en.Wedoany.com Reported - NVIDIA CEO Jensen Huang stated unequivocally in an interview on Wednesday, May 7, 2026, Eastern Time, that the next generation of artificial intelligence infrastructure will require a massive amount of optical connections, as computing demands are rapidly increasing and copper cables can no longer meet the requirements. He stated bluntly: "We are going to scale optics in a way that's never been done before, and frankly, no optics company has ever seen this kind of scale."
Huang made these remarks while commenting on the recent strategic partnership between NVIDIA and Corning Incorporated. He used this collaboration as an example to illustrate the inevitable logic of scaling optical technology: the interconnects between chips within AI data centers are rapidly transitioning from the traditional copper cable era to the optical interconnect era. Huang pointed out: "We are experiencing the largest infrastructure build-out in human history. AI is going to become a fundamental infrastructure globally, and the United States is certainly no exception." He added that the construction of AI infrastructure creates a unique opportunity to reinvest in American manufacturing and supply chains, a process made possible after decades of industrial offshoring.
From the perspective of the real urgency of technological substitution, Huang's judgment directly addresses the core bottleneck of current AI cluster interconnects. As the bandwidth demand for a single rack in large-scale GPU clusters rapidly surges towards hundreds of terabytes, traditional copper cables are approaching their physical limits in three dimensions: transmission distance, bandwidth density, and energy efficiency. Taking NVIDIA's new generation rack-level AI systems as an example, even in the copper cable era, they already incorporated approximately 5,000 high-speed copper cables. Such a scale of copper cabling faces severe signal attenuation and heat accumulation under the deluge of massive parallel data streams. In contrast, optical fiber transmits data via photons. Corning CEO Wendell Weeks has previously stated clearly that transmitting data in the form of photons consumes 5 to 20 times less energy than transmitting it in the form of electrons.
Huang's emphasis on the importance of optical connections at this time forms a logical echo with NVIDIA's previously released optical interconnect roadmap. At the GTC conference in March 2026, NVIDIA formally proposed a three-pronged parallel approach involving copper Scale-up, optical Scale-up, and optical Scale-out, and for the first time simultaneously advanced copper interconnect and co-packaged optics (CPO) interconnect solutions in architectures featuring NVLink 144 and above. Huang stated bluntly at the time that CPO is an essential core technology for the expansion and development of the AI industry. When evaluating the partnership with Corning this time, he further quantified this judgment for external audiences—the opening of supply bottlenecks is itself an inevitable step in the transmission of AI computing power from the chip layer to the infrastructure layer.
The data from the Corning partnership provides quantitative support for this judgment. On May 6, 2026, NVIDIA and Corning jointly announced plans to build three new advanced manufacturing plants in North Carolina and Texas. Corning will increase its U.S. production capacity for optical connectivity products tenfold and expand its U.S. optical fiber capacity by over 50%. The project is expected to create more than 3,000 high-paying jobs, and NVIDIA also obtained warrants to acquire Corning common stock through a $500 million equity investment. IT Home, citing industry analysis, pointed out that the core technical path of the collaboration points to CPO—replacing traditional copper cables with Corning's optical glass fiber in AI rack-level systems.
In the interview, Huang also particularly emphasized that the benefits of the current wave of AI investment extend far beyond technology companies. He noted the rising demand within the AI industry for electricians, construction workers, chip manufacturing employees, and data center infrastructure specialists. "The shortages we face and the demand for all these skilled tradespeople is really, really high," Huang said. He cited this as strong evidence that the AI infrastructure build-out process has already rippled through the entire economy. From the perspective of industry chain transmission, the NVIDIA-Corning partnership will not only accelerate the transition of CPO technology from laboratory validation to mass production but will also drive the synergistic expansion of the entire optical interconnect industry chain, including upstream optical materials, optical module packaging and testing, and downstream data center fiber cabling. As copper cables gradually give way to optical fiber, the transmission architecture of AI data centers is undergoing a paradigm shift from electrification to photonics, and the speed of this shift will directly determine whether the next generation of trillion-parameter model training clusters can be delivered on schedule under the dual constraints of power consumption and bandwidth.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
