Every few years, the energy market produces a sharp reminder that price stability is the exception, not the rule. The European wholesale price spikes of recent years made the point clearly. The next reminder will look different in cause and shape, but it will come. Energy is a commodity, commodity prices move, and the data centre is one of the most energy-intensive industrial workloads on the grid.
How a data centre operator weathers those moves depends on two things. The structure of their energy contracts. And the efficiency of the IT load they are running. The first is a procurement question handled at the energy desk. The second is a procurement question handled at the IT desk. The two functions do not talk to each other often enough.
That asymmetry is worth fixing.
How exposure actually works
Most large data centres operate on long-term bulk energy contracts that protect them from short-term spot price volatility. That is the right structure. It also creates an illusion of insulation that disappears at the contract renewal date. When the underlying market has moved by multiples, the renewal is brutal, and the cost gets passed to colocation, networking and cloud customers shortly after.
Two things make non-domestic energy contracts different from the domestic ones most readers think of first. Non-domestic prices are less volatile day to day, but they reset at contract intervals. And smaller users pay materially more per unit than larger users, which is the opposite of how procurement intuition often runs. Both points matter for any operator running multi-site or smaller-footprint estates.
What the IT load can do about it
The variable a data centre operator actually controls, in any energy market, is how much energy the IT load consumes per unit of useful work. That is the efficiency lever, and it is the only lever that pays whether prices go up, down or sideways.
In a high-price environment, the savings from running on more efficient hardware are amplified. The same physical efficiency that produces a modest saving at baseline prices produces a substantially larger saving at multiples of baseline. The carbon arithmetic moves the same way. The case for efficient hardware does not become true when energy prices spike; it becomes harder to ignore.
How we model it
Our default reporting uses reported average non-domestic energy prices, deliberately not headline crisis numbers. That keeps the comparison like for like across customers and across years. Where a customer has a specific contract position they want us to model against, we will. Where they want a stress-test against a high-price scenario, we will. Balancing cost, energy and the carbon attached to both is the work; the underlying price assumption is a configurable input, not a fixed claim.
The point we make to customers in any market is the same. The efficiency case is the most stable line in your procurement model. It pays when energy is cheap, and it pays more when energy is expensive. Procurement decisions made on measured evidence hold their value through the volatility.
That is the case for a measurement-first approach to data centre hardware. It gets clearer every time the energy market reminds the sector that prices move.