BofA: AI Capex Raised to $1.7 Trillion — Copper Foil, Fiberglass, Optical Fiber, Magnetic Materials, and Tungsten to Benefit
Claire Weston
Bank of America raised its 2030 global AI capex forecast from $1.4 trillion to over $1.7 trillion, arguing that AI competition is shifting from compute power to power infrastructure and upstream materials — copper, copper foil, fiberglass, optical fiber, magnetic materials, and tungsten stand to benefit as supply constraints persist.
Where does $1.7 trillion go?
BofA cites four drivers behind the upgrade: accelerating compute migration, intensifying cloud-platform competition, sovereign AI infrastructure buildouts, and faster enterprise AI adoption.
China's AI capex alone is projected to rise from roughly $140 billion in 2026 to $327 billion by 2030 — a 24% CAGR, accounting for about 20% of global spending.
This means → China is not a sideshow in this investment cycle. A fifth of the total spend lands there, and a large share of incremental materials demand originates domestically.
Will fiber optics really replace copper?
BofA directly pushes back on the "fiber-in, copper-out" narrative: copper remains irreplaceable for internal power delivery and short-range signal transmission inside data centers, and AI-driven power infrastructure investment indirectly lifts copper demand too.
China's data-center-driven copper demand is forecast to rise from 341 kt in 2025 to 1,190 kt by 2030 — a 28% CAGR — pushing its share of total Chinese copper demand from 2.1% to 6.4%.
On the supply side, BofA expects global copper supply to fall 1% in 2026, with deficits of 491 kt in 2026 and 754 kt in 2027. In plain terms = demand is accelerating while supply is shrinking — the gap widens each year.
Top picks: Zijin Mining (first choice), with buy ratings on Jiangxi Copper and CMOC.
Copper foil and fiberglass — a split between high-end shortage and low-end glut?
Copper foil demand is migrating from battery applications to high-end PCBs — the circuit boards that connect chips inside electronic devices. AI servers demand more PCB layers and higher-frequency signal performance, driving up demand for premium copper foil, while low-end capacity remains structurally oversupplied.
This means → not every copper foil maker benefits. The key is whether a producer's output meets AI-server specifications.
Switching from lithium-battery copper foil to PCB-grade foil requires additional capex and process upgrades; customer certification takes about a year or longer, so supply responds slowly.
In electronic fiberglass, high-speed PCBs and packaging substrates need substrates with low dielectric constant, low dielectric loss, and low thermal expansion — put simply = signals travel faster, lose less energy, and the board doesn't warp when heated. Qualified suppliers are scarce. BofA rates Sinoma Science & Technology a buy.
Optical fiber, magnets, and tungsten — how is the demand structure shifting?
Optical fiber demand is pivoting from traditional telecom to AI data centers. Overall supply is adequate, but preform capacity — the raw glass rods from which fiber is drawn — constrains high-spec fiber availability. BofA rates ZTT (Zhongtian Technology) a buy.
High-performance NdFeB magnets — powerful permanent magnets used in motors and cooling systems — are being pulled by AI data-center cooling and humanoid robotics, while rare-earth export controls limit supply. BofA rates JL MAG Rare-Earth a buy.
Tungsten demand is shifting from traditional cutting tools to PCB micro-drill bits for AI hardware. China's resource controls support pricing. BofA also rates CGN Mining a buy, on the logic that nuclear power is a key enabler for AI electricity demand, and uranium faces a structural supply deficit.
What is BofA's core thesis?
In one line: AI capex competition is increasingly becoming a competition over power and materials infrastructure.
This means → whoever can resolve upstream supply bottlenecks during the demand acceleration phase will capture outsized profits along this beneficiary chain.
This reflects a broader shift: AI investment opportunities are extending from "who makes the chips" to "who supplies the raw materials" — and the very barriers on the materials side — long certification cycles, strict resource controls, slow capacity conversion — are precisely what keep margins elevated.
Content is for reference only, not financial advice.