好色先生TV

01 / Introduction

Data centre (DC) investment is construction鈥檚 biggest countercyclical demand story. Global Data reports that the big five tech companies, led by Microsoft, invested $35bn into their DC fleets in 2023. Real estate investment is racing ahead too, and Linklaters has tracked US$22bn of deals in DC capacity in the first five months of 2024 alone.

The DC growth story is well documented. A combination of exponential growth in cloud computing and a gold rush for AI leadership has triggered an unprecedented race for power and fibre connections, data hall floorspace and technical equipment. Additionally, it is also triggering a wave of DC refurbishments as owners upgrade facilities in line with client requirements.

Growth at this pace brings pain as well as promise. Access to power has become hotly contested, particularly as DC development has been blamed for power and water shortages in locations as diverse as London, Dublin and Amsterdam. As a result, owners and developers have shifted their focus of investment away from the core FLAPD markets of Frankfurt, London, Amsterdam, Paris and Dublin to a wider range of second-tier markets including Italy, Spain and the Nordic countries.

Given the importance of first-mover advantage in the digital world, assured day-one delivery is the overriding critical success factor for DC development. The risk cascade flows from the end user 鈥� which could be, for example, an AI inference service 鈥� through the DC occupier to the developer or constructor. Penalties are substantial, but delivery is complicated by shortages of plant and equipment and by exacting user-testing at completion. As projects are scaled up to 80MW and beyond, representing an initial construction investment of close to 鈧�1bn, the risk exposure for clients and contractors is huge.

>>Also read: Market forecast: Demand is set to rise as the economy stabilises

>>Also read: Human or artificial intelligence? It鈥檚 all a matter of trust

Unlike most real estate sectors, which have a regional focus, DC clients are global operators delivering regional programmes. Some DC requirements 鈥� including edge processing, regulated personal data and secure facilities 鈥� are location sensitive, but a great many compute and storage requirements are footloose. Clients need to be remarkably flexible 鈥� adapting their designs to meet new technology requirements while also adapting their procurement strategies in line with local circumstances, as well as a scarcity of equipment, industry capacity and capability.

In this article we will examine the complexity of the DC procurement landscape 鈥� including the different characteristics of key players in the market and also ways in which DC procurement differs from other complex building types. We highlight practical solutions to the challenges that clients and their supply chains face in a gold rush market that has become increasingly important as a source of industry workload.

02 / Client approaches to data centre delivery

The market is very diverse, with over 1,250 data centres in Europe alone. Most are smaller, dedicated enterprise DCs providing secure facilities for users such as investment banks. Overall, enterprise DCs are thought to account for around 40% of global critical IT load.

Users increasingly access their DC services either via the cloud from a hyperscale provider like Microsoft or Amazon, or via their own servers operating from a carrier-neutral co-location centre. Development of DC capacity has mostly consolidated around the hyperscalers and co-location developers, which both operate tier-three DC facilities featuring high levels of redundancy in the plant and diverse power supplies and data feeds. Forecasts by Synergy Research Group predict that 80% of critical IT load will be via shared provision by 2029. Hyperscalers are the most important tenants of co-location providers.

Common characteristics

There are some key characteristics that both hyperscalers and co-location developer share:

• Need for speed Rapid development driven by demanding end-user requirements.
• Cross-border operation The need to adapt ways of working to suit local markets.
• Exposure to capacity constraint All developers are competing for scarce capacity in a rapidly accelerating market. Implications include paying premium costs for equipment.
• Operating at scale The ability to aggregate demand for equipment and construction materials across a programme. This strength is balanced against a dependence on the capacity and performance of the major equipment supply chain.
• Fleet standardisation Reference designs provide a starting point for DC development but are usually subject to significant adaptation to align to local regulation and end-user requirements.
• Campus scale planning Both types of developer are adopting a phased approach where possible, delivering incremental capacity in 10MW-20MW self-contained units as part of mega鈥憇cale delivery.

Co-location centre clients

The core business of the co-location client is the development and operation of carrier-neutral DCs. Outside of China, the largest global operators are Equinix, Digital Realty and NTT. There is a thriving mid-size market of rapidly growing providers, including Colt, CyrusOne and QTS.

Co-location centre operators offer a variety of lease models ranging from a powered shell 鈥� where a tenant is responsible for all plant and equipment serving data halls plus the technical fit鈥憃ut 鈥� to a more turnkey solution aimed at smaller occupiers.

Key points of difference for a co鈥憀ocation operator are:

• End-user focus Developers are ultimately focused on their tenants鈥� needs and stakeholder management is comparatively simple, with drivers for change coming mostly from end users.
• Viability sensitive A shortage in DC space means that rental rates are currently rising, but this cannot be guaranteed long term. Commercial returns are sensitive to risks around inflation, contractor availability and change management.
• Flexibility and adaptability Developers need to respond to client demands in real time as leases are finalised, such as new requirements for high-performance liquid cooling. Change management processes need to capture the full cost and time implications before the lease agreement is finalised.

Hyperscale clients

As already highlighted, hyperscalers represent the most concentrated source of demand for DC capacity, both as tenants and as developers. Even so, DCs are an operational enabler for businesses that are overwhelmingly focused on customer needs and technology solutions.

The five digital giants 鈥� Alphabet, Amazon, Apple, Meta and Microsoft 鈥� directly owned 20% of IT load in 2023 and operated a further 20% in leased co-location space.

Within a direct development, hyperscalers have the opportunity to manage the process end-to-end. This potentially enables deep standardisation at the level of system, process and product, but paradoxically also involves a much more complex stakeholder environment affecting development priorities, technology changes and so on.

Key points of difference for a hyperscaler are:

• Start-to-end delivery Total control of the construction process makes it easier to introduce innovation such as close-coupled cooling, which involves the integration of complex MEP solutions within the data racks themselves.
• Workload Hyperscalers are extremely responsive to changes in their client markets, so patterns of DC workload can fluctuate significantly 鈥� this has implications for workload planning and for contractors鈥� DC market strategies.
• Change management There is significant scope for change throughout a development programme as clients adjust to developments across their markets and technology stack. This puts a premium on a project team鈥檚 ability to respond to and manage change impacts in a fully integrated digital environment.

Implications

Co-location and hyperscale clients share many similarities, including the technologies that they deploy, the supply chains they share and the constraints that make it difficult to scale capacity.

The biggest challenge within the DC sector is market concentration, which is driven mostly by hyperscalers鈥� rapid expansion plans.

Higher levels of demand, driven particularly by parallel working on competing schemes, are expected to result in further capacity constraint.

While hyperscalers are arguably less responsive to market conditions than pure-play developers in the co-location segment, they have the advantages of deeper pockets and greater flexibility in their approach to the market. Direct procurement by hyperscale clients can be expected to become an even larger market segment.

03 / Key procurement challenges and solutions

Data centre projects form part of very large, highly complex and fast-moving programmes. These present significant challenges for the client and construction supply chain in accessing resources, co-ordination and quality control. Well-known and widespread problems associated with permitting and the sourcing of diverse power and fibre  connections highlight that construction is just one source of complexity and risk affecting the client.

Design standardisation and customisation

The starting point for most DC projects is a standardised design solution, typically developed to the equivalent of RIBA Level 3. This will usually be developed to Level 4+ by the client鈥檚 design team incorporating adaptions in line with local building regulations.

Hyperscale clients will typically retain responsibility for the design during construction, whereas co-location operators need to be able to transfer more risk through design and build contracts in line with their commercial model.

Long lead-in items

Clients have to undertake direct procurement of programme-critical plant alongside the general contractor (GC) appointment. Once total costs of installation and integration are accounted for, the value of these packages will be around 30%-35% of total capital expenditure, depending on resilience and redundancy requirements.

Both hyperscale and co-location clients will manage frameworks for key equipment including generators, UPS, switchgear and so on, forecasting demand for equipment on the basis of demand curves. Sole sourcing is rare due to single-point-of-failure risks. Equipment procurement is typically undertaken in competition, although agreements may include provisions for volume discounts and inflation risk-sharing. Equipment providers can facilitate faster delivery at a premium price.

Integration of free issue plant and equipment

OFCI (meaning owner-furnished, contractor-installed) is the means by which long lead-in plant and equipment is integrated into the works. The equipment manufacturer is typically responsible for witness testing, installation and plug-in and for the performance of the kit.

As part of the OFCI model, both the consultant and contractor teams have critical roles in co-ordinating quality control, delivery, storage, installation, integration and final commissioning. The GC appointed for the services installation (MEP) will typically have responsibility for integration of the equipment into the wider MEP build, while the consultant team will cover off all client-side responsibilities, including the management of witness testing.

Balance of plant

The remaining scope of MEP, termed balance of plant (BOP), will typically equate to a further 20% to 25% of the total capital expenditure. The scope will include the detailed design and the co鈥憃rdination and installation of complex, labour-intensive and hard-to-prefabricate distribution systems for power, air and water, in addition to fire suppression systems, security and the building management system.

The design of BOP facilitates the 鈥減lug-in鈥� installation of OFCI equipment, requiring close co鈥憃rdination with equipment manufacturers, particularly with respect to the design and configuration of building management systems.

Data hall fit-out

A lot of innovation in the sector is focused on data halls, and particularly in relation to the management of soaring heat loads. Conventional solutions based on computer room air-handling (CRAH) units in data halls are being superseded by close-coupled solutions that facilitate much more efficient cooling. With greater integration of cooling systems within racks themselves, there is more opportunity for plug-and-play installation.

In a co-location centre, clients typically secure access to a data hall one week ahead of sectional completion, facilitating some early assembly work before system switch-on. Security considerations with respect to contractor clearance and access management are critical in a part-operational DC.

Prefabricated and modular solutions

Prefabricated systems play a huge role in delivery of the modern DC, facilitated by client-led OFCI procurement. Containerised generators and UPS systems are naturally suited to a modular approach focused on a self-contained system.

Hyperscalers have further opportunities to increase the scope of modularisation in more complex assemblies, including data racks that will integrate multiple systems such as cooling and controls and which may also include some client-derived intellectual property (IP).

The certainty offered by pre-installation factory testing is an important source of value. Even so, multiple suppliers are needed to mitigate the risk of single points of failure.

Management, quality control and witness testing

Resourcing requirements associated with supervision and testing on DCs are much higher than for many other project types, typically 25% of the workforce, and are at the heart of some of the resource constraints affecting the sector.

In addition to the intensity and complexity of the services installations, the compartmented nature of DC design means a lot of management resource is needed to maintain contact and quality control. Resource constraints may also affect the client鈥檚 ability to participate in witness testing and sign-off, increasing the pressure on any float in the programme.

Phased development

Over the past four to five years, the size of DC developments has increased at an astonishing rate. Whereas 20MW schemes costing 鈧�200m to 鈧�250m would have been considered large before covid, many clients on both the hyperscale and co-location side of the market are considering schemes of between 150MW and 200MW, with a potential construction value of around 鈧�2bn.

Programmes on such a scale will always be phased, and there are some critical considerations for the successful management of phasing 鈥� including the make-up of the delivery team, pressure on wider site infrastructure including logistics and even the wider impact of a development on the local community and economy.

One of the most pressing challenges for DC teams is the assembly of appropriately skilled teams. Even with explosive growth in the sector, clients and contractors are all working in a tight labour market. Use of digital systems to supplement skilled inputs and to access global skills pools will be even more important as the sector goes through its current growth phase.

04 / International procurement considerations

One distinct feature of the European DC market is a rapid expansion of locations for new schemes. The core FLAPD cities of Frankfurt, London, Amsterdam, Paris and Dublin have long been seen as prime development locations, but shortages of suitable sites and available power have triggered growth in favoured tier-two locations such as Madrid, Lisbon and Milan as well as more remotely located hyperscale developments in the Nordic countries.

Overlaying the complexities of DC procurement onto local markets in Europe adds new dimensions, including different contracting models, preferred construction methods and local supply chain structures. Developing local knowledge and working with the grain of the market makes it easier to access resources.

Core markets including France, Germany and Italy have large contractors and reasonably deep supply chains. By contrast, clients in Scandinavia need a more flexible strategy to access local capability. In addition to adapting procurement routes to local practice, detailed design will need to be aligned to local standards, which can affect areas including: layouts for fire escape, MEP design for energy efficiency, fire safety and the specification of on-site renewables such as photovoltaics.

France

Workload is largely focused on three major GCs: Bouygues, Eiffage and Vinci, each of which have a large in-house workforce that could potentially result in some capacity constraint. Tier-two contractors team up in joint ventures to pursue work but find it hard to demonstrate capability. Mercury from Ireland has also established a credible position in the French market.

The code-based legal system in France means contracts operate differently, with the GC working to a performance specification with limited explicit reporting and governance obligations.

In practice, arrangements outside the contract, typically led by the consultant team, make the process work in line with client expectation for greater levels of control and transparency.

Given the added complexity, more entrepreneurial co-location developers have often been early movers in the French market. The complexity also acts as a barrier to entry for international contractors. Extra capacity is expected to be created in the domestic market by tier-one GCs making further contractor acquisitions and by tier-two GCs establishing more joint ventures.

Germany

GC appointments are typically split between building shell and MEP, with the building shell GC undertaking most of its delivery using an in-house workforce. The MEP GC also takes responsibility for the OFCI co鈥憃rdination.

High-tech construction is experiencing a boom in Germany, with Wafer Fab and Giga Factory investments competing for experienced GCs. This means DC clients must focus on attracting their supply chain by shifting the balance of commercial terms in favour of the GC and by adapting standard shell designs to align with preferred building solutions. Although GCs do take design responsibility, the client is required to develop a near-fully detailed design to secure regulatory approvals, losing some of the benefits of standardisation and automation.

Ireland

Ireland has long been a favoured location for inward investment and is a core location for hyperscalers. The DC sector is an important part of a relatively small construction market, which means clients must work closely with contractors, sharing the forward pipeline to support capacity planning.

Hyperscalers support supply chain training initiatives, so the skills crunch is being actively addressed. Irish contractors have developed a deep expertise in DC delivery, including prefabricated solutions, which they now actively export throughout Europe.

Although clients will have less difficulty in accessing contracting expertise in Ireland, permitting is becoming increasingly difficult as a result of power grid capacity constraints.

Scandinavia

Nordic construction markets are relatively small, so local GCs typically do not have the scale to manage the full risk of DC delivery. Options available to clients include the employment of a global GC on a fly-in, fly-out basis or package-based procurement where the client鈥檚 team takes on a greater co鈥憃rdination role.

Package contracting works well in the region, accessing for example civils contractors with experience of groundworks in hard rock. By contrast, clients may need to bring in external MEP package contractors and other specialists to bolster regional capacity.

05 / Conclusions

With the AI market forecast to grow from US$300bn in 2024 to over US$1.4tn by 2030, the DC sector is locked into an investment and innovation super-cycle placing huge demands on the supply chain. Growth will be focused on a relatively small group of hyperscale and co-location clients, which will continue to compete against one another for resources while depending on each other for growth and access to data hall capacity.

Even as DC programmes grow in scale, environmental, power supply and supply chain constraints will remain a challenge, requiring innovative design solutions and flexible procurement models to assure end date certainty.

DC clients will continue to work closely with their project teams to find new ways to grow their delivery capacity including the greater adoption of system and process automation.