Price tells you what a crowd believed this morning. Capital expenditure tells you what management committed to years ago and cannot easily undo — so it is the slower signal, and the harder one to fake.
Capital expenditure as filed with the SEC, summed across the several XBRL tags companies use for it — querying only the most common tag under-reports energy and utilities specifically, which are the sectors worth comparing. Compared with the same quarter a year earlier, because capex is seasonal. “Thin” marks a sector where fewer than three in five constituents had filed when this was built.
What AI actually depends on, one layer at a time. Click a branch to go further down the chain — data centers need chips, chips need packaging, packaging needs a film that one company in Japan makes. Click any company to see its role and whether you can buy it.
Each step depends on the steps to its right. Start at AI (who builds and sells it), move to the data centers it runs in, then everything those data centers need, down to minerals and gases. PRIVATE means there is no listed share to buy.
The chips get the headlines. The scarce thing is a site with a grid connection — which is why the bitcoin miners with power contracts are being bought for their substations, not their hash rate.
Ranked on this library’s own 36-month upside multiplied by the scorecard, so a high target only ranks if the business behind it scores well. Upside is capped at 300% for the ranking, otherwise one outlier would order the whole list. These are the companies the research covers that serve this theme — not every company that does.
Steady. Loosening in REFINED metal, tightening in the ore behind it. ICSG counts finished copper leaving smelters and sees a surplus; the banks forecast the concentrate feeding those smelters and see a deficit. Chinese smelter capacity has grown faster than mine supply, which produces both at once. Neither figure is wrong and they are not about the same thing — see the forecast range above before drawing a conclusion.
Read this before you trust the number. The previous ICSG forecast had 2026 at a 150 kt DEFICIT. The April 2026 release flipped it to a surplus. A number that changed direction in six months is a forecast, not a fact.
The refined balance is what gets consumed; the concentrate balance is what constrains it. A copper thesis needs to say which one it is about.
ICSG forecast · 25 Apr 2026
Tightening. A sixth consecutive deficit year either way. The disagreement is about how large, not about direction.
Read this before you trust the number. Coverage of the same World Silver Survey quotes both a record 215 Moz deficit and one near 46 Moz for 2026. They cannot both be the headline number, so both are shown. A deficit of 215 Moz is roughly a fifth of annual supply; 46 Moz is a rounding error by comparison, and the investment case is completely different depending on which is right.
Silver is the one metal where industrial and investment demand compete for the same ounces, so a structural deficit reaches the price faster than in most commodities.
Silver Institute / Metals Focus · 16 Apr 2026
Steady. Still buying, but slower than last year — 130 t by July against 160 t over the same months of 2025. The official-sector bid is intact rather than accelerating.
These are REPORTED purchases. Some buying is never reported, so the true figure is higher by an unknown amount — which is exactly why nobody can give you a precise number here.
Official-sector buying has been the marginal bid under the gold price, so its pace matters more than any single month's move.
World Gold Council · Sept 2026
Tightening. Mines supplied 87% of what reactors needed. The rest came from stockpiles, reprocessing and down-blended weapons material — sources that deplete. That gap is the whole uranium thesis, and it is visible without paying for a forecast.
The 1–3 year FORWARD balance is genuinely paywalled — UxC charge around $8,500 a year and the WNA around £1,500, with no free substitute worth publishing. So this shows the last full year of actuals, which are free, instead of a forecast nobody here can verify.
A structural gap filled by finite secondary supply is a different thing from a temporary shortfall.
World Nuclear Association figures · 2024 full year
Live data by TradingView. Block size = market cap, colour = move over the chosen timeframe. The % on each button is that market’s main index in local currency (for crypto: the CMC Crypto 200 index). Scroll to zoom into a sector.
▸ Information Technology XLK · 74 stocks | +30.6% | +813% | +24.8% | 61% | 31% |
▸ Industrials XLI · 83 stocks | +10.6% | +251% | +13.4% | 30% | 2% |
▸ Financials XLF · 76 stocks | +7.2% | +236% | +12.9% | 17% | 9% |
▸ Consumer Discretionary XLY · 47 stocks | −5.2% | +220% | +12.4% | 30% | 11% |
▸ Communication Services XLC · 23 stocks | +1.6% | +158%* | +12.2% | 48% | 17% |
▸ Health Care XLV · 60 stocks | +19.1% | +197% | +11.5% | 13% | 23% |
▸ Energy XLE · 21 stocks | +45.3% | +172% | +10.6% | 33% | 10% |
▸ Materials XLB · 25 stocks | +19.4% | +162% | +10.2% | 24% | 8% |
▸ Utilities XLU · 31 stocks | −6.3% | +125% | +8.5% | 10% | 6% |
▸ Consumer Staples XLP · 33 stocks | +11.1% | +104% | +7.4% | 12% | 9% |
▸ Real Estate XLRE · 30 stocks | +7.5% | +94% | +6.9% | 7% | 7% |
Insider data unavailable this run: OpenInsider returned no rows