Jensen Huang Confirms the “Copper & Fiber” Coexistence Era

China4w ago update xebapp
15,776 0 0

Jensen Huang has explicitly stated that the future of AI infrastructure relies on a “Dual-Era” where both copper cables and optical fiber coexist. This approach settles the market debate over which technology will dominate by assigning them specific roles based on distance and power efficiency.

1. The Strategy: “Copper for Scale-up, Fiber for Scale-out”

Huang clarified that “optics replacing copper” is a misunderstanding. Instead, NVIDIA’s architecture uses them in tandem:

  • Copper (Scale-up): Remains the primary choice inside the server rack (short-distance, <3-5 meters) because it is inexpensive, durable, and has zero power consumption.
  • Fiber Optic (Scale-out): Dominates inter-rack and long-distance connections. As AI clusters expand to tens of thousands of GPUs, fiber is the only way to move data across vast data centers without massive signal loss. 

2. Validation of Hock Tan’s View

Broadcom CEO Hock Tan previously argued that copper would remain the standard for intra-rack connectivity for as long as possible. NVIDIA’s recent GTC 2026 comments confirm this, specifically supporting companies like Astera Labs (ALAB) and Credo (CRDO). Their technologies, such as Active Electrical Cables (AEC) and Retimers, are critical for extending copper’s reach as lane speeds hit 200G and 400G. 

3. Increased Global Demand for Fiber Optic

The demand for optical fiber is not just increasing; it is becoming a strategic bottleneck for AI scaling. 

  • Strategic Investment: NVIDIA recently announced a massive $3.2 billion partnership with Corning (GLW) to build three new factories in the U.S., aiming to boost fiber production by over 50%.
  • Volume Explosion: A single AI “factory” (large-scale data center) requires tens of thousands of miles of high-end optical fiber to connect the entire cluster.
  • Technological Shift: Beginning with the Feynman platform (expected 2028), NVIDIA plans to use Co-Packaged Optics (CPO), which places “light engines” directly next to the chips, further accelerating the transition from electrical to optical transmission at the chip level. 

Comparison of Roles

Feature Copper (AEC/DAC)Fiber Optic (Modules/CPO)
Typical Range0–3 Meters3 Meters to 2 Kilometers
Best ForIntra-rack (Within one cabinet)Inter-rack (Between cabinets)
PowerVery Low to ZeroHigh (unless using CPO)
Growth DriversNVL72 / NVL144 scale-upLarge-scale AI clusters / CPO

For long-term investors, the “direction” is clear: 

AI is driving a massive infrastructure build-out that requires more capacity for both materials. Copper provides the localized muscle, while fiber provides the global nervous system for AI. 

Impact of US-China Trade and Supply Chain Restructuring

While trade demand is surging, the supply chain is showing a clear trend toward regionalization:

  • U.S. Localization of Production: NVIDIA is actively funding Corning to build three advanced manufacturing facilities in North Carolina and Texas. This strategic move aims to boost domestic optical fiber capacity by 50%, specifically designed to mitigate risks of potential supply chain disruptions.
  • China’s Supply Chain Dominance: China currently accounts for 56.3% of global optical fiber production. With its formidable advantages in cost efficiency and massive production scale, Chinese suppliers are expected to remain major beneficiaries of the global AI infrastructure wave.

Conclusion:


Optical fiber is no longer just a basic construction material; it has become a mission-critical component that determines the scalability of AI systems. As NVIDIA pushes forward with next-generation architectures like Vera Rubin and Feynman, the global optical fiber trade is set to enter an unprecedented “Explosive Year.”

© Copyright Notice

Related Posts

no comments

none
no comments...