NextFin News - The delicate equilibrium of the artificial intelligence supply chain was jolted this week as shares of Lumentum Holdings and Coherent Corp. retreated from recent highs, following clarifying remarks from Nvidia CEO Jensen Huang regarding the future of data center connectivity. Speaking at a major industry forum on Tuesday, March 17, Huang tempered expectations for a wholesale industry shift toward optical networking, confirming that Nvidia intends to utilize a hybrid approach of both copper cabling and advanced optical solutions for its next-generation AI factories. The admission sent a chill through the photonics sector, which had spent much of the month rallying on the assumption that copper’s physical limitations would force an immediate, total migration to fiber optics.
The market reaction was swift and unsympathetic. Lumentum shares, which had been buoyed by a massive $4 billion joint investment announcement from Nvidia earlier in March, fell more than 7% in mid-day trading, while Coherent saw a similar decline. Investors had interpreted the earlier investment as a signal that Nvidia was preparing to abandon copper entirely in favor of co-packaged optics (CPO) to solve the heat and power bottlenecks inherent in massive GPU clusters. Instead, Huang’s endorsement of copper for short-reach connections within the rack suggests that the "death of copper" has been greatly exaggerated, at least for the 2026-2027 hardware cycle.
The technical tension at the heart of this volatility lies in the "reach" of the signal. Copper is significantly cheaper and more energy-efficient than optical fiber for very short distances, typically under two to three meters. As Nvidia scales its Blackwell and successor architectures, the density of these chips creates immense heat. While optical fiber can carry data further with less interference, the cost of the transceivers required to convert electrical signals to light remains a significant hurdle for data center operators. By signaling a commitment to both mediums, Huang is effectively telling the market that Nvidia will prioritize cost-efficiency and reliability over a pure-play technological leap where copper still suffices.
This nuanced stance complicates the narrative for Lumentum and Coherent, two companies that have become the primary proxies for the "optical transition" trade. Earlier this month, on March 4, Nvidia’s $4 billion commitment to these two firms was hailed as a watershed moment, intended to shore up the U.S.-based supply chain for advanced laser components and silicon photonics. That investment remains intact, and the long-term trajectory for photonics is still upward, but the timeline for total dominance has been pushed back. The volatility reflects a classic "buy the rumor, sell the fact" scenario, where the "fact" turned out to be a more pragmatic engineering reality than the revolutionary shift bulls had priced in.
For the broader AI infrastructure market, the persistence of copper is a win for traditional cable manufacturers and a reality check for the photonics industry. It suggests that the "gigawatt-scale AI factories" U.S. President Trump has championed as a cornerstone of national competitive advantage will be built on a patchwork of technologies rather than a single silver bullet. While Lumentum and Coherent are still essential partners for the long-haul and inter-rack connections that define these massive facilities, they must now contend with a market that realizes the transition to an all-optical world will be a marathon, not a sprint.
The divergence in stock performance highlights the sensitivity of "AI-adjacent" stocks to every syllable uttered by Nvidia’s leadership. As the primary architect of the current AI boom, Huang’s technical preferences dictate billions of dollars in capital allocation. For now, the message is clear: the industry will squeeze every last bit of utility out of copper before fully committing to the light. This pragmatism ensures that while the photonics revolution is inevitable, it will not be instantaneous, leaving investors to recalibrate their expectations for a more gradual evolution in data center architecture.
Explore more exclusive insights at nextfin.ai.
