Key Points:
- Nvidia has committed at least $6.5 billion over three months to silicon photonics companies to address critical AI data transfer bottlenecks.
- The company made separate $2 billion commitments to optical component leaders Coherent, Lumentum, and Marvell.
- By shifting from copper wires to light pulses, co-packaged optics can reduce interface power consumption by a massive 3.5x.
- The strategic investments triggered a major stock market rally, with optical suppliers’ share prices doubling in early 2026.
U.S. chipmaking giant Nvidia has launched a massive, coordinated investment drive into silicon photonics to solve the most critical bottleneck facing the artificial intelligence (AI) industry. According to data reported by CNBC on Friday, May 29, 2026, the company has committed at least $6.5 billion to various photonics-focused companies in just the past three months. The strategic push aims to replace traditional copper-based data transmission in AI factories with next-generation optical technology that uses light rather than electricity.
This aggressive shift marks a significant transition in how the industry scales AI data centers. In 2021, the primary constraint was raw compute power, while in 2023 it shifted to memory. Today in 2026, with hyperscaler capital expenditures running at hundreds of billions of dollars, the constraint has moved again to the “highway between the chips”—the interconnects. Traditional copper wires have hit a physical wall, as transferring massive amounts of data over copper consumes too much energy, generates excessive heat, and degrades signals over long distances.
The bulk of Nvidia’s $6.5 billion capital deployment has targeted established optical component and laser manufacturers. In early March, Nvidia announced separate, non-exclusive $2 billion strategic partnerships with both Lumentum and Coherent. These agreements are far more than passive financial bets; they include multi-billion-dollar long-term purchase commitments and capital to expand U.S.-based manufacturing. By securing guaranteed access to advanced laser and optical networking capacity, Nvidia is protecting its supply chains from future capacity shortages.
Later in March, Nvidia announced an additional $2 billion partnership with semiconductor designer Marvell Technology. This collaboration ties directly into Nvidia’s NVLink Fusion architecture, custom AI infrastructure development, and advanced silicon photonics. Marvell, which acquired photonics startup Celestial AI, is developing cutting-edge electro-optic polymers and silicon photonics. This deal ensures that Nvidia can integrate high-bandwidth, optical interconnects directly into its custom server racks, maintaining its commanding lead in high-performance networking.
To secure the physical “pipes” of this optical network, Nvidia also partnered with 175-year-old glassmaker Corning, investing $500 million in the company. Under this long-term agreement, Corning will expand its U.S. optical connectivity manufacturing capacity tenfold, increase its domestic fiber production by more than 50%, and build three new manufacturing plants in North Carolina and Texas, creating over 3,000 local jobs. Concurrently, Nvidia participated in optics startup Ayar Labs’ $500 million Series E funding round, which aims to scale up co-packaged optics (CPO) for high-performance computing.
This focus on co-packaged optics represents a critical technological milestone for the hardware industry. Traditionally, optical transceivers reside far from the main processor, requiring electrical signals to travel across a circuit board. CPO solves this by mounting the optical interconnects directly onto the same substrate as the GPU. This close packaging can cut interface power consumption from approximately 30W to as low as 9W, delivering a stunning 3.5x efficiency gain that can dramatically lower the multi-billion-dollar electricity bills of modern AI factories.
During Nvidia’s annual GTC conference, founder and Chief Executive Officer Jensen Huang outlined the strategic necessity of this transition. He stated that in the age of AI, software runs on intelligence with tokens generated in real time by AI factories for every interaction and every context. Huang explicitly warned that the world’s current silicon photonics manufacturing capacity falls far short of the available supply. By investing early and aggressively across the entire optics supply chain, Nvidia is ensuring it can scale its infrastructure without hitting a physical wall.
The massive investment drive has triggered a spectacular, sector-wide rally among optical and networking stocks on Wall Street. Since the beginning of the year, Coherent shares have surged 104%, Corning has jumped 108%, Lumentum has skyrocketed 134%, and Marvell has soared 141%. Investors are increasingly recognizing that the real value in the next phase of the AI boom lies in the physical connections linking thousands of GPUs. On Nasdaq, Nvidia’s shares traded at around $214.25, bringing its market value to approximately $5.23 trillion.
While some independent analysts and researchers note that the widespread commercial rollout of co-packaged optics faces technical and packaging hurdles, they expect large-scale adoption to begin by 2028. By securing early partnerships, capacity rights, and technological leadership, Nvidia has successfully built an unassailable moat around its hardware business. The move proves that the ultimate winner of the AI race will not just be the company that designs the fastest chip, but the one that masters transmitting data at the speed of light.
