The Next Wave Of AI Is Optics, And The EUV ETF Owns It

Coverage initiation: the Corgi Lithography & Semiconductor Photonics ETF is rated Buy, targeting AI infrastructure demand with a stated shift from memory to optics. The fund provides concentrated optics exposure, with optical/optical interconnect exposure at ~29% weight, aiming to avoid memory-heavy names and Chinese transceiver suppliers. The note argues that recent regulatory developments plus Nvidia’s platform evolution should favor US/European optics suppliers as the optics market grows alongside infrastructure buildout.

Analysis

This is more a capex-basket rotation than a clean single-name earnings catalyst. The market implication is that incremental AI spending is moving up the stack toward optical engines, packaging, and interconnect, which should expand margins for a narrow set of photonics vendors while compressing pricing power for commoditized memory and lower-end transceiver suppliers. The immediate effect is usually multiple expansion in the “picks-and-shovels” names, but only if hyperscaler budgets are not already fully allocated to compute nodes.

The second-order loser is any supplier dependent on a finite AI infrastructure dollar that does not participate in the optics content shift. That argues for relative shorts in memory-heavy exposure if HBM pricing rolls over, but not outright sector shorts unless cloud capex slows. For NVDA, the read-through is mixed: the platform keeps pulling the ecosystem forward, but if optical attach rates rise faster than accelerator unit growth, some value leaks away from the GPU layer into the network and packaging stack.

Catalysts are 1-3 months: platform announcements, hyperscaler capex commentary, and export-rule headlines that can re-rate US/EU optics names versus China-linked transceiver supply chains. The 6-18 month thesis is stronger if optical content per rack keeps compounding; it breaks if memory remains the larger bottleneck or if NVDA’s roadmap preserves cheaper electrical interconnect longer than expected. The consensus may be underestimating how quickly optics becomes a budget constraint rather than just a growth theme.

This is not a blanket bullish semiconductor call. If anything, the move is likely underdone in the photonics subset and overdone in broad semis, because the incremental dollars are becoming more selective and less evenly distributed across the stack.

Key Decisions for Investors

• Initiate a relative-value long COHR / LITE basket vs short MU over the next 1-3 months; the thesis is capex mix shift toward optics with cleaner earnings sensitivity than memory. Risk/reward is attractive if optics attach rates re-rate before memory pricing reaccelerates; falsifier is a renewed HBM shortage or strong MU guidance.
• Buy a 3-6 month call spread on COHR on any pullback rather than chasing strength; this limits premium burn while preserving upside if hyperscaler capex confirms the optics thesis. Stop if the name fails to outperform SOXX over the next earnings cycle.
• Use SMH as a hedge, not a core long, until the market proves optics content is additive rather than budget-neutral. If SMH lags a photonics basket by 5%+ over 4-8 weeks, that supports a stronger pair trade.
• Treat NVDA as a monitor, not the primary expression: if NVDA commentary shifts toward higher optical attach rates or network bottlenecks, add to the optics basket; if not, fade the trade and rotate back into compute leaders.
• Watch for China/export-policy headlines; easing there would directly weaken the US/EU supplier advantage and is the main event-driven downside to the relative-value long.