Stocks In AI Bottleneck Soar 1,000%. Now Nvidia Is Demanding 20X More Supply

Rosenblatt says Nvidia asked suppliers for a 20x increase in indium phosphide laser capacity through 2030, while suppliers are only willing to commit to 12x, underscoring how quickly optics is becoming a critical AI bottleneck. The firm sees the InP optics market expanding from about $1.9 billion in 2025 to $22.75 billion by 2030, with Lumentum projected at $9.0 billion of capacity and Coherent at $4.3 billion. The report is constructive for optics suppliers tied to AI networking demand, but it also highlights execution and overbuild risk.

Analysis

The market is likely still underpricing the second-order constraint here: not just more optics demand, but a manufacturing bottleneck that will re-rate the entire optical supply chain for years. The most attractive exposure is not the highest-beta “AI optics” names, but the companies with a credible path to scale manufacturing without torching margins. That favors vertically integrated vendors with process control and wafer-level efficiency gains; it also suggests the real P&L inflection may arrive in 2027-2030 rather than immediately, as hyperscaler buildouts translate into committed supply agreements and capacity utilization.

Competitive dynamics matter: if Nvidia is effectively pre-booking future laser capacity, suppliers with tighter customer relationships and broader networking portfolios should gain negotiating leverage, while subscale component players risk being squeezed on price once capacity catches up. The bullish read on Coherent and Lumentum is that they can convert demand into operating leverage; the bearish read on Applied Optoelectronics is that a lot of the upside may already be reflected in the stock, with execution risk around product mix and next-gen transceivers. A less obvious winner could be Broadcom if CPO adoption shifts value from discrete optics toward silicon-photonics integration and switch/routing content.

The main risk is a familiar one: supply additions outrun end-demand and the optical cycle reverts faster than investors expect. That risk is not imminent over days, but it becomes meaningful over 12-24 months if hyperscaler capex growth normalizes, AI inference economics improve less than expected, or competing architectures reduce bandwidth intensity per model. Another near-term catalyst for volatility is any sign that customers are delaying CPO qualification, which would compress multiples in the more speculative names first.

The consensus may be overconfident that every optics participant benefits equally. In reality, the winners will be those with durable yields, packaging expertise, and balance-sheet capacity to fund buildout; everyone else is just a call option on a capital cycle. The setup supports a barbell: own the quality compounders, fade the story stocks where valuation already discounts perfect execution.

Key Decisions for Investors

• Go long COHR vs. short AAOI for a 3-9 month pair trade: Coherent has clearer operating leverage and manufacturing scale, while AAOI appears most vulnerable to multiple compression if AI optics expectations merely meet, rather than exceed, current pricing.
• Accumulate LITE on pullbacks over the next 4-8 weeks: highest direct exposure to CPO content with the best mix of strategic positioning and capacity expansion, but size modestly given cyclicality risk.
• Initiate a long AVGO position as a structural hedge on the supply chain shift: if optics value migrates toward integrated silicon-photonics and networking silicon, Broadcom captures wallet share with lower execution risk than the pure-play optics names.
• Use LWLG only as a small speculative option ticket, not a core equity position: upside is large if CPO adoption broadens, but the stock is highly sensitive to qualification delays and any slowdown in hyperscaler capex.