Revitalising aging data centres in the era of Artificial Intelligence (GenAI)

Data centres often conjure an image of a pristine and immaculate site, newly built with all the bells and whistles to support the growing demands for colocation/ GenAI services. However, for 47% of global facilities (built over 11 years ago) this is not the case and their aging mechanical and electrical (M&E) equipment can drastically impact valuations when disposing/ acquiring these sites.

As data centres age, the M&E equipment deteriorates. This leads to expensive maintenance/ replacement costs that need to be appropriately accounted for within the business plan, if not then they can impact the transaction valuation. The future reliability and Power Usage Effectiveness (PUE) of the facility will also be in question. We have recently seen a number of transactions where aging M&E has significantly impacted the Enterprise Value (EV). This is due to the necessary uplifts required in the Seller’s business plan to account for the additional maintenance capex.

Investors should also be assessing how the facility could be retrofitted to increase the rack density and take advantage of increased client workload demands (average rack design densities sit around 8kW per rack, whereas Gen AI workloads require 30kW+ per rack).

A proactive approach to planning and maintenance, alongside a sensible business plan, could play a vital role in driving maximum value from the asset.

In this article we will look at the era of aging digital infrastructure and outlining some of the themes that can impact valuations. We repeatedly see 3 key assessment criteria that must be considered by both sellers and buyers during the M&A process; 1) what are the considerations and cost implications of end of life equipment, 2), the impact of the changing regulatory environment must be considered, and 3) how can these facilities be retrofitted to meet increasing customer demands.

What are the considerations and cost implications of end of life equipment?

Sellers/ Buyers need to carefully consider the impact of end of life equipment within the business plan. This will be out of necessity rather than optionally and many facilities will face spiraling maintenance capex costs. Typically, most smaller items (e.g. UPS batteries) are replaced every 5-7 years and represent relatively minimal cost (£250k+ / MW). However, when facilities run closer to 15-25 years old then the big ticket items (Generators, Transformers etc) start to reach end of life and these can represent significant spiraling costs (£1,000k+ / MW). Therefore, the question becomes: should you replace the equipment, like for like, post transaction or do you upgrade your facility to meet increasing customer workload demands?

Main Category	Asset Group	Typical equipment cost (£k/ MW)
Ventilation & Air Con	Evaporative Condensers	250+
Ventilation & Air Con	CRAC/ CRAH	450+
Electrical	HV Transformer	100+
Electrical	Generator diesel, oil	350+

^{Notes:

CRAC/ CRAH - Computer Room Air Conditioning/ Handler

HV - High Voltage

kW - Kilowatt}

The impact of the changing regulatory environment

Another consideration that must be made is: what are the regulators expecting? Owners/ Operators have made rapid PUE improvements between 2007-2014, dropping from 2.50 to 1.65, but have recently plateaued around 1.5-1.6 (see below). For most regulators this will be seen as not efficient enough and they will likely mandate much lower metrics (1.15-1.25), as we’ve already seen in Germany via the German Energy Efficiency Act where the PUE targets are set to below 1.2. These requirements will likely be adopted across the rest of Europe and Globally as data centres continue to put pressure on local grid capacity. This means that all of these facilities will require expensive cooling system upgrades to be carried out (£400k+/ MW), which may not be correctly costed for within the business plan.

Beyond PUE there will also be considerations around waste heat recovery, impacts on local water systems and the wider local environmental impact. All of this has significant cost implications that will need to be carefully considered.

How can these facilities be retrofitted to meet increasing customer demands?

With average rack design densities sitting around 8kW/ rack, this poses a potential risk as customer demands and processing requirements increase. Customers will expect much higher densities from colocation providers and even higher for GenAI workloads (30kW/rack+) to meet their future demands.

All of this represents substantial cost implications as M&E equipment and fit out typically accounts for 40%+ of the total build cost. This could mean replacement costs of between $4.8-7m/ MW to retrofit existing kit to make the site fit for future demands. Therefore, it is important to assess the future market demands as part of any transaction and cost these appropriately within the business plan.

Conclusion

When considering buying or selling any data centre it is important to carefully consider these 3 critical questions; 1) how will end of life equipment impact my investment/ costs, 2) what is the direction of travel for regulatory requirements (improved PUE etc), and 3) is the design suitable for future customer demands? Carefully assessing these points is imperative to the success of any M&A process.

If you need help assessing these items as part of your transactions then the team here at PwC have the expertise and experience to support. Please reach out and we’d be happy to set up a discussion with our qualified team of Uptime Accredited Data Centre Tier Designers.

Sources: