Nvidia hardware momentum meets a new choke point: copper, not cash, HIVE Digital’s Frank Holmes says

Nvidia’s push to scale AI computing is running into a constraint that has little to do with chip financing and everything to do with building the systems that hold and power them. In remarks highlighted in a Yahoo Finance report, HIVE Digital Chairman Frank Holmes said that Wall Street can line up large amounts of money for data center buildouts, but the scarcity that matters is upstream, in the form of copper availability.

Holmes’ argument centers on how much copper is required to wire a major data center power footprint. He said that “one gigawatt of data center” would consume about 50,000 tons of copper, framing it as a structural supply issue rather than a solvable financial one. The implication is that even if capital is available, projects can be delayed or redesigned when key inputs are constrained.

The Yahoo Finance piece also referenced Holmes discussing financing platforms as “huge,” and claimed that Nvidia “just lined up $500 billion.” The report ties that figure to the broader push toward AI infrastructure capacity, but it stops short of detailing the specific financing structure, counterparties, or the exact components of that total. As presented, the point is less about the mechanism of financing and more about what still can prevent expansion: raw-material and infrastructure throughput.

Holmes’ comments place copper alongside power as an operational bottleneck for AI data centers. While the report does not provide a full procurement breakdown, it suggests that copper, like other grid-adjacent inputs, can become a gating item once dozens of gigawatts of new demand emerge faster than production and logistics can respond. In that context, the “scarce isn’t money, it’s copper” framing becomes a critique of how markets sometimes over-focus on funding availability while underweighting build-time realities.

For Nvidia, the immediate relevance is indirect but real. Nvidia is a primary supplier of accelerators and related software in AI training and inference, and those systems require rapidly expanding data center capacity. The companies that build those data centers, and the suppliers that deliver electrical infrastructure, sit downstream of chip demand. If copper supply and installation timelines tighten, it can affect scheduling, power-per-rack configurations, and the pace at which compute capacity becomes available for deployment.

Holmes also pointed to China in the report, indicating that the demand and supply story has geographic weight. The article does not elaborate on which part of the supply chain is under strain, what regional procurement risks apply, or whether copper constraints are binding at particular stages of construction. Still, the mention of China underscores that AI infrastructure scaling is global and that constraints in one major market can ripple through equipment orders, logistics, and construction schedules elsewhere.

What remains unclear from the Yahoo Finance coverage is how quickly copper-related limits can be eased. The report does not quantify current copper inventories, contract lead times, or the extent to which wiring can be optimized through alternative designs, higher efficiency cabling, or different data center architectures. It also does not specify whether the $500 billion figure reflects committed capital, funding pipeline estimates, or a mix of equity, debt, and supplier financing. Until those details are pinned down in filings, earnings materials, or named transaction announcements, the copper bottleneck should be treated as a thesis about constraints, not a precise forecast for Nvidia’s revenue cadence.