Nvidia Bets $2B on US Optical Factories

This move is larger than it looks, totaling $4 billion with a parallel $2 billion investment and multi-year purchase agreement with Lumentum. It's a strategic vertical integration play, moving Nvidia from simply designing chips to controlling the critical data-moving infrastructure that connects them, a significant step in owning the entire AI hardware stack. The investment directly targets a fundamental physics bottleneck in scaling AI. As GPU clusters for training massive models grow, traditional copper wiring suffers from signal degradation and high power consumption, creating a data-transfer wall. Optical interconnects are essential for the multi-terabit bandwidth needed to keep thousands of GPUs fed with data. A key technology being accelerated is Co-Packaged Optics (CPO), which integrates optical transceivers directly onto the same package as network switch chips. This approach drastically cuts power consumption and reduces signal loss from over 22 decibels to just 4, a critical gain for the massive east-west traffic patterns inside AI data centers. For robotics, this hardware layer is foundational. Training general-purpose models for humanoids, like Nvidia's Project GR00T, requires data throughput at a scale that pushes current infrastructure to its limits. This investment builds the high-speed, low-latency nervous system for the "AI factories" that will birth the next generation of embodied AI. The partnerships are highly specific: Lumentum is a leader in CPO-optimized laser modules, while Coherent provides deep expertise in silicon photonics and the transceivers that convert electrical signals to light. By securing multibillion-dollar purchase commitments and future capacity rights from both, Nvidia is locking down the supply chain for these essential, specialized components. This move positions optical components as a supply chain chokepoint that Nvidia now has preferential access to. As hyperscalers and competitors race to build their own AI infrastructure, they may face constraints on the very parts Nvidia is now funding, creating a significant competitive advantage in the buildout of next-generation systems.