Data centres: Correlation
Prices it moves with, and why
The proxy is the US commercial electricity tariff. Power is the main running cost of a data centre, so operators chase cheap, reliable and clean supply through long-term contracts, while rapid demand growth from AI is pushing up local grid costs.
Proxy US PPI: commercial electric power index · monthly, Jan 2006 to Sep 2026There is no free public price for data centres itself, so this tab uses the closest public series as a stand-in. Read the numbers as an indication, not as the price of this exact item.
What its price moves with
Selected linkData centres and Energy-intensive industry
Monthly changes 0.83 strong correlation 12-month changes 0.63 moderate correlation Last 5 years 0.85 monthly changes Lead or lag none No clear lead or lag Price function, Jan 2006 to Sep 2026 Data centres ≈ 2.56 × (Energy-intensive industry)0.80 R² 0.95 On this fit, a 10% higher energy-intensive industry price goes with a 8.0% higher data centres price over the long run. Units: index and index.Chart: Data centres and Energy-intensive industry prices rebased to 100 at Jan 2006, log scale
Both prices rebased to 100 at Jan 2006, on a log scale so equal percentage moves look the same size.Chart: Data centres price against Energy-intensive industry price, monthly, log scales, with fitted line
Each dot is one month; paler dots are older (2006), darker dots recent (2026); the ringed dot is the latest month. The dashed line is the fitted price function. Both axes are logarithmic.Commercial and industrial tariffs come from the same utilities and fuel costs, so they move together closely. Monthly moves correlate at 0.83, the strongest link here, a shared driver of generation and grid costs.
strong co-movement month to month · 249 months Show chartsHousehold and commercial electricity share generation and network costs, so tariffs move together. Monthly moves correlate at 0.73 and yearly at 0.89, a strong common driver set by utility costs and regulation.
strong co-movement month to month · 248 months Show chartsGas plants supply much of the power data centres draw and are often built for them. Yearly moves correlate at 0.37 with European gas, weak because US commercial tariffs follow American gas with regulatory delay.
little or none co-movement month to month · 249 months Show chartsChillers and air handling remove the heat servers make and are a big share of a data centre's energy use. Yearly moves correlate at 0.72 with the cooling equipment index, mostly a shared US price trend.
little or none co-movement month to month · 249 months Show chartsTransformers, uninterruptible power supplies and converters feed every server hall, and data centre demand has helped lengthen transformer order books. Yearly moves correlate at 0.66 with the transformer index, a common driver from grid and building investment.
weak co-movement month to month · 249 months Show chartsLarge tech firms are signing contracts for nuclear power, including restarts and small reactors, to supply data centres around the clock. The uranium proxy shows no link with tariffs, at 0.01 monthly, since fuel is a small part of nuclear cost.
little or none co-movement month to month · 249 months Show chartsHow to read these numbers
Correlation of monthly changes (the number in the coloured pill) compares how much each price moved, month by month, since 2006 or the first month both series exist. It runs from −1 to +1: near +1 the two usually rise and fall in the same months, near 0 there is no link, and a negative number means they tend to move in opposite directions. Above 0.7 is strong, 0.4 to 0.7 moderate, 0.2 to 0.4 weak.
12-month changes compare year-on-year moves, which shows whether two prices share the same cycles even when the month-to-month noise differs.
Price function is a straight-line fit of the two prices on logarithmic scales: price ≈ k × (partner price)β. β is how much this price moves, in per cent, for a 1% move in the partner over the whole period, and R² is the share of the variation the fit explains. Two prices that both drift up with inflation can fit well without any real link, so read it together with the monthly correlation.
Lead or lag checks whether the monthly correlation is clearly higher when one price is shifted by 1 to 6 months.
Correlation is not cause. The reasons on each card explain the economic link; the numbers show how strongly it has shown up in prices. Proxy series are US producer price indexes or related prices and can blur the real link.
Data: World Bank Commodity Price Data (Pink Sheet), October 2026; US Bureau of Labor Statistics producer price indexes, IMF primary commodity prices and US EIA prices via FRED, Federal Reserve Bank of St. Louis, retrieved November 2026. Calculations by this site.