

Real money. Real position. Real receipts.