Back to News
Market Impact: 0.35

Nvidia CEO Jensen Huang Highlighted a New AI Bottleneck. 3 AI Stocks That Could Benefit.

Artificial IntelligenceTechnology & InnovationCompany FundamentalsAnalyst Insights
Nvidia CEO Jensen Huang Highlighted a New AI Bottleneck. 3 AI Stocks That Could Benefit.

Nvidia is positioning for AI networking bottlenecks by investing $2B each in Coherent and Lumentum plus $1B in Nokia, underscoring rising demand for faster data transfer via silicon photonics. Coherent’s results cited include +27% revenue to $1.8B, +55% EPS to $1.41, and data center/communications revenue up 41%, while Lumentum reported revenue nearly doubled to $808.4M and guides $960M–$1.0B for the current quarter. The article is constructive but flags concentration risk for Coherent (data center/communications is 75% of revenue) and intense competition for Lumentum (forward P/E ~50).

Analysis

This is less about one-off supplier wins and more about the market discovering that AI monetization is now constrained by data-movement economics. That is favorable for the optical stack, but the largest near-term beneficiary may still be NVDA because it can steer standards and lock up capacity before the bottleneck becomes widely priced. The second-order effect is better inference economics for hyperscalers: lower latency and power per bit should improve AI unit economics and extend capex budgets rather than simply shifting spend from GPUs to optics. COHR looks like the cleaner expression because it has strategic scarcity value and less obvious multiple risk than the more crowded names. LITE has more operating leverage, but that also means it is the most vulnerable to any delay in qualification, design wins, or pricing pressure as competitors fight for the same slots. NOK is the weakest-quality way to play the theme: if this remains a side project instead of a real earnings stream, the stock can give back gains quickly once the narrative fades. Consensus is likely underestimating the lag between strategic investment headlines and actual revenue capture. Photonics capacity, customer qualification, and network architecture changes usually take quarters, not weeks, so the immediate move may outrun the fundamentals. The thesis breaks if next-cycle guidance from data-center and networking suppliers shows slower bookings, or if hyperscalers pivot toward software/network optimization that reduces hardware intensity.