THERE IS NO QUESTION THAT POWER IS THE BOTTLENECK IN DATA CENTRE BUILDS RIGHT NOW .
E D I T O R ' S Q U E S T I O N
MICAH GERTSON , ASSOCIATE PARTNER , ALTMAN SOLON
There is no question that power is the bottleneck in data centre builds right now , especially when trying to address AI . For greenfield builds , getting a new grid offtake agreement for 10 + MW can take years depending on the utility . For brownfield DCs , supporting higher rack densities by upgrading the power distribution and cooling inside the data halls is complicated enough without disturbing your customer ’ s mission-critical IT infrastructure .
“
THERE IS NO QUESTION THAT POWER IS THE BOTTLENECK IN DATA CENTRE BUILDS RIGHT NOW .
In the greenfield context , we have observed three key power strategies from our leading DC clients which helps them to significantly benefit from the AI wave :
1 . Prioritise power – Make powerbanking as important as landbanking by adding on-the-ground power procurement teams to your land prospecting team . Knowing how to work with utilities across each country / geography is a granular task greatly improved by in-market presence . For larger international players , this likely means hiring 10s of people dedicated to power development versus the few that many operators have today .
2 . Create a ‘ menu ’ of sites – DC tenants deploying IT infrastructure for AI move quickly and are themselves under significant time pressure to monetise their assets ( mainly GPUs ). Having a menu of 10 + MW site options for which you have developed power helps operators provide optionality for AI-driven tenants . Location doesn ’ t have to be perfect these days , e . g . suburban Milan or Madrid or rural locations in the Nordics can be very interesting to a tenant if it gives them the ability to deploy 10 + MWs of GPUs in six to nine months .
3 . Extract value from powered land – As an investor , look to invest in operating platforms that have larger land and power banks and at least one scaled AI-driven DC tenant . Even if they won ’ t develop every site into an operating DC , we see an increasingly healthy market for selling powered land- so there are more ways than ever to monetise the power team ’ s efforts .
In the brownfield context , we find addressing AI workloads at scale to be much more opportunistic . The main strategy that an enterprise colocation operator can employ is to set up some rows or cages of reardoor heat exchange racks to offer to their customers as a proof-of-concept . This kind of localised liquid cooling deployment limits the requirement for large-scale plumbing refits for liquid loops and CDUs , while allowing you to offer high density racks of 50 + kW to take advantage of unutilised power capacity .
AI-driven DC demand is an exciting but also challenging opportunity to address . This segment is still nascent , and there is no silver bullet – which means there are likely good opportunities for most operators and investors .
32 www . intelligentdatacentres . com