The global economy might be suffering from inflation and low growth, but that hasn’t stopped the first half of 2023 from being the busiest on record for datacenter takeup in Europe, according to new research.
A report [PDF] from commercial real estate outfit JLL claims there was record demand for datacenter capacity during the period from April to June this year alone, with 114 MW of take-up across Europe’s leading markets including Frankfurt, London, Amsterdam, Paris and Dublin – referred to by the charming acronym of FLAP-D.
London remains the largest market with a total of 902 MW, according to JLL, amounting to a 35 percent share of the total Tier 1 capacity. However, here there was a fall-off in the volume of new capacity added so far this year, at only 7 MW.
In contrast, Frankfurt saw 44 MW of additional takeup during the second quarter, bringing the city’s total for the year so far to 80 MW, in comparison to just 26 MW at the same point last year.
Capacity in Amsterdam sits at 458 MW, with 10 MW added in Q2, but there has been no significant new supply added to this market since a moratorium on new datacenter development ended in 2020.
Paris has seen 24 MW of new capacity added so far in 2023, while Dublin gained an additional 12 MW, increasing the overall market size by 6 percent to 199 MW, JLL said.
According to the report, at least some of this wave of new uptake can be attributed to pent-up demand. The last couple of years saw significant levels of pre-leasing as customers aimed to secure new datacenter space while it was still being developed, JLL said, while the supply chain issues that were seen across the industry led to completion dates being pushed back.
The report also notes that the emergence of generative AI has been having an impact on datacenters, an effect also previously commented on by research outfit Omdia. In fact, the high compute power requirements needed for the training of AI models is driving a fundamental change in the very design of datacenters, JLL claimed.
It states that high-performance GPUs have been added to the compute infrastructure in order to meet the requirements of training AI models, and these are pushing the need for liquid cooling rather than air cooling.
JLL’s Head of EMEA Datacenter Transactions Tom Glover observes that infrastructure for training AI models is less reliant on latency, so proximity to the end user may be less of a consideration.
This may lead to some datacenters being strategically located where access to renewable energy, available land and water for cooling is not an issue, he claimed. However, this would seem to imply such a datacenter being built from the outset to be largely dedicated to running AI training workloads.
Glover said that private equity firms are ready to invest based on the growth potential of AI. He cited Blackstone as earmarking properties built to handle the rising data needs from a surge in AI, while DigitalBridge has announced a fund of $1.2 billion geared towards AI investment.
However, a separate report earlier this year claimed that the European datacenter industry was facing challenges in meeting the growing demand for capacity, due to issues such as materials and heavy equipment to build sites being in short supply.
It stated that there was an insufficient pipeline of datacenter development planned in Europe over the next three years to meet the forecast increase in demand.
Source : Theregister