AXTI

Current Thesis

This is the InP-substrate pick-and-shovel for the AI optical-interconnect boom, and our prior read was wrong. We ran 3x DEFER (4/19–5/15) on "management top-ticked the squeeze with the $64.25 secondary — no chase." In hindsight that $550M raise (8.56M shares, 4/21) funded the InP capacity doubling that re-rated the stock from [entry redacted] to a $143.16 ATH on 5/26. The squeeze framing was a category error: this is a real, accelerating revenue inflection, not a thin-float retail pop. Theme is ACCELERATING. Stock is now ~$103–105 (6/1–6/4), -28% off the ATH — a pullback to the rising 20-EMA, not a broken setup. We missed the $64→$143 leg; the job now is a disciplined re-entry, not a blind mid-chop chase.

Bull Case

• Q1 2026 (reported 4/30): revenue $26.9M, +39% YoY; InP revenue $13.6M = >50% of total for the first time; non-GAAP gross margin 29.9% vs 21.5% prior quarter — margin inflecting with mix shift to InP.
• InP backlog hit a record $100M (Q1 call 4/30), up from ~$60M cited weeks earlier — hard visibility into the AI data-center optical build-out.
• Q2 2026 guide (4/30): revenue ≥$34M and first GAAP + non-GAAP profitability in years — the inflection is guided, not hoped.
• Capacity doubling in 2026 and again in 2027 (4/30 call) — supply-constrained InP = pricing power as 800G→1.6T transceivers ramp.
• NVIDIA's 3/2/2026 $2B-each optical partnerships with Coherent and Lumentum ($4B total) validate optical interconnect as THE AI bottleneck; AXT sits upstream supplying the InP substrate those lasers are built on.
• Hyperscaler capex +67% YoY in 2026 (Stifel, May) with most incremental interconnect now optical → InP demand is structural, not a meme.
• Cluster confirmation 6/2: Aehr (+15%), AAOI, AOSL, AXT all bid together — sector-wide optical/compound-semi move, not a lone-name squeeze.
• Northland PT raised to $125 from [entry redacted] (Outperform, early June); estimates lifted 5/8 — sell-side is catching up, the documented narrative-momentum edge.

Bear Case

• -28% from the $143.16 ATH (5/26) to ~$103 (6/1) in days; 6/3 intraday range $103.81–$115.12 — violent, no confirmed higher-low. Could be distribution, not a flag.
• Valuation ~50–65x sales: ~$5–7B market cap on ~$27M quarterly revenue. A single InP order push-out re-rates this in half.
• RSI ~73 (6/4) still elevated; previous close $115.70 → $103 = a failed bounce, momentum cooling.
• 52-week range $1.47–$143.16 shows a thin, reflexive name that round-trips hard — the reflexivity that took it to $143 cuts both ways.
• Story is 100% forward backlog/guide; Q1 revenue is still only $26.9M. Any slip on the 2026 capacity-doubling timeline breaks the thesis instantly.
• We were structurally late (3x DEFER, missed $64→$143). Chasing mid-pullback with no structure is the lowest-quality way to express a thesis we already missed once.
• China/Tongmei subsidiary geopolitics + STAR Market IPO timeline remain an overhang/wildcard.

Setup & Price Structure

• Archetype reclassified 6 → 2 (Picks & Shovels). No longer a thin-float retail squeeze — it's a multi-billion-cap InP substrate supplier with a real revenue inflection. Drop the 1%/name squeeze cap, but the volatility still demands disciplined sizing.
• Levels: ~$103–105 now (6/1–6/4); ATH $143.16 (5/26); secondary clearing $64.25 (4/21) decisively reclaimed — the old "re-base above $64.25 on >1.5x vol" trigger FIRED in May and we missed it.
• Rising 20-EMA est. ~$95–110 (data noisy; the quoted 20-SMA $60.61 is stale and lags the move badly). Price is sitting on/just above the rising 20-EMA = textbook pullback-to-support for an accelerating name.
• Clean entry trigger: higher-low above ~[entry redacted] then reclaim of $115–120 on >1.5x relative volume → upgrade to HIGH and size up. Blind add into the $103–115 chop = MEDIUM probe only.
• Stop discipline: daily close below $90 = pullback became distribution → abort. No averaging down.

Catalyst Calendar (next 30 days)

• Continuous (June): photonics/AI-optical news flow + EUV photonics-ETF inflows; Benzinga attention stack still printing (6/2 "moving higher"). Flow-driven, no single date.
• ~2026-06 (est.): optical/data-center conference read-throughs (OFC-adjacent, hyperscaler capex updates) — watch COHR/LITE/AAOI tape as real-time confirmation.
• 2026-07-29 (est.): Q2 2026 print — the next hard binary (guide ≥$34M rev + first profitability in years). This is OUTSIDE the 30-day window, so no near-term earnings blackout; do NOT pre-position until ~3 trading days out.
• No dated catalyst inside the 6/4–7/4 window → catalyst_date null. Trade the tape/structure, not an event.

What Would Change Our Mind

• Upgrade to HIGH / size up: higher-low above ~$100 + reclaim $115–120 on >1.5x volume, with COHR/LITE/AAOI confirming.
• Abort / stay out: daily close below $90 (loses rising 20-EMA/base), OR Q2 (7/29) revenue under the $34M guide, OR InP backlog declines QoQ from the $100M record.
• Theme-break tell: if the photonics cluster (COHR/LITE/AAOI) rolls over while AXT holds, the InP-specific bid is hollow — fade. Conversely, AXT lagging a ripping cluster = single-name problem.
• Saturation watch: CNBC mainstream coverage + a second photonics ETF + retail-euphoria peak = trim trigger, not a buy.

Correlation Notes

• Direct cluster (real-time read-through): COHR (Coherent), LITE (Lumentum) — the InP laser makers AXT supplies, both NVIDIA-backed $2B each (3/2). AAOI (Applied Optoelectronics) and AEHR (Aehr Test) move in lockstep (6/2).
• Theme ETF: EUV (photonics ETF, launched May) — AXT is a holding; ETF inflows = mechanical bid, outflows = mechanical pressure.
• Upstream demand: NVDA optical capex + hyperscaler capex (+67% YoY 2026) is the ultimate driver — watch GB300 / 1.6T transceiver ramp commentary.
• Legacy read-through now weaker: IQE.L (GaAs) was the old GaAs-ASP tell, but AXT's story has shifted to InP/data-center; COHR/LITE are now the cleaner reads.
• Negative correlation: a sharp risk-off or AI-capex-pause headline (e.g., a hyperscaler capex cut) hits this harder than the index given the 50–65x sales multiple.