Keen to capitalise on the rapid expansion of artificial intelligence, firms are weighing up how best to get data centre exposure while managing long-term risk.
By Jack Arrowsmith, London
In its Q1 earnings call, Blackstone CEO Steve Schwarzman declared that the firm believed it has become “the largest investor in AI-related infrastructure in the world”.
And it’s not just the world’s largest alternative asset manager: investors across private markets are racing to meet the capacity demands of tech companies.
The facilities offer compelling trends which suit the investment needs of firms looking to commit capital for the long-term.
“In Europe we think the market will grow by six-fold in the next seven years,” says Rafael Coste Campos, a partner in the real estate team at Bain Capital.
According to data from McKinsey, global AI-related data centre demand could increase 3.5x on 2025 levels by 2030. Backed by the major hyperscalers, these investments offer firms a durable long-term tenancy.
“A typical lease with a hyperscaler is 20-25 years, fully guaranteed. Then you have a bond from one of the strongest companies in the world,” he says.
Bain Capital’s exposure includes Bridge Data Centres, a platform operating across southeast Asia which has scaled from 60MW to almost 2GW under the firm’s ownership.
But the unprecedented levels of energy demand from the facilities mean they carry their own risks for investors.
In its base scenario, the International Energy Agency (IEA) projects that global electricity consumption for data centres will reach 945 Twh by 2030, up from 415 Twh in 2024 and comparable with all of Japan’s energy consumption today. Meeting that demand is a formidable challenge.
“One of the greatest obstacles for new developments in core markets is grid connection. Obtaining new power connections can take years,” says Steven Zehr, a partner in the infrastructure team at Stafford Capital Partners.
“When we underwrite a deal, we want to see that access to power is taken care of, and connectivity is strong,” he adds.
Limited energy availability, coupled with ageing grid infrastructure and complex planning processes mean projects can be held up by long lead times to get online, especially in Europe.
Grid connection in the data centre hubs of Frankfurt, London, Amsterdam, Paris and Dublin (collectively known as FLAP-D) can take an average of seven to ten years according to the IEA.
And with demand as high as it is, pricing is an inevitable headwind for expansion.
Watchdog Monitoring Analytics found that for customers of PJM Interconnection, the electricity provider covering over 67 million people in the northeastern US, total cost of wholesale power increased 76% in Q1 2026 compared to the previous year.
This coincided with a total cost of capacity increase of 398%, with the watchdog adding that data centre load increased customers’ bills by $13.8bn.
This means hyperscalers are seeing their own costs rise, and may ultimately face pressure to compensate consumers for rising prices attributed to data centre expansion.
The financial heft of these entities is likely to help in absorbing these costs, given how central the infrastructure has become for their operations.
And if grid access is achieved, the investors stand to benefit from the supply challenges.
“Once you have very strong demand on one side and supply remains structurally constrained, you see durable earnings growth for the asset owner,” he says, adding that the extent of the supply challenges in Europe make the risk of oversupply negligible.
The NIMBY challenge
There is the additional risk, however, that mounting public opposition could see the facilities being required to source their own power.
In Ireland, new data centres which have a maximum demand from the grid of 10MVA or more (approximately 8–10MW) must provide power generation and/or storage capacity to match.
While this does not yet amount to full self-sufficiency – the facilities can continue to draw power from the grid – it does signify a shift towards placing more burdens on the data centres themselves.
“There is a fundamental difference in the cost of capital across the AI infrastructure value chain. Someone building a data centre is hoping for a mid-20s IRR, while someone investing in the energy infrastructure will see a low double-digit return,” says Coste Campos.
He adds that he expects data centre capacity will still be able to grow from European cities which are expanding meaningfully, such as Madrid and Milan. The cities are not currently mandating self-sufficiency on the part of new data centres.
Local opposition can also restrict the size of new projects. The proposed $100bn Stratos data centre in Box Elder County, Utah, led by investor Kevin O’Leary, has been at the epicentre of mounting backlash.
Following extensive public pressure and an intervention from Utah Senate President Stuart Adams, who demanded that the project be shrunk by 75%, O’Leary has agreed to scale back the planned 40,000-acre facility to around 20,000 acres.
While the facility is arguably an outlier – it would have been one of the largest data centres in the world – it serves as an example of how local populations are increasingly looking to oppose data centre expansion.
To mitigate these risks, some investment firms are turning to smaller scale solutions. Wafra, the US-based alternative asset manager owned by Kuwait’s public pension fund, is pursuing the construction of edge data centres.
“As the data produced by large scale data centres is delivered across networks, you need to have assets that are closer to population centres to allow for the balancing of traffic and for latency sensitive workloads to be completed closer to the end user,” explains Michael Coleman, managing director in the firm’s real assets and infrastructure team.
These smaller scale assets can afford to be less intrusive than the larger projects, and easier to integrate into existing energy infrastructure. Wafra recently set up Beyond Data Centres with digital infrastructure developer Tony Rossabi, to build these facilities.
Coleman says that for the new platform “these edge data centres are generally below 10MW, so the workloads are smaller and it is easier for them to fit into the grid.”
“They can more or less fit in existing industrial parks, meaning they don’t need to be in people’s backyards,” he adds.
For firms accustomed to mid-market ticket sizes, this also offers a more diversified approach than investing in a single large data centre.
Coleman says that “building ten 10MW facilities instead of a single 100MW data centre avoids concentrating risk in a single asset”.
The adjacency funnel
The risks associated with investing in data centres mean some firms are turning to adjacent industries which stand to benefit from the boom.
Real estate investor Heitman is managing industrial land located near data centre developments to support their construction and maintenance.
“Across our industrial portfolio, we’ve been doing a lot of leasing to data centre buildout companies. We’re currently investing in industrial outdoor storage across Virginia,” explains Dan Vickerman, head of commercial investment research at the firm.
“The capital expenditure here is much lower than for a data centre, and the use of the sites is flexible,” he says, meaning it can be repurposed for other use cases, should the pace of the data centre buildout begin to slow.
Adjacencies can also be used to support the evolving energy requirements of the sector.
“The only answer to [the energy generation] problem is going to be nuclear, through a combination of existing large-scale reactors, and small modular reactions,” says Michael Lustbader, managing partner at Arlington Capital Partners, a US firm which invests in government-regulated sectors such as defence and technology.
In April the firm acquired ENERCON, an engineering company which focuses on nuclear, which it merged with existing engineering asset Pond & Company.
It sees the investment as a way to benefit from the broader expansion of data centres, by serving the nuclear energy assets which could be their main power source.
“During a gold rush, you know that the picks and shovels are going to do well,” he says.
“We don’t want to have to pick an individual data centre, but we know that they are all going to need power.”
As nuclear continues to develop, the technology could also be used to help data centres themselves become independent from the grid.
Small modular nuclear reactors (SMRs) are one potential solution. The small scale reactors could sit alongside a data centre to power it, removing the risks associated with connecting to the grid.
In Europe and North America, the technology is yet to reach commercialisation. The first SMR units are expected to come online in the US in 2030-2031, with additional deployments through the mid-2030s.
But as the data centre buildout faces constraints, private markets will need to continue innovating to ensure they can capture the alpha from the rising demand for AI technologies. With a menu of investment options before them, an assessment of long term risks becomes more vital than ever.
As data centres hits constraints on earth, space presents another option. With greater access to energy, cooler temperatures, and more room than our planet, it could be a valuable store of compute power. We discuss this on Stories of Transformation with NewSpace Capital CIO Felix von Schubert.
