AI data centers and metals: a turning point in demand or inflated hype
Reuters columnist Andy Home weighs how real AI data center demand for metals is, and where grid links, policy and construction delays stand in the way.
- published on Newsmetry
- original source published
the same feed is available to bots and dashboards through the api. connect the api →
what happened
Reuters columnist Andy Home asks whether demand for metals from AI data centers will prove a true game-changer or mostly hype. According to a World Economic Forum white paper, data centers use 60 to 75 metric tons of metals per megawatt. Gallium, germanium and tin stand out in market impact, with data centers projected to account for 6.0%, 2.4% and 1.2% of global annual usage respectively in 2035. Copper ranks fourth, and Wood Mackenzie expects copper and aluminium usage to peak in the early 2030s.
Wood Mackenzie estimates that once grid connections are counted, demand for copper and aluminium runs three to four times above projections based solely on the buildings. Yet a study by the Oxford Smith School and broker Marex points to delays: the average US wait for a grid connection was four years between 2018 and 2023, while a data center can be built in 18 to 24 months.
why it matters
For markets, metals demand depends not only on AI investment but also on power infrastructure and government policy. China has restricted exports of gallium and germanium, which is why other countries are looking for alternative sources. Copper and tin are sensitive to sentiment in the technology sector, which recently suffered a sell-off.
market impact
Copper and tin: AI infrastructure demand supports metals, but construction timelines and policy make the outlook uneven
- Copper mixed — grid and power demand supports it, while efficiency gains and connection delays weigh
- Tin mixed — circuit board demand supports it, while the tech sell-off pressures prices
tags
instruments
on polymarket
sources
- Metals in the Age of AI — Game-changer or hyperscale hype?: Andy Home — Reuters, 14 November 2026 at 07:00 UTC
prepared by the newsmetry newsroom with ai assistance from the sources listed below. not investment advice.
alerts on your topics, the morning digest and the api — in your account.
get stories in real time ↗