AI has rewritten the data centre design brief

Baynes is direct about how fundamentally AI has changed the conversation. “If you’re not talking AI, then you’re not really talking about the current state of the market today,” he says. The shift is most visible in rack density. “If you look at a typical cloud data centre, it’s 10 kilowatts per rack. With AI training and the use of Nvidia GPU cluster technology, we’re now looking at up to 100, 200 kilowatts-plus per rack. The way we design data centres has to change fundamentally.”

The grid cannot keep pace

That demand is colliding with a UK grid that Baynes says is already oversubscribed. “The amount of grid connection requests in the pipeline at the moment outstrips the ability to supply today from the grid,” he says. Asked where the sharpest bottleneck lies, grid capacity, planning permission, or the pace of renewable build-out, his answer is unambiguous. “Grid capacity is the biggest constraint at the moment to growth. Access to energy is the biggest challenge that data centre operators face today, not just in the UK, but around the world.”

Speed, he argues, has become the defining currency of the sector. “There’s a lot of investment that has to go into data centre builds. You have to have pretty deep pockets in this space, so you need to get your return on investment. The way to do that is to get to market as quickly as possible.”

Data centres as active grid participants

Rather than treating the grid purely as a constraint, Baynes sees an opportunity for data centres to become contributors to grid stability. “Data centres could be an active asset on the grid, which can ease some of the burden in terms of demand, not just consuming, but actually managing the workloads and flexing that with the grid as they go.” That could mean facilities scaling back consumption at peak times and feeding capacity back during quieter periods, giving the grid more flexibility to support wider demand.

Community engagement as licence to operate

Baynes is candid about the reputational challenge facing the sector. “There’s a view on data centres which might not be favourable in some quarters. We’ve seen some moratoriums in places because of the demand for energy.” His prescription is early, transparent engagement with local authorities, and a tangible community benefit, not just infrastructure investment. He points to waste heat reuse as one practical example, such as channelling excess heat from a facility into a local leisure pool or residential heating scheme. “These discussions and early collaboration can help with the relationship with local councils, to perhaps speed up permitting and get over some of the hurdles councils face.”

Hitting emissions targets without slowing growth

On the UK government’s target of an 87% emissions cut by 2042, Baynes believes it is achievable, but only under specific conditions. “Yes, but only if digital growth and energy transition happen together, hand in hand. AI needs to be supported by cleaner grids, greater investment in renewables, and more efficient technologies.” He cites the shift from diesel to HVO in backup generators as one example of incremental change already underway, alongside battery storage and diversified energy strategies. “Data centres can actually themselves contribute to that, they can be supportive and help the government reach some of these targets through technology, through innovation, and through early adoption of being a more sustainable and more renewable functional building.”

RELATED STORIES

How Schneider Electric is scaling liquid cooling and power for AI data centres

Schneider Electric partners with Foxconn to scale AI data centres