The creation of AI, including data centers and supporting infrastructure, creates one of the biggest financing opportunities in today’s markets. But investors must navigate growing competition, supply constraints and concerns about non-performing assets as private debt establishes its role in the sector and capitalizes on the $700bn opportunity. Aysha Gilmore reports that…
The push for Artificial Intelligence (AI) has grown rapidly and has led to greater concern for software companies – a staple of private equity portfolios. And while much of the funding for AI-related businesses has gone to the public markets, private credit managers are eager to get in on the action.
But to do that, investors need to navigate increased competition, supply constraints and problems related to non-performing assets.
The huge amount of money needed to support the development behind AI means that there is a big enough pie for everyone to grab a piece of.
The development has been a key topic in credit markets for the past six months, as “the AI capex cycle is in full swing,” according to Mahmoud El-Shaer, fixed portfolio manager at Wellington Management.
No longer are tech giants like Google and Meta investing in AI from their own profits alone. Increasingly, they are turning to external financing, with large debt issuances and even equity issuances from hyperscalers.
Meanwhile, many private equity market managers are building investment strength through their infrastructure arms. Last month, private equity firms Apollo and Blackstone teamed up with Broadcom to launch a $35bn (£26.2bn) AI infrastructure platform. In total, the platform is expected to provide 20 gigawatts of computing power for AI labs through 2028, the firms said.
KKR launched Helix Digital Infrastructure, a new company formed in June to fund AI infrastructure and other alternative asset managers, including Ares Management, have set multibillion-dollar fundraising goals for data center investments.
However, this funding is still only visible. According to a Morgan Stanley study published in May, global data center capex (excluding power) is expected to exceed $3.2tn until 2028. Of this, approximately $1.75tn is expected to be financed through debt products, with private debt accounting for approximately $700bn.
A growing field
AI infrastructure investment remains a “nascent arena” for private debt, explains Anant Kumar, global investment strategist at Benefit Street Partners.
Currently, most funding comes from government markets, including high-yield and investment-grade bonds, with private markets still too small to accommodate some of the larger deals. However, capital is beginning to flow into the asset-based finance (ABF) market, particularly where borrowers seek flexibility and speed, areas that play to the strengths of private credit.
Kumar narrates Another Credit Investor that Benefit Street Partners, Franklin Templeton’s private equity arm with assets under management (AUM) of $93bn), is currently lending to data centers within AI infrastructure as part of its ABF portfolio, and is looking at chip and turbine deals.
He explains that over the past six to nine months, the business case for AI has become “much clearer”, with value now being seen in workplaces and return on investment more visible.
“I think one of the fears is that this is all a bubble, like the telecom bubble, that we’re building all this capacity before the bubble bursts. I don’t think this is the case,” Kumar said. “Right now, the construction of AI that is happening is fully implemented – users are testing tokens and data centers are fully operational. Unlike the telecom bubble, it is not like you are building a lot of predictive power. For this reason, many companies will collect money from all available sources – investment grade, high yield, direct ABF lending.”
Increased competition
However, despite strong demand and a compelling investment case for AI infrastructure, the growing data center market and declining risk profile raise questions about whether private debt can compete with banks in financing the sector. Along with this, that developing foundations can end up causing overcrowding.
Joel Holsinger, head of alternative credit at Ares Management, recently told ACI that within the ABF firm’s portfolio, despite these areas still being attractive, “digital infrastructure and data centers are approaching the end of their relative value for us”.
This is largely due to the large amount of money that is already flowing into the data center space, especially on the lending side, and banks starting to operate, especially in the US, he says.
“Banks are building capacity in the data center space, so there’s a lot of concentrated risk,” Holsinger said.
AI is increasingly dominating venture capital, accounting for nearly a third of funding in the property sector in the UK and close to three-thirds in the US, with a few mega-rounds driving a growing share of overall investment.
Read more: Private credit could get a ‘boost’ with higher rates despite software concerns
SRT deals
However, Holsinger says that while the bank’s involvement makes digital infrastructure less visible among ABF’s holdings, it opens up deals within the risk transfer (SRT) market. As banks seek to increase lending for AI infrastructure, they are also looking to offload exposure to data centers.
“We see secondary derivative interest as freeing up capital in that data center segment, but primary derivative is less interesting to us,” Holsinger said.
Roopa Murthy, head of EMEA infrastructure credit at Ares, added that the opportunity is becoming more prominent in Europe, where regulatory pressure is growing stronger.
“We’ve seen banks struggle to focus on this space, and they’re looking at new ways to reduce exposure,” he said. “This includes working with private mortgage lenders through SRT-style structures.”
While the US market may be oversaturated, European data centers and related AI infrastructure still require “significant” amounts of private debt investment, said Anne-Laurence Roucher, head of the private markets group at Edmond de Rothschild.
“Half of the world’s data centers are located in the US, and the US only has about 4 percent of the world’s population,” he said. “So Europe is likely to continue to invest heavily in data centers, as well as power generation, given its significant energy needs.”
Bubbles and bumps
Indeed, the supporting infrastructure, especially power and network, is central to the investment case for data centers, and without it, the investment case becomes very weak.
Kumar notes that concerns about an AI bubble are less pressing than the structural issues facing the sector.
“I’m not worried about important prospects; I’m worried about the ability of border labs to get access to the computers they need,” he said.
Beyond these constraints, there is also the risk of technological obsolescence, said Roucher, Edmond de Rothschild has allocated digital infrastructure through a portfolio of infrastructure loans.
Data centers require upfront investment, yet the rapid pace of innovation in GPUs and computing architecture raises the risk of non-performing assets for investors, he says.
“There has been an increased risk of technology and infrastructure over the last decade as AI has become more connected to data centers,” Roucher explained. “You need more data center capacity for AI, and you need more power for AI.
“Therefore, the loan provided can be temporary to ensure that it covers the risk of default as it is quickly repaid.”
Another key bottleneck lies in the supply of hardware, while the semiconductor supply chain remains highly concentrated, with companies such as TSMC producing the bulk of advanced chips. The main risk here is that any country disruption can have a very negative impact on AI and data center development.
Additionally, since data centers rely heavily on water for cooling, resource shortages are an added factor in expansion and a major risk for investors.
Weather hazards
At the same time, the use of electricity and water, as well as construction work, puts a great strain on the environment. And despite the stormy climate agenda in the US and elsewhere, participating in the creation of AI infrastructure could pose challenges for climate-focused investors like European asset manager Ambienta.
“Thankfully it’s a growing sector, but we need to link that growth with environmental sustainability,” said Ran Landmann, head and chief investment officer of Ambienta’s credit strategy. “In fact, the growth of AI and data centers does not address that.”
Still, Landmann tells ACI that Ambienta is exploring ways to invest in AI development. While they are not currently investing directly in data centers, they are exploring opportunities across ecosystems, he says.
“It’s a growing industry, and we need to find an intersection in terms of how our companies play an active role in this massive growth, while also participating in environmental solutions that make it sustainable,” added Landmann.
“We are touching that field, as you build more data centers, there are technologies that support this construction to be sustainable. Examples are companies that provide solutions in cooling technology in data centers, an area that has been identified as a growth area for us.”
Read more: Executives are hunting for software winners and losers amid the AI scare
Skepticism and market selling
Despite the strong momentum, questions remain about whether AI expectations have been overstretched. At the beginning of June, markets experienced a sell-off caused by weaker-than-expected results from semiconductor company Broadcom, which failed to meet expectations for AI chip demand.
This has triggered a broader tech trend, compounded by macroeconomic factors including stronger-than-expected US employment data. Although markets have recovered somewhat, the episode highlighted existing concerns about the sustainability of AI-driven growth.
Kumar, however, dismisses the volatility as temporary, suggesting that “there will always be pullbacks as nothing goes up automatically”.
“If I take one part of the semi space, which is memory, the demand for memory is supported for a period of many years and I don’t think that the sale of one day affects that thesis,” he said. “Therefore, I would argue that this daily fluctuation is noise.”
At the moment, the consensus seems to be that AI-driven decline is unlikely in the near term. And with forecasts estimating that private debt needs to provide $700 billion to build it is inevitable that the industry will reinvest in AI infrastructure and its supporting backbone.
This article originally appeared in the July issue of Alternative Credit Investor, to view this issue click here.
