AI supply-chain thesis — mapping bottlenecks, focus companies, and supply-chain exposure for investors.
**Bottleneck theme:** Photonics / Optical Interconnect **Focus:** $AAOI — Applied Optoelectronics, Inc. ## Why this matters The AI-driven move from 400G → 800G → 1.6T optical transceivers is one of the cleanest secular bottlenecks inside the data center. NVIDIA's Spectrum-X and Quantum-X switching, Broadcom's Tomahawk 6, and Marvell's 1.6T DSPs all front-end an exploding number of optical ports. **Linear Pluggable Optics (LPO)** — which strips the DSP out of the transceiver and runs the analog signal directly between switch ASIC and laser — is the architecture hyperscalers want for AI training fabrics because it cuts ~50% of transceiver power, reduces latency, and shrinks the BOM. Co-packaged optics (CPO) is the longer-term endgame, but LPO is the *bridge generation that ships in volume in 2026*. AAOI is the most leveraged listed pure-play on this transition. It is vertically integrated — designs InP, fabricates lasers, builds the photonic subassemblies, and assembles modules — and has just landed the first volume 1.6T LPO socket at a major hyperscaler. ## Q1 2026: the inflection took the share price from a small-cap to ~$11B+ market cap - **March 2026: $200M+ first volume order for 1.6T transceivers** from a major hyperscale customer. Shipments begin Q3 2026, ramp through Q4. The customer is expected to return to >10% revenue concentration — long-form code for a top-3 hyperscaler. - **March 2026: $53M 800G order** from a separate hyperscaler. - **April 2026: incremental $71M 800G order** from the same hyperscaler, taking 800G commitments at that customer to **$124M since mid-March**. - **March 2026: volume 1.6T LPO shipments started** — multiple months ahead of larger competitors. - **Management 2026 revenue guide >$1B** vs consensus ~$962M. - **Texas manufacturing capacity** expanding to **>500,000 units/month of 800G/1.6T by year-end 2026**. - Stock has gained roughly **1,100% in the past year**. ## What makes AAOI structurally different 1. **Vertical integration on InP**: Most module makers buy DFB / EML lasers from Coherent, Lumentum, or Sumitomo and have to take whatever supply comes. AAOI fabricates its own InP lasers in-house — a cost edge plus tighter control during periods of EML scarcity (which has been a chronic constraint at 800G+). 2. **LPO product / time leadership**: First volume shipper of 1.6T LPO. LPO is a stickier socket than DSP modules because the optimization is co-engineered with the switch ASIC vendor and customer firmware. 3. **U.S. + Taiwan dual-fab footprint**: After tariffs widened the China-cost penalty (60% → 70% on import wafer substrates), domestic and friendly-shore manufacturing became table stakes for hyperscaler procurement. AAOI's Texas expansion is timed for this. 4. **Microsoft Right of First Refusal (ROFR)**: Long-disclosed strategic relationship that includes a ROFR on certain transactions. Provides a theoretical floor — and a possible takeout optionality. ## What this is *not* - Not a long-term LPO monopoly. Coherent ($COHR), Lumentum ($LITE), Fabrinet ($FN) and Innolight (private) are all credibly competitive. AAOI's edge is timing, vertical integration, and design-in stickiness on the first 1.6T sockets. - Not a CPO play directly — that's NVIDIA + Coherent + TSMC + ASML/AMAT silicon photonics. AAOI's role in CPO is uncertain. ## Position in the supply chain AAOI buys raw InP wafers (from $AXTI / Sumitomo / II-VI subsidiary), pulls in passive optics and packaging materials, fabricates DFB / EML lasers in-house, and ships modules pluggable-form (QSFP-DD800, OSFP-1.6T) to switch OEMs (Arista, Cisco, Celestica) and direct to hyperscalers. The end customers are the same hyperscalers everyone is pointing at: $MSFT, $META, $AMZN, $GOOGL, plus $ORCL. ## Risks - **Customer concentration**: $200M order returning a hyperscaler to 10%+ of revenue is a feature now and a risk later. AAOI has historically had 30-50% concentration with one customer; that volatility cuts both ways. - **Valuation gap**: The stock has run ~1,100% in 12 months; many sell-side targets sit *below* current price. Wall Street Zen issued a sell. Even a strong execution outcome leaves limited near-term margin of safety. - **Dilution**: AAOI has repeatedly tapped equity to fund capex (Texas expansion alone is ~$150M). Continued issuance is likely. - **Larger competitors entering 1.6T**: Coherent and Lumentum are not far behind; once they qualify at the same hyperscaler, AAOI's share could compress quickly. - **EML supply constraint**: The industry is short on EMLs; AAOI's vertical integration helps but is not unlimited capacity. - **CPO timing risk**: If hyperscalers leapfrog from 800G LPO straight to CPO at 3.2T, the 1.6T TAM is shorter than current consensus. ## What we're watching - Q3 2026 first-shipment milestone for the $200M 1.6T order - 500k/month 800G+1.6T capacity ramp execution at Texas - Second hyperscaler 1.6T design-in announcement - Coherent / Lumentum 1.6T LPO qualification timing - Microsoft strategic actions under the existing ROFR - Any 800G-EML supply commentary on the Q1 / Q2 2026 calls
Browse all AI supply-chain theses · Macro trends · Industries · Product categories