In the past couple of years, artificial intelligence (AI) and the data centers that power it have taken over the infrastructure and energy sectors. What was once a relatively niche market in the telecoms sector, comprising hyperscalers, telecoms, and real estate companies, now encompasses infrastructure fund managers, private equity firms, land developers, power and cooling companies, and others.

Stratospheric market projections for the amount of high-performance computing (HPC) required to power complex AI and cloud systems and the resulting demand for additional electrons has made digital infrastructure a central theme for infrastructure investors and the lenders that support their ambitions. Telecommunications sector transactions reached a record USD 245bn in 2024 and the sector is likely to surpass that in 2025, according to Infralogic data.

Those who stand to benefit the most are first comers, and especially the large asset managers with reach across both the power and digital infrastructure sectors, as well as their lenders, industry professionals told Infralogic.

While experts consulted for this piece warned that there are bound to be small busts along the way, they all agreed that many of those investing in the space stand to benefit from the AI boom.

Market projections for AI infrastructure

While a decade ago, data center power demand was steadily around 200 TWh globally, it’s projected to reach nearly 1,200 TWh by 2030, according to a Goldman Sachs Research study from 29 August. For the US alone, the jump would go from 65 TWh to nearly 600 TWh, according to the study.

All this growth will require massive capital expenditures. Goldman Sachs estimates that in 2025 and 2026, the five highest-spending US hyperscale companies will have a combined USD 736bn of capex. In Brookfield’s Essential Insights paper published last month, the company estimated that AI-related infrastructure spending will be over USD 7trn in the next decade.

Infrastructure and energy players have scrambled to benefit from this boom and the vast opportunities it promises for investment in the real assets needed to power AI. UBS’ Chief Investment Officer Ulrike Hoffmann-Burchardi projected that power and resource companies can expect over USD 3trn in annual investment by 2030 to satisfy the electricity demand from data centers and AI systems. In his recent white paper, Hoffmann-Burchardi stated that AI and the electrification to support it will drive “over 50% of global corporate profit growth in the next decade.”

“Real estate used to be this little niche in this area, fiber infrastructure, cell towers, and data centers. The industry has become so much more, and now there are different levels of risk and different investors to finance each of them,” said an executive at a data center development firm backed by one of the largest equity players in the space. “There is [now] a much broader pool of investors and therefore reward expectations will be adjusted accordingly.”

First come, first profit

Though wary of discussing financial details, many industry professionals gave one answer regarding who was going to benefit from this boom the most: everybody, specifically those who already had their foot in the door.

Pim Rothweiler, managing director and head of telecoms & tech at Natixis, explained that the data center tide is lifting most boats, but experience can give companies a leg up.

It is not just experienced players in the hyperscale and lending spaces that are benefiting; early arrivers in the power space also have a competitive advantage.

Before the rise of AI, data center developers had a relatively smooth process to find a utility to supply power with little infrastructure investment required. Now, many more data center developers and hyperscalers are approaching multiple utilities for countless project proposals. Brookfield’s research estimates that this will cause “development timelines to double and interconnection queues to stretch out three to six times longer in key geographical areas where the digital buildout will occur.”

It caught the utilities by surprise, stated Joan Hutchinson, managing director for off-take advisory at Marathon Capital. Data center developers are dumping interconnection requests for load; utilities, ISOs, and states don’t have a streamlined process to go through these requests.

“Some saw the writing on the wall and were strategic about it, finding appropriately sized parcels of land and making load interconnections early,” added Hutchinson. “That could apply to anybody who had foresight.”

The big dogs

When it comes to making a profit in the data center space, scale is a huge competitive advantage. This can be true of data center developers, hyperscalers, investors, and lenders alike.

Some large asset managers are already seeing returns on their digitalization investments.

During a 2Q23 earnings call, Sam Pollock, CEO of Brookfield Asset Management (BAM), said that Brookfield expected recent data center investments to initially earn single digit yields that will eventually grow materially.

“These [data center] investments are expected to generate high-to-mid teen returns, which could be even higher depending on the success of capital recycling,” Pollock told investors. “We plan on developing almost one gigawatt of capacity over the next three years, which we anticipate will increase last year’s EBITDA by over 5x.”

Not even two years later, in the 1Q25 call, David Krant, CFO of Brookfield Infrastructure, stated that results for the firm were up 5% over the prior year, underlining that one of the key reasons was “strong contracting within our data center business.”

Despite already having substantial data center and power investments, BAM announced that it was launching a “dedicated strategy focused on the development of AI infrastructure” in a letter to shareholders published on 6 August as part of its second quarter results.

Taking the next step in that strategy, the company, through Brookfield Manager Holdings Ltd, registered Brookfield Artificial Intelligence Infrastructure GP S.à r.l. on 25 August.

Brookfield is not the only firm to reap the rewards of a wide-reaching digitalization plan; other large firms are benefiting from engaging with data centers from all angles.

“We’re focusing on data centers, not just through real estate or infrastructure or credit or equity, but across every investment platform at the firm,” said Rick Campbell, senior Managing Director with Blackstone Credit and Insurance (BXCI). “This is how we aim to deliver the best results for our clients. It’s a megatheme for all parts of the firm.”

In July, Blackstone Infrastructure managed funds and Blackstone Real Estate announced plans to invest over USD 25bn in Pennsylvania’s digital and energy infrastructure. The firm backs data center developer QTS, which has multiple sites in the state ready to develop data centers.

Blackstone also backs CoreWeave, a hyperscale company, which has a data center in Pennsylvania, as well. In May 2024, Blackstone provided USD 4.5bn as part of a USD 7.5bn debt raise to help develop the hyperscaler’s fleet.

Fellow alternative asset manager BlackRock also provided CoreWeave with USD 429m as part of the debt raise. BlackRock backs Alliant Energy, which has partnered with Blackstone’s QTS to build a USD 10bn data center in Cedar Rapids, Iowa.

KKR and Blackrock’s GIP together own CyrusOne, a data center provider which they acquired for over USD 15bn in 2022. The duo more recently inked a USD 50bn partnership with Energy Capital Partners (ECP) for data center and power development in 2024. In turn, ECP has a USD 25bn partnership to build Texas data centers with Abu Dhabi’s ADQ.

These major asset managers are stacking up data center investments and joint ventures, while simultaneously backing some of the largest energy and power companies, as well, demonstrating the benefits of supporting AI from multiple infrastructure angles.

“You have people coming from real estate, of power, of gas, of data center development; everybody is bringing their angle. What we have found is that there are very few parties that really bring everything,” said Hutchinson. “Everyone has to realize how big the data center business is relative to the sectors they came from and how much interdependence there’s going to be between all the parties to be successful.”

“We believe the ability to scale or act in size is what people value,” said Campbell. “Companies need more than an answer today, but someone they can grow with. Scale will be a huge competitive advantage.”

The debt picture

On the lender side, data center debt financings are just as large an opportunity as for equity investors, but that opportunity set is shifting.

For many years, tech companies like Meta, Alphabet, and Amazon were putting data center builds on their balance sheets, absorbing land and hyperscale infrastructure costs. While smaller and newer companies and colocation providers were already turning to lenders to secure development debt, these large hyperscalers are now also looking to take this burden off their balance sheet.

“They [hyperscalers] are starting to prefer to have land and data centers banked rather than to hold it on their balance sheet. There will be huge financing opportunities there, but they’re going to be for gold star, top-tier parties,” said Marathon’s Hutchinson.

According to recent JLL data, up to 80% of data center development financing now comes from debt. Norton Rose Fulbright predicted that data center financings were expected to reach USD 60bn by the end of this year, USD 16bn more than in 2024.

Debt issued in the data center space in the US has grown exponentially since 2020, reaching USD 65.25bn issued in 2024, up from USD 2.96bn in 2020, according to Infralogic data. So far this year, USD 56.87bn has been issued, well ahead of last year’s pace at the same time of the year.

According to Infralogic data, the top loan providers in the data center sector in the US are SMBC, SocGen, MUFG, TD Bank Group, JP Morgan, Blackstone Group, Mizuho, CoBank, Natixis, and Santander.

In addition to construction debt, the sector is seeing different types of facilities, including warehouse and holdco deals, according to Norton Rose’s analysis. Market professionals noted that these large transactions can mirror power and renewables project financing structures but also add their own elements.

The deals have been compared to some of the massive, multi-billion-dollar LNG financings seen in recent years and greenfield gas-fired plant financings, which can also require billions of dollars, depending on the size of the plants.

Earlier this month, DigitalBridge-backed Vantage Data Centers was revealed to be in talks with lenders for a USD 38bn financing package to fund data centers in Wisconsin and Texas with MUFG and JPMorgan leading the deal.

“For commercial banks focused on plain vanilla project finance, they are increasingly seeing better value in financing hyperscale data centers than in fully contracted power plants from an economic perspective,” said Ralph Cho, co-CEO of private credit lender Apterra Infrastructure Capital.

However, when comparing data centers with LNG or gas-fired plants, some market professionals noted that data centers have a relatively lower construction risk. With a good location, these assets are straightforward to construct and operate, said the credit asset manager.

Other experts disagreed, stating that data center deals are not always a shoo-in for success.

“Look at the upfront and underwriting fees for banks. They can be super expensive for short tenors and uncompetitive relative to power, midstream, and infrastructure deals,” said a senior manager at a financial advisory firm.

To other lenders, the price for these deals or any premium associated with them is part of the low-risk nature of these deals. It can be called structured-Meta-risk or Microsoft or Google-risk, explained Rothweiler.

That premium, over where the company’s bonds would trade, or how Meta or one of the other hyperscalers would finance themselves, is the margin that lenders are looking at, according to Rothweiler.

Opportunities around data centers

As tech giants, developers, and others pour billions into AI infrastructure, adjacent infrastructure, like liquid cooling systems, present further opportunities for investment.

“If you think about data centers, as a lender, one of the neat parts is that if it costs one dollar to build a data center, there’s roughly three dollars of equipment that goes into it depending on the facility,” said Blackstone’s Campbell. “It’s not just about lending to the data center itself. There are picks and shovels around it, with very high-quality counterparties providing equipment, and O&M contractors and other key providers.”

One reason for the rise in adjacent opportunities is the growing size of AI projects requiring more extensive GPU systems (graphics processing units) and, in turn, requiring more chips and more advanced cooling systems, according to the 8 September North America Data Center Trends H1 2025 report from CBRE.

Another, the study noted, was the proliferation of legislation limiting water use for data centers, prompting more manufacturing improvements to advance cooling systems.

“We view construction risk, with the right people and the right contractors, as relatively low compared to building other types of assets,” explained Campbell.

Hoffmann-Burchardi endorsed this idea in his recent white paper, advising investors to adopt a more balanced and diversified approach across both the data center and AI value chain, including AI laggards like internet and software companies.

BAM announced during its recent earnings call that it is now focusing on investment opportunities near data centers, such as fiber networks, manufacturing, recycling, and liquid cooling systems.

“If, for whatever reason, a year from today, the relative value is not there, then we are still somewhere in the ecosystem benefiting from this megatheme,” said Campbell.

A hyperscaler bubble?

While first-comers, large-scale equity investors, and lenders are predicted by many industry professionals to ‘win’ the AI infrastructure race, hyperscalers are not guaranteed success just because of their size today. AI technology companies have been making headlines recently for not turning a profit.

The New York Times dove into how most AI tools are currently not profitable, but will have to generate large cash flows to recoup their developers’ investments. Futurism noted that a McKinsey report from this year revealed that AI data centers will need to spend USD 6.7trn to keep up with demand.

Even OpenAI CEO Sam Altman compared the AI boom to the late ‘90s and early 2000’s dot-com bubble in an interview with the Verge.

“Part of the calculus is recognizing that even the largest hyperscalers can lose market share,” said Apterra’s Ralph Cho. “History shows how quickly giants like Yahoo and Netscape fell out of favor. These firms are doing everything possible to defend their competitive edge.”

While it remains to be seen who will come out on top at the end of the data center boom and which investments will be the most profitable, one group that is clearly seeing the cost burden is the ratepayers, especially in areas where data centers are or will be concentrated.

Due to the amount of power needed to support data center growth, experts estimate that wholesale electricity prices could rise by an average of 8% by 2030, according to a study by Carnegie Mellon University, North Carolina State University, and Sutubra Research from June 2025.

In Northern and Central Virginia, the country’s data center alley, prices are expected to rise by 25% in five years, due to the immense demand for 100 TWh of electricity and the cost of keeping 25 GW of coal plants that should be retiring, according to the study.

Keep dancing while the music’s playing

Although not all AI companies can come out on top, market participants remain confident in executing mega deals with these hyperscalers.

“From a lender’s perspective, you’re essentially advancing against long-term rent payments from investment-grade tenants – high-quality companies that are locked in for 20 years,” said Cho. “It’s a bet on a very low probability of default.”

Just as the hyperscalers are looking to top-tier banks for financing, so too are banks looking for hyperscalers that have the best track record.

“We have to make choices as well. Capital is ultimately finite, whether it’s bank capital, private capital, or anything in between,” said Rothweiler.

Hyperscalers are already valuable companies, so it’s unclear if you can call them winners in this boom; they’re already winning, explained Hutchinson. “What we’re betting on in this industry is the credit of hyperscalers such as Amazon and Microsoft. So far, I don’t think there’s a reason to be concerned. But if there is a disruption, we’d be very exposed, because it’s only a handful of parties who are bringing significant credit to the space.”