By any measure, investment levels in the vast, technologically rich buildings holding the world's data have been remarkable. This year, the boom in AI tools and infrastructure looks to rise higher still and various tech companies are committing even more Capex to their investments.
On 2 December 2025, Johannes Callet, Partner in Germany and Lacy Gratton, Partner in London met with industry experts Alex Eagles, Executive Director at PGIM, and Paul Mortlock, Head of European Data Centre Capital Markets at CBRE, to find out what's driving capital decisions, and how current trends are likely to shape the data centre investment landscape of tomorrow.
State of the market
Daily reports track the eyewatering ambition of big US tech companies (the hyperscalers, such as IBM, Meta and Oracle) while our desks already groan under the weight of email traffic, AI tools and video call technology. Globally, datacentre capacity is currently 59GW but, according to JLL, a further 7GW will reach completion this year to service the demand. Over the next two years, Morgan Stanley estimates global spending on data centres will reach nearly $3tn.
As a major investor in data centres, PGIM recently raised $1.7 billion in equity through a closed-ended global fund. Its seven-year investment strategy involves finding strategic land and investing in hyperscalers. Critically, PGIM partners with operators and tenants to build either turnkey or powered shell data centres, leveraging their knowledge and networks to help both hyperscalers and co-locators operate efficiently.
Against this backdrop, PGIM's 'buy, let, sell' model means navigating some turbulent and intensely competitive supply and demand dynamics in the race for a return on capital. Over the course of the webinar, the discussion picked up on some key features driving investment – now and for the foreseeable.
Liquidity, liquidity, liquidity
Investment opportunities are currently hampered by liquidity and a lack core capital for take-outs. This is evidenced by few exits (instead there is substantial M&A activity in Europe) but as Paul explained, there are still several options. Debt can be refinanced, or a platform exit may be achieved through the sale of a majority or minority stake. Both strategies effectively recycle capital to another financial investor. Data centre joint ventures, IPOs or Yield Company (yieldco) structures provide other exit routes. Individual data centres can be sold, and some campus assets are engineered from the outset to enable this, but the assets are becoming so significant that this is challenging.
So, what could bring more liquidity to the market? The obvious answer is for interest rates to come down, said Alex, since returns depend heavily on the cost of capital. Even long-income players, currently looking at a powered shell with a 25-year lease in place, are also considering the 20-year gilt, which today trades at 5.3%.
Still, as the market matures and core capital enters the space, greater liquidity and more diverse investment structures are likely to emerge, further institutionalising this dynamic asset class. Core capital groups are the natural off-takers to groups like Gravity Edge – which pursue value-add strategies – or the infrastructure investors currently prevalent at the data centre platform level. A lack of fund vehicle opportunities has created a gap that could also be addressed as the market matures.
Traditional players are focused on highly profitable co-location, but choosing the right tenant is key to future-proofing. Offering investment-grade credit, hyperscalers will sign leases of 10–15 years. The neo-clouds, an emerging phenomenon whose lower-cost graphics processing unit (GPU) infrastructure has generated approximately 700MW of take-up in Europe, tend to sign shorter leases of seven to 10 years while in some cases showing unproven balance sheets.
Picking your moment
Before leases are signed, data centres must be built. Development costs, with their power, plus build plus land numerator, have inflated significantly in recent years, and there is a reduced spread between development yields and stabilised yields. The result is increasing the numerator to justify viability, with rents now significantly surpassing alternative use. There is some nervousness amongst the hyperscalers that the core European DC markets are too expensive. The market needs to further mature, in-line with the US, in particularly North Virginia (NoVA) where major activity continues to support both rental and investment comparables, including Build-to-Suit and powered shell deals. Still, extremely low vacancy rates reveal a stark supply and demand imbalance. As Paul put it: "We can't build these facilities quickly enough."
Those looking for higher returns tend to tolerate more development risk, facing power and permit uncertainty, participating in construction and environmental risks, or bearing the commercial risk of leasing by entering earlier in the lifecycle. Pouring the concrete floors to support thousands of humming servers is not even the costly bit of data centre development. The expensive part is the fit-out, mechanical and operational kit that requires a highly skilled operating partner. It's common for developers to enter into a master agreement with a hyperscaler, ultimately looking to set up a partnership that reaches beyond the single development project. Given the intensive capital outlay and highly specified product, tenants become stickier which results in less churn.
Latency, latency, latency
For those looking for location operators to deploy capital into, the discussion focused on latency and the need to be within close proximity to the network 'edge' to facilitate the communication of packets between the public internet server provider (ISP) networks, and private hyperscale networks. There will always be some degree of latency, exacerbated by distance and capacity, but it is already possible to see how latency has dictated the development of major EU markets over time. Indeed, investors continue to focus on the hearts of these agglomerations of economies, for example, areas such as Slough, London, Paris, Amsterdam and Frankfurt.
Investment strategies usually prioritise locations close to other operators and co-locator gateways, within proximity to the network edge. Positioning data centres within clusters of internet exchange points allows tenants to connect. However, the investment into public and private fibre networks has allowed hyperscalers to extend their reach into (cheaper) non-traditional markets, for example Google with approx. 500MW either in development or in planning in North-East London. It is understood that this is served by a newly established edge connection at Telehouse IXP in the Docklands.
Do you have the power?
If you want power, you have to pay for it. Costs tend to be lower in the Nordics, France and Iberia, so these regions remain attractive for less latency-sensitive applications, such as photo archives or more cost-sensitive occupier groups, including neo-cloud providers.
Investment decisions are most impacted by power supply constraints, however, and governments were called upon to make concerted efforts to update the energy grid. By gobbling up existing power capacity, the data centre market risks effectively ensuring an increased cost for all power users.
"Power has become the number one commodity," Paul noted, and "lack of power is causing enormous challenges". Fresh power applications in Tier one markets (such as London and Frankfurt) often face a wait of six to eight years or more, contingent on infrastructure upgrades. In Tier two markets (including Milan, Madrid, Warsaw, Oslo, and Zurich), wait times are usually reduced but still substantial, often four to six years.
Some investors are prepared to wait for power connections, but others are not, leading to divergent investment strategies driven by location. New markets are emerging where pockets of power are available, including Zaragoza in Spain and Sines, in Portugal. Pockets of power also appear in the north of England and Germany.
Location, location, location
As jurisdictions continue to develop their own data regulation laws, investment will feel the impact of data sovereignty.
In theory, deciding where to site a data centre is effectively a choice of governing framework, although it's not quite that simple. EU data governance requirements, for example, are arguably moving towards companies limiting exposure to non-EU providers and yet currently, Northern Virginia is the largest Tier one market in the world, touching 70% of all global data. In fact, as Alex noted, a massive 90% of EU data is currently held by non-EU providers.
The future role of data sovereignty is further complicated given that we may see a distributed rather than a centralised network in the next five to ten years. The internet was never really able to scale effectively, explained Alex: the original 1980s infrastructure is highly vulnerable, so data remains dependent on private networks. As we move from device-to-server architecture, “reestablishing direct communication device-to-device (peer-to-peer) is essential to keep data safe, and so does it really matter where the data is stored?". Most users of cloud infrastructure are never really sure where their data actually is anyway and, besides, transferring data ownership remains conceptually challenging.
With so much to explore, data sovereignty will be the subject of our March webinar. Transferring data ownership becomes easier when you move the network 'edge' from the IXP to the device – as the device has a dedicated Internet Protocol address – and the data is being transferred IP-to-IP. You can prove ownership if the transmission of data is transacted between counterparties with an address they own. You are essentially removing the middleman, ie no Network Address Translation (NAT) gateways or servers – it's all about scaling the public computer network.
Conclusion
The European data centre investment market stands at a critical juncture, calculating its approach to unprecedented demand, severe power constraints, constrained liquidity and evolving tenant requirements.
As it matures into 2026 and beyond, the conversation highlighted how latency considerations will remain paramount, with land values following power availability. Understanding these basics is essential for well-judged capital deployment. "We have conviction in the fundamentals," said Alex, "hopefully the investment market will follow."
Couldn't attend?
Sign up below to watch a recording of the webinar.
Watch webinar
