NextFin News - NVIDIA has committed $4 billion in strategic investments to Coherent and Lumentum, a massive bet on the optical plumbing required to sustain the next generation of artificial intelligence. The deal, announced this week, splits the capital equally between the two photonics leaders, with $2 billion earmarked for each to accelerate U.S.-based manufacturing and research into silicon photonics and laser components. By securing these critical links in the supply chain, NVIDIA is moving to eliminate the looming bandwidth bottlenecks that threaten to stall the expansion of "gigawatt-scale" AI factories.
The investment marks a fundamental shift in how the world’s most valuable chipmaker views its role in the ecosystem. No longer content to simply design the silicon, NVIDIA is now financing the infrastructure that allows those chips to communicate. As AI models grow in complexity, the traditional copper-based electrical signals used to move data between processors are hitting physical limits. Light-based optical interconnects offer the only viable path to the massive bandwidth and lower power consumption required for the Blackwell architecture and its successors. According to CNBC, NVIDIA CEO Jensen Huang described the partnership as a move to build the "world's most sophisticated silicon photonics."
For Coherent and Lumentum, the $2 billion infusions are transformative. The agreements are multiyear and non-exclusive, but they come with multibillion-dollar purchase commitments and future capacity rights. This effectively de-risks the massive capital expenditure required to build advanced fabrication facilities on American soil. Lumentum, which has long been a key supplier for telecommunications and consumer electronics, will now pivot more aggressively toward the data center. Coherent will use the funds to scale its production of 800G and 1.6T optical transceivers, the high-speed "translators" that turn electrical data into light pulses.
The timing and geography of the deal are equally significant. By mandating an expansion of U.S. production, NVIDIA is aligning itself with the broader industrial policy of U.S. President Trump’s administration, which has prioritized domestic semiconductor and high-tech manufacturing. This "onshoring" of the optical supply chain provides a hedge against geopolitical instability in Asia, where much of the world’s optical assembly has historically been concentrated. It also ensures that the most sensitive research and development into next-generation laser technology remains within the domestic sphere.
The competitive landscape is also reacting. NVIDIA’s move follows a $3.25 billion acquisition of photonics startup Celestial AI by Marvell Technology last year, signaling that the industry’s heavyweights are in a frantic race to own the "optical layer." While NVIDIA’s H100 and B200 chips currently dominate the market, the real battle is shifting toward the interconnects. If a data center cannot move data between its 100,000 GPUs fast enough, the raw compute power becomes a stranded asset. By locking up the capacity of the two largest players in the space, NVIDIA is effectively building a moat around the physical connectivity of the AI era.
The financial implications for the broader market are stark. This $4 billion outlay represents a significant portion of NVIDIA’s cash reserves, yet it is a drop in the bucket compared to the hundreds of billions in revenue the company expects from data center sales over the next three years. The real winners are the specialized equipment manufacturers and materials scientists who underpin the photonics industry. As the industry moves toward 1.6-terabit speeds and eventually co-packaged optics—where the laser is integrated directly onto the chip—the technical barriers to entry will only rise. NVIDIA has decided that in the race for AI supremacy, it is better to own the light than to be left in the dark.
Explore more exclusive insights at nextfin.ai.
