Data centres have swiftly become one of the most capital-hungry asset classes in the global investment landscape.
That transformation was on full display at the Datacloud USA Keynote Panel on ‘The Evolution of Data Centre Financing,’ where the panellists discussed how the industry has shifted from hundreds of megawatts to gigawatt-scale campuses, and from $200 million projects to multi-billion-dollar developments.
Hosted by Kemal Hawa, partner at Kirkland & Ellis LLP, the panel featured a cross-section of stakeholders in the sector: Kevin Mammel (CFO, Rowan Digital Infrastructure), Brian Pryor (MD, Houlihan Lokey), Felix Zhang (partner, Ares), Joan Hutchinson (MD, Marathon Capital), and Cynthia Thompson (co-founder & chairperson, Cloudburst Data Centers).
Together, they unpacked how capital structures are evolving, what private credit and ABS markets can (and can’t) support, and how the rise of “Neo Cloud” clients are shaking up underwriting norms.
Kevin Mammel opened with a striking observation: “A couple years ago, we were doing 20–30 megawatt deals — $200 to $300 million projects. Now we’re doing gigawatt campuses with multi-billion-dollar financing arrangements.”
The scale has changed dramatically, but capital markets are still catching up. Felix Zhang from Ares noted that in 2021, just $10 billion in data centre asset-backed securities (ABS) were outstanding. Today, that figure has ballooned to $50 billion, a fivefold increase. Yet even that growth is only scratching the surface.
“If you consider the $3 trillion in capex needed globally over the next three years, ABS will fund only a sliver of it,” Zhang said. ABS markets, while growing, are mostly applicable to stabilised, leased assets, leaving a gap in financing for pre-lease or early-stage developments.
That’s where private credit is stepping in.
“At Ares, and among private credit providers, we’re plugging the gaps, providing capital before monetisation, at either the corporate or asset level,” Zhang explained. The question now isn’t whether capital exists, but whether the market is deep enough to recycle short-term construction loans into long-term solutions.
Banking bottlenecks and the rise of private capital
According to Joan Hutchinson, the scale of financing now required is pushing traditional lenders to the limit.
“We’re underwriting $60 billion annually in construction debt. Last year, only about $9 billion in ABS was issued,” she noted. “There’s a liquidity issue in the commercial banks, and that $60 billion needs to be taken out and refinanced.”
This gap is driving a critical need to develop new markets for post-construction refinancing. Private capital, insurance firms, and new ABS teams are entering the fray, but they’re moving cautiously.
Cynthia Thompson of Cloudburst noted that much of the early-stage development is still funded through equity or creatively structured joint ventures. “We’re seeing power/data centre duos forming, bringing in partners from oil and gas or real estate,” she said. “Sometimes these investors provide ‘risk dollars’ – expensive, but critical for early stages.”
Still, that risk capital remains hard to source. “You can’t expect the ABS market to fund projects with no tenants signed,” she said. “Not yet.”
Neo cloud clients: new tenants, new risks
Another pressing issue is the influx of so-called “Neo Cloud” clients who are purchasing at hyperscale but don’t carry the same credit ratings as the traditional tech giants.
Brian Pryor identified this as one of the most pressing challenges for developers, “Neo clouds have insatiable demand but lower credit quality. The financing markets are still trying to figure out how to underwrite them.”
He cited CoreWeave as the leading example. Now a public company with a $55 billion market cap and high-yield bonds rated B1, it’s become “underwritable” as a single-tenant customer, albeit at higher cost. But others in the category haven’t reached that level of transparency or scale.
“This will be the last financing product to develop,” added Zhang. “A single-asset financing for a non-investment-grade tenant is possible, but it will come at the highest cost. Lenders prefer portfolios and diversity.”
Long lead equipment: A $100 million head start
Another pressure point is the financing of long-lead equipment, such as transformers, GPUs, and switchgear, which can carry lead times of up to 36 months. For developers, it’s a chicken-and-egg problem: secure the gear without a signed lease, or risk losing time-to-market.
Thompson said Cloudburst tackled this by embedding financing considerations into their founding strategy. “We curated our investor base from the start around the suppliers of those long-lead items,” she explained. “To be taken seriously by hyperscalers, we had to show we had gear in hand.”
Mammel echoed that sentiment. Rowan Digital began with a strong equity backer in Quinbrook Infrastructure and initially pursued “power shell” deals, minimising the equipment it needed to buy. With their first lease stabilisations in place, they’ve now raised $300 million in holdco-level financing to fund future projects.
That model is becoming increasingly common. “In some cases, tenants even bring their own inventory,” noted Hutchinson. “That can be the deciding factor in choosing one customer over another.”
What’s next?
As gigawatt-scale developments become the norm, the financing toolkit will have to keep evolving. The panellists agreed: more joint ventures, creative capital structures, and diversified investor pools will be needed to match the sector’s explosive growth.
“The market is going to figure it out,” Mammel said. “Just like Facebook figured out how to monetise eyeballs, the Neo Clouds will figure out how to monetise AI demand. And the capital will follow.”
RELATED STORIES
