Insights from Nomad Futurist Foundation Ambassador James Rix
Every week the press is awash with announcements on the latest and greatest: new campus, largest in Europe, APAC, or USA, etc. It seems like each announcement tries to outdo the last, with every project needing to claim at least 200MW, or risk being seen as ‘lightweight’ or not credible.
Let’s try to unpack some of this.
Currently, we do not know how much energy each data centre actually consumes. The EU is beginning to uncover this through regulatory reporting.
Usually, the ‘bragga watts’ in these announcements are for 100% load, at some unspecified point in the future. As discussed previously, very few facilities ever operate at full capacity.
Impressive as a 300MW campus may sound, what is the timescale from RFS to actually reach 300MW? Let’s examine a few details of a fictitious owner/operator, Bragga Watts DC Inc (BWDC). If BWDC is building a hyperscale campus, the 300MW may be divided into multiple data halls, e.g., 15MW each, with four halls per building, potentially resulting in five 60MW buildings on a single site, or some other combination.
Will BWDC fit out all five buildings at once? Unlikely. Even a hyperscaler would phase the development, as the capital cost to build the entire site in one go is rarely feasible. Additionally, if BWDC isn’t a hyperscaler itself, it may take time to secure a key tenant for all buildings.
The anchor tenant may gradually occupy space, installing in two halls in year one, then a further two halls in year two, repeating for other buildings.
Therefore, from the announcement of a 300MW site, BWDC may initially go to market with only two buildings, costing around USD $1.2 billion, with the remaining three buildings following later.
Four years after announcing 300MW, only 120MW may be installed, possibly consuming less than 100MW of IT load. It could take another five years before the subsequent buildings are built out, meaning the full 300MW may take up to a decade to realize.
Cynics may point out that AI deployment could consume unused capacity. While AI hasn’t been directly included in the calculations above, its impact on power usage remains variable – the engineering answer is, as always, ‘it depends.’ For the purposes of discussion, this provides a solid starting point.
Want to dive deeper into the realities behind data center megaprojects? Join the Nomad Futurist Foundation to access insights, debates, and industry perspectives shaping the future of digital infrastructure.
