Copper settled the week of May 22 near $6.40 a pound, holding most of a year-to-date gain that has the AI buildout’s single largest physical input still climbing while every other indicator in the materials stack lags. The US Federal Permitting Improvement Steering Council moved in April 2026 to bring both AI data centers and an Alaskan copper project under FAST-41 in the same quarter. Beijing’s top smelters spent the same window cutting output at a treatment and refining benchmark of zero. The recent Trump and Xi leaders summit, in the White House recap of May 17, produced no removal of China’s rare earth or magnet controls. The pieces of the post-2024 critical minerals architecture are now visible. Copper is the load-bearing column.

What’s happening

• US copper futures held near $6.40 per pound through late May 2026, with the market posting a second consecutive monthly gain tied to AI data center buildout and clean energy demand (IndexBox, May 2026).
• Hyperscale AI campuses are now sized in 50 to 150 MW power blocks. At an industry intensity of 27 to 33 tonnes of copper per MW of installed capacity, a single 100 MW site absorbs several thousand tonnes of copper before the upstream grid reinforcements are counted (Investing.com analysis, May 2026; Copper.org).
• The Federal Permitting Improvement Steering Council granted FAST-41 coverage to QTS Richmond Technology Park Data Center 5 in April 2026, the first data center project ever to enter the program. The QTS campus is seven buildings, roughly 3.2 million square feet, and supports AI workloads. QTS told federal officials it was simultaneously building hyperscale deployments at nearly 4 GW (Data Center Knowledge, May 20, 2026).
• FPISC separately granted FAST-41 coverage to Alaska’s Arctic copper and critical minerals project, run by Trilogy Metals. Trilogy CEO Tony Giardini cited US copper import dependence as the strategic gap the listing was meant to address (Data Center Knowledge, May 20, 2026).
• The annual TC/RC benchmark between Antofagasta and the major Chinese smelters settled at USD 0 per tonne in January 2026, the lowest level ever agreed. China’s top smelters agreed to cut production by more than 10 percent in 2026, and Beijing halted around 2 million tonnes of planned new smelter capacity (Fastmarkets, 2026).

Brazil angle

Vale is having the run no Western copper major is having. Q1 2026 copper production hit 102.3 kt, up 13 percent year on year, with record output at both Salobo and Sossego (Vale 6-K, April 28, 2026). The company has stated a path to roughly 500,000 tonnes of copper by 2028 through brownfield improvements at Salobo and Sossego, with the new Bacaba operation replacing Sossego’s depleting open pit and extending the asset by at least eight years (Mining.com).

The near term constraint is operational, not structural. Sossego’s SAG mill is scheduled for a 110-day planned shutdown in the second half of 2026, which will pull Q3 and Q4 numbers down even as the trajectory improves. The Brazilian copper read this quarter is not about new mines. Carajás-region brownfield expansion at Vale is doing the heavy lifting, and Brazilian copper is moving the SDX needle in a window where Pensana, Pilbara and Sigma Lithium are flat or down. Brazilian copper’s permit to production timeline is structurally shorter than the US average. Worth saying out loud.

US angle

The Trump administration is running two parallel plays. Project Vault, announced February 2, 2026, with $10 billion of Export-Import Bank loan capacity and roughly $2 billion of private capital, is the strategic stockpile leg (Bloomberg, CNBC, February 2026). FAST-41 expansion is the permitting leg. Both presume the structural problem is American, not Chinese: the US does not lack copper resources, it lacks the timelines and the equipment supply to bring them online fast enough to match data center load growth.

Jigar Shah, former director of the Department of Energy’s Loan Programs Office, told Data Center Knowledge in May 2026 that transformers are running 18 to 24 month lead times, wires and cables have inflated 152 percent since 2019, switchgear is up 77 percent, and “nearly half of US data centers planned for 2026 are facing delays or cancellations.” NERC has projected that summer peak demand could rise by 224 GW over the next decade, driven largely by data center growth. Wood Mackenzie has estimated US generator step-up transformer demand has climbed 274 percent since 2019, with power transformer demand up 116 percent.

Read FAST-41’s first data center grant (QTS Richmond) and its Alaska copper grant in the same breath. Washington is now classifying copper supply as inseparable from AI data center supply. That is a useful tell of what the federal government thinks the binding constraint actually is.

China angle

China refines roughly 50 percent of global copper. Its share has grown from around 15 percent in 2005 to half of global refined supply in 2025 (Andaman Partners, IEA). The zero TC/RC settlement for 2026 is not a Western shock but a self-inflicted Chinese problem: domestic smelter capacity outran concentrate growth, the benchmark collapsed, and Beijing is now constraining its own supply chain to defend smelter margins. Top smelters cutting more than 10 percent and 2 million tonnes of new capacity halted is the kind of move that, if sustained, modestly tightens global refined copper into 2027.

The summit framing matters. The White House’s May 17, 2026 recap of the Trump and Xi leaders summit acknowledged that China would address US concerns about rare earth shortages, but did not call for removal of the magnet, gallium, germanium or graphite controls in force (KFGO, US News, May 17, 2026). On copper itself, no announcement. Beijing is running the smelting side and the rare earth side as separate negotiating tracks. Reading them together: China kept the architecture. The US got a small win on rare earths and is left to build the rest.

What it means

The TAI-M sub-index closed May 22 at 101.8, up 6.4 percent year to date (Tantalum Strategy, indexes.json, May 22, 2026). Copper is doing most of the heavy lifting in that move. TAI-P, more volatile and now at 104.2, has been the headline number through Q1 on uranium and nuclear utility equity, but the materials side is now confirming the power base. That has not been true for most of the post-base-date period.

For investors: copper exposure is the cleanest single read on the AI buildout right now. For policymakers: FAST-41 expansion is a real signal worth tracking quarter to quarter, not a banner. For Brazil: the Salobo run validates the Brazilian copper case at the moment global supply tightens, and the Carajás brownfield model is structurally better positioned than the US greenfield permitting cycle.

What to watch

• Sossego SAG mill restart in late 2026 after the planned 110-day shutdown. The inventory drawdown through Q3 will shape Vale’s full year guidance and the Brazilian copper read into 2027.
• Next FAST-41 grants in June and July 2026. The cadence at which the Permitting Council adds copper, transmission and substation projects tells you whether the program is binding or symbolic.
• Annual TC/RC negotiations in late 2026. A second zero or sub-$5 benchmark would lock in additional Chinese capacity cuts and tighten LME inventories into 2027.

A note on the indexes

TAI and SOV50 readings cited above are Tantalum Strategy’s proprietary thematic indexes, live since January 2026. They are not benchmark grade. Constituent weights and the editorial overlay are documented in scope on /indexes/methodology; the kernel is desk editorial judgment. Naive replication from public commodity feeds alone will diverge.