AI's Copper Hunger: Data Centers Driving Up Cable Prices

As 2026 begins, copper prices remain at elevated levels. On February 5, COMEX copper settled around $5.82 per pound, reflecting a slight daily pullback but still significantly higher—up over 30% in some comparisons from early 2025 levels—amid sustained global demand.

You might wonder what this has to do with everyday life. Beyond pricier power cables or EV charging cords, the real driver this time is the AI boom – massive data centers devouring copper at an unprecedented rate, indirectly pushing up costs all the way to the Ethernet cable behind your home router.

Why AI Data Centers Are Copper's New Super-Consumer

Training and running large AI models require enormous GPU clusters. A single hyperscale data center can house tens or even hundreds of thousands of graphics cards. The short-distance, high-speed connections inside and between nearby racks (typically under 5-15 meters) overwhelmingly rely on high-speed copper cables, specifically Direct Attach Copper (DAC) or Twinax cables. For instance, NVIDIA's GB200 NVL72 rack system – a liquid-cooled AI supercomputer with 72 Blackwell GPUs – uses over 5,000 high-performance NVLink copper cables, totaling more than 2 miles (3.2 km) in length.

A quick clarification: these aren't the familiar Cat8 Ethernet cables you buy for home use:

  • Speed — DACs support 400G, 800G; Cat8 tops out at a stable 40G.
  • Design & Use Case — DACs are thick, fixed-connector Twinax cables built for dense, intra-rack GPU-to-GPU links; Cat8 is slim, flexible twisted-pair with RJ45 plugs for office or home runs up to 30 meters.
  • Advantages — DACs have near-zero extra power draw, lower latency, and far better cost-efficiency – critical for AI clusters that are hypersensitive to both energy bills and budgets.

Because of these benefits, data centers heavily favor copper DACs for short-reach interconnects. Industry forecasts indicate that AI-driven data centers could consume around half a million tons of copper annually by 2030, with demand ramping up significantly in the interim years. Other estimates suggest AI-powered facilities will average about 400,000 tonnes per year over the next decade, reflecting the massive scale of these interconnects.

Price Ripple Effects: From Hyperscale Racks to Your Living Room

The chain is straightforward on the enterprise side:

  • Rising copper prices → higher costs for DACs and other copper-intensive cables → increased build-out expenses for data centers → potential pass-through to cloud/AI service fees over time.

At the consumer level, the impact has been muted so far. Top Amazon Ethernet cable sellers like Jadaol have kept prices largely stable for nearly a year, often maintaining discounts, thanks to intense competition in the category. Copper price swings also take time to flow through to retail—factories and vendors typically lag in repricing due to careful cost planning. This means manufacturers are currently absorbing much of the increased raw-material burden.

By contrast, copper-heavy sectors like air conditioning have already announced industry-wide price increases to pass on costs.

For now, the extra expense of the AI copper boom is largely being shouldered upstream—but sustained pressure could eventually change that.

Outlook: A New Supercycle Just Beginning? 

In the near term, 2026 copper prices will swing with AI capex pace, Fed policy, and global recovery signals. Longer term, sustained AI and decarbonization growth could usher in a true copper supercycle, with total data center demand (heavily AI-influenced) potentially reaching 2.5 million tons by 2040. Efficiency gains in cable design and higher recycling rates may ease some pressure, but copper's role as an indispensable industrial metal is only growing stronger.

One humble Ethernet cable now ties together the grand narrative of the AI age. The next time copper makes headlines, remember – it's probably coiled up somewhere in your home too.

RELATED ARTICLES