Silicon Photonics Investment Is Ramping Fast as AI Clusters Outgrow Copper Wiring. 2 AI Stocks Stand to Win.

The investable edge is not just “more AI spend,” but a reallocation of that spend from compute to interconnect as cluster scale makes power and thermal efficiency the gating factor. That tends to help the pure-play optical suppliers first, but the bigger strategic winner is still the platform owner with the software and networking stack; if optics remove a bottleneck, NVDA can ship more full racks and preserve pricing power. The flip side is that once this becomes a standard architecture choice, component suppliers can get squeezed on mix and margin as hyperscalers dual-source aggressively.

Near term, the market will likely trade the story faster than the cash flows. The first 1-3 month catalyst is commentary on order visibility, lead times, and whether co-packaged optics is moving from lab validation to production ramps; any hesitation there would hit the higher-beta names hardest. Over 6-18 months, the real test is whether transceiver revenue converts into sustainable gross margin or simply funds more capacity while customers negotiate down pricing.

The contrarian risk is that consensus is treating this as a clean secular winner when it is still a cyclical, customer-concentrated parts business. If hyperscaler capex pauses, or if thermal/yield issues delay co-packaged optics, COHR and LITE can underperform sharply even if the long-term thesis survives. Conversely, if NVDA networking attach expands faster than optics supply, the bottleneck migrates upward and the market may prefer the platform names over the suppliers.