Uranium and Copper: a Squeeze Is Likely, a Supercycle Is Not

Copper trades near $14,540 per tonne on the London Metal Exchange (LME). Yet the loudest calls for a uranium and copper supercycle, driven by artificial intelligence (AI), pitch the metal at $30,000-$40,000 per tonne within five years, a figure no mainstream forecast supports.

That gap between narrative and evidence matters more than usual right now. AI data centres are drawing heavily on power grids, reactor construction is concentrated in Asia, and both uranium and copper show signs of structural tightness.

The harder question is which parts of the story hold up. Some of the thesis is backed by data. Some is overstated. And the real bottleneck may not sit in a mine at all.

Here is a weighted read on what the evidence supports for uranium, what it supports for copper, and which risks decide whether either earns the supercycle label.

Where is the real AI power bottleneck: the mine, the reactor or the grid?

The intuitive version runs like this: AI needs electricity, electricity needs fuel and wiring, so buy the miners. It is a tidy chain. The problem is that the first link to break is not fuel or metal.

It is delivery.

AI data centre power demand is straining grids that were designed for gradual residential and commercial growth, which is why connection access, not fuel supply, becomes the first constraint on how fast the load can materialise.

In the UK, a prominent British advanced computing facility has been delayed awaiting a grid connection. Broader UK connection delays of 8-10+ years have been reported, with some offers dated 2037 or later.

Grid queue snapshot Ofgem reportedly shows a large queue of data centres awaiting connection, though the figures have not been independently confirmed.

Equipment tells the same story. Wood Mackenzie data reportedly show US generation step-up transformer demand (the units that raise a power plant’s output voltage for transmission) up 274% between 2019 and 2025, with Q2 2025 delivery times of 143 weeks for those units and 128 weeks for power transformers. These figures have not been independently confirmed. Gas turbines from GE Vernova, Siemens Energy and Mitsubishi Power are also reported to be sold out into the late 2020s, though Wood Mackenzie’s Ben Boucher expects lead times to ease.

Why industrialists are buying their suppliers

When the queue is the constraint, owning the supply chain becomes a rational move. Companies are integrating vertically for six main reasons:

• Avoiding grid-connection queues
• Securing firm power through nuclear power purchase agreements
• Capturing scarce equipment such as turbines
• Getting around transformer and switchgear shortages
• Reducing reliance on central transmission through behind-the-meter systems
• Locking up fuel through long-term uranium contracts

In Britain, industrial buyers completed multiple acquisitions of power-equipment makers across 2025-2026. Claims that every British maker in one component niche was bought within six weeks, with order books full for five years, remain unconfirmed.

What this tells you: if grid access gates AI demand, the fuel and metal demand it generates may arrive later than headlines suggest. Niche equipment makers could capture value first.

How nuclear buildout and tech deals translate into fuel demand

If electricity delivery is the near-term choke, uranium demand depends on a longer chain. Understanding it shows why AI alone cannot tighten the market.

The sequence works in three steps. A reactor is built over many years. Its operator, usually a utility, then signs long-term fuel contracts. Those contracts pull on mine output, conversion (turning mined uranium into a gas), enrichment (raising the share of fissile uranium) and secondary inventories such as stockpiles.

Each stage limits how fast supply can respond. Mining is also concentrated: Kazakhstan, Canada and Namibia produced about 75% of 2024 mine output, with Kazakhstan alone at 38%.

Tech companies have entered this chain directly.

Two strategies sit inside that table. Restarts lock up existing plants with known fuel needs. Pre-orders for unbuilt advanced reactors carry licensing, fuel and construction risk. For your purposes, restarts are the credible near-term signal; reactor pledges are options, not orders.

Announcements are cheap and construction is not, which is why many new nuclear programmes stall before a reactor is ever built, and why pledges should be weighed as options rather than as confirmed fuel demand.

Asia builds, the West announces

About 80 reactors (79.4 GWe) are under construction globally, roughly 55 of them in Asia. The World Nuclear Association counts 39 in China, while another tally puts it at 36; the difference comes from inclusion criteria. China reportedly accounted for nine of eleven construction starts in 2025 (unverified).

The US is reported to have zero under construction, against policy targets of 10 large reactors underway by 2030. Those are objectives, not plants. A widely circulated comparison of 27 US projects versus 57 in Asia could not be confirmed.

What do the supply-demand numbers say for uranium and copper?

Start with the stronger case. Spot uranium averaged $89.63/lb at September 2026 month-end, while the term price sat at $96.50/lb. Term trading above spot suggests utilities are paying up to secure future supply rather than buying for today.

Forecasters see deficits widening. Sprott projects a cumulative shortfall of about 1.25 billion lb through 2045, and Pitt Street Research sees an annual gap of roughly 15M lb by 2030, rising to about 47M lb by 2035. Kazatomprom has reportedly cut nominal 2026 capacity to about 77M lb, and Bank of America and UBS forecasts of $88-$130/lb remain unverified.

Copper is where the evidence thins. Goldman Sachs expects a 2026 average of $12,650/t with a 490,000-tonne surplus, below today’s price. BMI’s $13,500/t and a Reuters poll average of $13,169/t (both unverified) land in the same range.

A refined copper surplus of the size Goldman expects would sit uneasily beside supercycle claims, since inventories build and prices soften when supply outpaces consumption even as long-term electrification demand remains intact.

Forecast reality check Goldman Sachs projects a 490,000-tonne copper surplus in 2026. The $30,000-$40,000/t supercycle target sits more than double the current price and has no mainstream forecast behind it.

The long-term gap is real. The International Energy Agency (IEA) sees a supply shortfall of about 25% by 2035, narrowed from 30%, while BHP expects demand above 50 Mt a year by 2050. New mines take around 17-18 years to develop. Yet S&P Global puts data centres at only about 6% of copper demand by 2040, as total demand climbs from 28 Mt to 42 Mt.

The read you should take: uranium has the firmer structural case, and copper’s gap is plausible but not a $30,000 story. “Supercycle” fits only if you accept long-dated assumptions.

Past performance does not guarantee future results. Financial projections are subject to market conditions and various risk factors.

What could break the thesis, and how should investors weigh it?

The bullish case has clear weak points, and most of them are about timing.

For copper, Goldman’s surplus is not alone; Fastmarkets reportedly expects a surplus just under 200,000 t in 2026 (unverified). Aluminium is substituting in some cable uses, and Chinese demand, tariffs and recession risk can all overpower long-term stories. Disruption cuts the other way: Reuters flagged roughly 125,000 t at risk in the Democratic Republic of Congo and about 200,000 t in Chile.

Aluminium substitution is one form of demand destruction at higher prices, and it shows how quickly buyers can trim copper intensity when costs rise, which weakens the case for sustained price spikes.

Uranium faces its own tests. As of September 2026, no mothballed US plant had resumed operation. Mine restarts, expansions and inventories could ease supply, while sanctions on Russian fuel tighten Western markets but raise costs and slow projects.

Equipment bottlenecks may also loosen as Siemens Energy, Mitsubishi and GE Vernova expand capacity. Add an AI spending slowdown, policy reversals, cost inflation and volatility, and the main threat becomes clear: being right too early.

That puts the weight on how you gain exposure. Producers, developers, equipment makers and term-contracted fuel carry very different timing risk.

Signposts worth tracking

1. Reactor restarts: Three Mile Island’s progress shows whether restarts deliver on schedule.
2. Term contract volumes: rising utility contracting confirms that demand is converting into orders.
3. Copper inventories: stock levels test whether the near-term surplus is materialising.
4. Chinese demand: the largest swing factor for copper prices.
5. Grid connection timelines: shorter queues would pull data-centre demand forward.
6. AI capex: hyperscaler spending plans set the pace for the whole chain.

Weighing a structural squeeze against a full supercycle

The evidence supports a squeeze more readily than a supercycle. Uranium tightness looks credible, backed by deficit forecasts, concentrated supply and a term price above spot. Copper’s long-term gap is plausible but tempered by an expected surplus in 2026 and a narrowing IEA shortfall.

AI accelerates demand without being its sole driver. In the near term, transformers, turbines and grid access are the chokepoints, which may push commodity demand later than headlines imply.

Your decision point is simple: anchor on verified supply-demand data, not headline price targets. Data is current as of 13 October 2026.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions.