What were once IT facilities are quickly becoming energy-intensive industrial operations.
The problem is that the underlying power architecture hasn’t kept pace.
Traditional alternating current-based systems rely on multiple conversion steps, each introducing inefficiency and heat.
At large scale, even marginal inefficiencies create significant cost exposure. An indicative calculation could show that in a 100 MW data centre, just 1–3% avoidable power loss across conversion stages well within the range observed in industry guidance2 can equate to:
- 1–3 MW of continuously wasted power
- ~8–26 GWh of lost energy annually (the equivalent of powering a small UK town for a year)
- £1–4 million+ in annual electricity cost (depending on the Power Purchase Agreement)
That lost energy is also converted into heat, compounding cooling demand and further increasing operational expenditure.
At small scale, that’s manageable. At hyperscale, it becomes a real constraint that drives up costs, increases cooling requirements, and limits how far infrastructure can scale.
That’s why the data centre industry is increasingly shifting toward high-voltage direct current (HVDC), particularly emerging 800 VDC architectures enabled by medium-voltage (MV) Solid-State Transformers (SSTs), while 1500 VDC systems remain a longer-term future direction for hyperscale facilities.
The SSTs enable simplification of the power chain and reduce conversion losses. These systems deliver more usable energy to the processor, reduce thermal load, and make it easier to scale dense AI environments.
