Make money doing the work you believe in
SIVES $2B mrkt cap today….LITAE $2B in sale -$70B mrkt cap….COHR $6B sales….$70B mrkt cap…imma sit on this one for a few years
. Copper (DACs/passive copper cables or emerging AECs) still dominates the shortest reaches (<1–3m) today because it’s cheap and simple for low-to-mid speeds. But at 800G/1.6T+ (the AI era), copper hits hard physical walls: signal loss, electromagnetic interference (EMI), excessive power draw/heat, and reach limits. Optics wins on bandwidth density, power-per-bit efficiency, scalability for massive GPU clusters, and total cost of ownership—especially as AI training/inference clusters scale to tens or hundreds of thousands of GPUs.
This isn’t zero-sum replacement; it’s market expansion. Optics invades copper’s turf (intra-rack, on-package, chip-to-chip) while enabling bigger clusters overall. CPO and optical I/O are the big structural shift here—NVIDIA and hyperscalers are driving it hard.Optics TAM Today and in 2030Here’s the consensus from market data, McKinsey, LightCounting, and players like Lumentum and Coherent (who are deep in the supply chain):
Optical transceivers (the broadest “optics” proxy, heavy data-center/AI weighting): ~$10–15B in 2024/2025 (data centers already ~40–50%+ of this). Projections to $21–46B by 2030–2034 (CAGR 13–17%). Some AI-specific slices grow faster.
Broader optical interconnect market (cables, modules, SiPh components, etc.): ~$16B in 2024 → ~$34–35B by 2030 (CAGR ~13–14%).
AI/data-center-specific optics (the real growth engine, including pluggables, CPO, lasers, optical circuit switches/OCS, DCI): Lumentum: ~$18B today → >$90B by 2030 (~40% CAGR). This is the most bullish (and cited) industry view—covers the full AI networking stack where optics displaces copper.
Coherent: CPO alone is a $15B+ SAM by 2030 (explicitly called “SAM expansion as optics replace copper”). Their total new-growth optics engines (CPO + OCS + thermal) add ~$20–23B on top of existing ~$50B+ base.
Data-center optics/cabling slice (intra-DC front-end/back-end, high-speed >800G/1.6T): McKinsey sees strong 14%+ CAGR through 2029, with >1.6Tbps segment alone potentially hitting ~$24B. Hyperscalers shifting ~87% of back-end to 800G+ by 2029.
Ultra-short-reach / on-package optics (the direct copper-killer): Smaller today (~$2B) but growing fast to several billion as CPO and optical chiplets ramp.
Copper’s current footprint (for context on the shift): Short-reach copper cables/DACs/AECs are still meaningful—projected >$10B TAM by 2029 for <3m links, with AEC modules alone going from ~$0.7B (2025) to ~$3.6B by 2034. Data-center wire/cable overall (copper + fiber) is ~$13–22B now. But its share erodes at high speeds; copper is increasingly limited to the very shortest/lowest-power niches while optics expands the total pie.
Bottom line on 2030 TAM: The optics opportunity in AI data centers is widely expected in the $70–90B+ range annually (per Lumentum/Coherent/LightCounting views), up 5–6x from today’s AI optics base. This is driven by the copper-to-optics migration plus overall AI cluster growth. Conservative transceiver/interconnect forecasts land ~$25–35B, but the AI-specific numbers from suppliers on the ground are materially higher because CPO/ELS/optical I/O multiply component density (more lasers, PICs, etc. per rack).SIVE’s lasers are a high-margin, critical choke-point component in this stack—especially the high-power CW DFB arrays for ELS in CPO/SiPh platforms. Volume ramps with every hyperscaler deployment, and SIVE is already shipping into prototypes with multiple partners. The thesis has legs if they execute on capacity and design wins. Risks include competition (Coherent, Lumentum, etc. also make lasers), execution on scaling their Glasgow fab/partner ecosystem, and any delays in CPO standardization. But the macro tailwind is massive.