3 weeks ago
70
Everyone’s talking about how much electricity AI data centers consume. Fewer people are talking about what those data centers are physically made of. HIVE Digital Technologies wants to change that conversation.
During TheStreet Roundtable, HIVE Executive Chairman Frank Holmes laid out a problem that should keep AI infrastructure planners up at night: a single gigawatt-scale data center consumes roughly 50,000 tons of copper. And there isn’t nearly enough of the stuff to go around.
The metal bottleneck behind the AI boom
The context here is Nvidia’s memorandum of understanding, signed on August 10, 2026, to mobilize over $500 billion for AI infrastructure projects alongside six major asset managers. That capital is earmarked for data centers, GPU clusters, and the physical backbone of the AI revolution.
Capital, as Holmes pointed out, is the easy part. The hard part is the raw materials.
Copper prices have climbed to $6.67 per pound by late August 2026, a 16% increase since the start of the year. That trajectory isn’t a blip. It’s the market pricing in a structural imbalance between how much copper the world produces and how much the AI buildout demands.
Analysts from Wood Mackenzie and the International Energy Agency have both flagged anticipated deficits in refined copper supply, driven in large part by surging data center construction. The problem compounds: mining new copper deposits takes years of permitting and development, while AI infrastructure timelines operate on quarters, not decades.
HIVE’s pivot from Bitcoin to AI
HIVE Digital Technologies has been quietly repositioning itself from a Bitcoin mining operation into an AI and high-performance computing company. The transformation has been anything but subtle in terms of deal flow.
In June 2026, HIVE signed a $220 million engagement for AI services. That was followed by a five-year, $350 million GPU cloud agreement announced on August 17, 2026.
The crown jewel of HIVE’s infrastructure ambitions is a planned 320 MW Gigafactory in the Greater Toronto Area, with an estimated capital expenditure of CAD 3.5 billion and a target launch in the second half of 2027. That facility is one piece of a broader plan to reach approximately 860 MW of global capacity.
What this means for the copper market and AI economics
The math gets uncomfortable quickly. If a single gigawatt-scale facility needs 50,000 tons of copper, and Nvidia’s consortium is marshaling $500 billion to build out AI infrastructure at unprecedented scale, the cumulative copper demand from these projects alone could represent a meaningful percentage of global annual production.
There’s also a geopolitical dimension. Copper supply chains are concentrated in a handful of countries, primarily Chile, Peru, and the Democratic Republic of Congo. Any disruption in those regions, whether from labor disputes, regulatory changes, or political instability, would amplify an already tight market.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·