The Market Doesn’t Know CPO Yet

Damnang · Damnang’s Substack · June 07, 2026 at 00:52 · ⏱ 20 min read  | Read on Substack ↗
Summary
The article argues that the co-packaged optics (CPO) supply chain is entering its early production ramp, and most investors lack the granular understanding of bottlenecks and value layers needed for the next phase of the optics trade. It maps the entire chain from InP substrates to system assembly, highlighting where capacity expansions, monopoly suppliers, and technology races create asymmetric investment opportunities, and teases specific stock selections to follow in a later section (not included in the provided text).
  • •AXT's Q1 2026 InP revenue hit $13.6M (over half of total revenue) with a backlog exceeding $100M; the company raised $632.5M to double InP capacity in 2026 and again in 2027.
  • •Soitec’s Smart Cut technology effectively locks up the SOI wafer market for photonics, with no viable alternative at production scale.
  • •Tower Semiconductor disclosed $1.3B in SiPho revenue contracts for 2027 with $290M in prepayments, signaling strong open-foundry demand outside TSMC’s COUPE bundle.
  • •TSMC’s COUPE volume production is scheduled for 2026, and its SoIC packaging capacity is heading toward 30,000–40,000 wafers per month by year-end, creating a single-vendor bottleneck.
  • •NVIDIA signed a partnership with Corning to expand optical connectivity capacity tenfold, with cumulative investment commitments reported up to $3.2B, validating connector bottlenecks.
  • •Lumentum placed multiple orders for Aixtron’s G10-AsP MOCVD equipment, the first public signal of in-house epi expansion for high-power CW lasers.
  • •High-power CW lasers for CPO (400mW+) are 6–11x stronger than existing SiPho transceivers (30–70mW), and ELS modules are designed to be field-replaceable, creating recurring replacement demand.
  • •The CPO supply chain splits into architecture-dependent layers (engines, COUPE packaging, connectors) and architecture-independent layers (InP substrates, epi, MOCVD equipment, fiber materials) that benefit regardless of CPO penetration.
Read time 20 min
Length 20,811 chars
Category finance
Ideas
Damnang Substack author, Damnang’s Substack
Article states SOI for photonics is 'effectively down to one' supplier — Soitec's Smart Cut. Every optical engine requires an SOI wafer, making Soitec a monopoly beneficiary of CPO volume growth regar
Article states SOI for photonics is 'effectively down to one' supplier — Soitec's Smart Cut. Every optical engine requires an SOI wafer, making Soitec a monopoly beneficiary of CPO volume growth regardless of architecture. Risk: Single-point dependency; any technology shift away from SOI (e.g., bulk silicon or other platforms) could erode relevance.
SOI
Damnang Substack author, Damnang’s Substack
Article identifies TSMC's COUPE and SoIC packaging as the highest single-vendor dependency in the CPO supply chain, with capacity set to reach 30-40k wafers/month by end of 2026. TSMC captures the pac
Article identifies TSMC's COUPE and SoIC packaging as the highest single-vendor dependency in the CPO supply chain, with capacity set to reach 30-40k wafers/month by end of 2026. TSMC captures the packaging bottleneck and benefits from NVIDIA's CPO ramp. Risk: Capacity shared with other demand; any yield issues at the packaging seam could delay CPO adoption and impact TSMC's revenue recognition.
TSM
Damnang Substack author, Damnang’s Substack
Article quantifies AXT's InP revenue surge ($13.6M in Q1 2026), $100M+ backlog, and $632.5M capacity expansion. InP substrate demand is structural from CPO and is a multi-year bottleneck with brownfie
Article quantifies AXT's InP revenue surge ($13.6M in Q1 2026), $100M+ backlog, and $632.5M capacity expansion. InP substrate demand is structural from CPO and is a multi-year bottleneck with brownfield expansion advantage over Sumitomo. Risk: Capacity execution risk; GaAs line conversion may lag; customer concentration on a few laser makers.
AXTI
Damnang Substack author, Damnang’s Substack
Article notes NVIDIA signed a partnership with Corning to expand optical connectivity capacity tenfold with up to $3.2B in commitments, explicitly calling connectors a 'bottleneck candidate.' Corning
Article notes NVIDIA signed a partnership with Corning to expand optical connectivity capacity tenfold with up to $3.2B in commitments, explicitly calling connectors a 'bottleneck candidate.' Corning is the primary beneficiary of pre-secured fiber/connector demand. Risk: Execution on tenfold capacity expansion; potential competition from SENKO, Browave, or other connector suppliers.
GLW
Damnang Substack author, Damnang’s Substack
Aixtron holds >90% share of MOCVD for InP epitaxy. The article cites Lumentum's multiple G10-AsP orders as the first public signal of epi expansion, and equipment orders are the earliest inflection po
Aixtron holds >90% share of MOCVD for InP epitaxy. The article cites Lumentum's multiple G10-AsP orders as the first public signal of epi expansion, and equipment orders are the earliest inflection point for the CPO cycle. Risk: Concentration risk; if an alternative MOCVD platform emerges (e.g., Veeco) or if CPO ramp disappoints, orders may slow.
AIXXF
More from Damnang’s Substack
MicroLED: The Optical Technology All Three Memory Makers Are Watching
Jun 04, 08:01
The Company That Shows Up No Matter Which Sector You Look At: A Full Anatomy of STM
Jun 02, 16:13
Why Is IBM Rising Again
Jun 01, 22:34
Cerebras WSE-3: The Technical Achievement and the Physical Ceiling
May 31, 00:27
Edge AI Investing Guide: Where Capital Goes After Cloud AI
May 29, 07:15

This newsletter, published June 07, 2026, features Damnang discussing SOI, TSM, AXTI, GLW, AIXXF. 5 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Damnang  · Tickers: SOI, TSM, AXTI, GLW, AIXXF