Regional data centre strategy: At a crossroads?

Download Listen

As compute and storage demands have risen, the traditional data centre strategy of ‘build and they will come” has served the industry well. However, the demands of enterprises – enabled by AI-driven automation – are becoming more diverse. A one-size-fits-all approach will no longer work: Data centre strategies must become more varied and more customer-oriented.

The traditional data centre strategy: build and they will come

The data centre industry has enjoyed enormous growth over the last two decades as the Internet has spawned vast quantities of data and new ‘as a service’ business models in software and compute. Much of this growth has come in the form of hyperscale facilities form the major players – Google, AWS, Microsoft Azure, Meta, Alibaba and the like. Hyperscale data centres are more energy efficient, and their scale enables them to balance compute and storage demands effectively. Regional data centres – located in tier 2 and 3 cities – have provided colocation facilities for enterprises seeking local compute and storage that want to avoid the hassle and cost of running their own facilities.

With strong demand, the focus of data centre operators has been on solving supply-side issues: securing powered land, access to water for cooling, and planning permits. It is only a little simplistic to say that the data centre business has been a real estate play: securing sites, managing the build, and operating a cooled warehouse securely and efficiently. Filling up a data centre has been relatively straightforward: Hyperscalers can predict demand for their services accurately and so haven’t ‘overbuilt’ and regional data centres have been supply-constrained so there has been plenty of demand from enterprises and governments (and Hyperscalers moving closer to the ‘edge’) for their facilities. But this situation is changing.

New forces driving data centre transformation

Until recently, managing compute workloads was relatively straightforward for enterprises. There were broadly three compute and storage options:

1. For applications requiring ultra-low latency – such as manufacturing systems – on premises compute was the answer. This was also true where close enterprise control was deemed critical – where an application or company data was too sensitive to manage in a third-party data centre.

2. Colocation, usually in a local data centre, was an option where data sovereignty remained important – for example where the public cloud was deemed a risk owing the potential for data to be moved beyond a specified jurisdiction.

3. For many other workloads – those used to run software as a service, social media, data analytics, disaster recovery, etc. – the flexibility and cost benefits associated with public cloud was the solution.

Now, there are alterations both on the demand side, from customers, and on the supply side – from investors, regulators, and from the changing nature of data centre competitors and the structure of the industry – which call into question a strategy focused predominantly on building facilities. For many regional data centres and those ‘at the edge’, this strategy will come under pressure in the near term. For larger data centres, such as many in the FLAPD market in Western Europe, the traditional business model is likely to continue to be successful for longer. Nevertheless, even here, it will eventually come under threat.

Let’s explore these forces in a bit more detail.

Demand-side forces

Over the next decade or so, the range of local compute options available for companies will increase to meet their growing range of needs. The reasons for enterprise hybrid (cloud) compute solutions include:

• Increases in automation generally, and in AI-driven activities specifically, which require massive (hyperscale) compute to train Large Language Models (LLMs) and local low-latency compute for inferencing and acting quickly. AI is, therefore, driving compute growth at hyperscale compute facilities now and, as inferencing needs increase to meet a growing number of use cases including those involving AI agents, this will drive demand up for regional, network, and on-premises compute.

• More stringent demands for sovereign compute solutions from regulators meaning a growth in local public and private cloud solutions. We wrote recently about developments in the EU on data sovereignty and cybersecurity and the funding that the €1.2bn in funding that the EU has invested to encourage European companies to build a sovereign cloud and challenge the dominance of the US hyperscalers. We also touched on AWS’s plans to invest €7.8bn in building a European Sovereign cloud. The is a further impetus for increased local compute infrastructure.

• The desire from many companies to repatriate applications and data back from the public cloud to other more local solutions. We covered the reasons for this in a recent article. These include compliance associated with data sovereignty mentioned in the bullet above, but also include cost and control concerns. Indeed, in a survey by Barclays Bank during 2024 83% of CIOs said they planned to move workloads back to the private cloud from the public cloud – up from 43% in 2020.

• Concerns about sustainability – both the carbon footprint associated with powering data centres and the water to cool them. Distributed compute – either at the far edge or in regional data centres – provides the opportunity to defray cooling issues by providing heat to residential, office, and recreational buildings (e.g. swimming pools).

Supply-side forces

Historically, by contrast to the consolidation inherent amongst the hyperscalers, the regional data centre industry has been relatively fragmented – particularly in Europe. There has been a handful of larger players, such as Equinix, offering internet exchanges globally but, generally, the market has comprised of lots of smaller local data centres meeting the compute and storage needs of local businesses. Asset-lite managed services providers have been responsible for ‘knitting together’ solutions from a range of data centres to provide national and regional offerings for larger firms.

But, again, this is starting to change. As regional and edge compute is attracting more capital, data centre consolidation strategies are gaining traction, reshaping the market. This is happening at a national level in Europe – for example Pulsant, in the UK, has grown to a portfolio of twelve data centres across the UK by snapping up individual sites. Consolidation is also starting to occur at a regional level – with German headquartered nLighten leading the charge to build a pan-European regional data centre ‘platform’ through acquisitions in Spain, France, Switzerland, Belgium, Netherlands, and the UK.

Consolidation is further driving greater diversity in local compute solutions because the price investors pay for an asset is not just driven by ‘supply side’ attributes (data centre capacity, expansion potential, power availability (which is particularly constrained in the FLAPD markets and likely to increase demand in tier 2 and 3 locations), sustainability concerns of investors etc.) but also by its commercial operations. Vendors can attract higher prices by differentiating their sites from competitors based on their portfolio of services and customers. Higher growth and margin solutions and customers will drive asset prices up. Colo and public cloud alone are unlikely to yield a premium price if market demand is shifting to highly interconnected private cloud!

 

New regional data centre strategy required

Hybrid compute options

At STL, we believe that these demand- and supply-side forces will lead enterprises to adopt a range of compute and storage solutions to meet their specific portfolio of use cases. We outline the five core infrastructure solutions that enterprises will adopt in their ‘hybrid compute’ strategies in the chart below. Four of these – colocation, single-tenant private cloud, multi-tenant private cloud, and public cloud are relevant to regional data centres. Private cloud will continue to grow rapidly and regional data centres have a strategic choice about whether they participate in this market. Investment will be needed, and success is not assured: for example, failing to provide a cloud management solution that works for managed service providers and enterprises will result in low take-up of private cloud. But not competing is self-limiting as the market for colo slows and hyperscalers continue to use their scale to squeeze margins in public cloud.

Data centre strategy: core infrastructure solutions overview

The need for a services-oriented data centre strategy

Many regional data centres currently don’t offer private cloud, or it is a small part of their portfolio. Despite the risks of entering the private cloud market, ignoring these sectors and sticking to colo and public cloud will inherently limit revenue and margin growth. Either way, regional data centres must consciously decide on their service portfolio in a way they have not had to consider before. The compute world is becoming more complex for enterprises and this complexity requires a more nuanced commercial strategy from data centres. Management teams at regional data centres need to consider which customers they will serve and with what services. A regional data centre commercial strategy should seek to answer the following questions:

1. What use cases to address? For example, where and how to address the demand for AI applications?
2. Which vertical industries to serve? Horizontal strategies are likely to be replaced with a more verticalised approach as data centres need to show expertise which is relevant to the specific compute needs of their clients?
3. What portfolio of services to offer? Stick with colo and/or public cloud or embrace the growth (and complexity) of private cloud?
4. How to optimise channel strategy? – for example, how to work with managed IT service providers?
5. How important is connectivity For instance, a large amount of connectivity from multiple carriers is particularly important at the edge and for private cloud.
6. Whether to venture beyond infrastructure? Move into professional services and managing applications?
7. What opportunities exist to acquire new assets with a complementary services portfolio?

Download this article as a PDF