The private credit sector should focus on power generation as the most advantageous way to engage with the ongoing data center construction boom.

That’s the view of Nik Nunes, the CEO of Industrial Development Funding (IDF), an arranger of deals for major lenders in the energy transition and digital infrastructure sectors.

Nunes, a veteran of GE and Perella Weinberg, said the strategy stems from a recognition of utilities’ and independent power producers’ (IPPs) financial constraints due to capital expenditure requirements that are going ballistic to meet data center demand. Financing GPUs (or financing the purchase of NVIDIA chips for cloud platforms), while still part of IDF’s strategy, faces more risks due to a fast-moving technology cycle.

“Our core interest is actually helping utilities stand up to assets,” Nunes said in an interview with NPM. “That’s a big space where we think we can channel insurance capital from Blackstone, where they manage USD 400bn and we have a partnership.”

For example, utilities like Entergy, which has increased its capital expenditure plans from USD 12bn to over USD 30bn over four years, face limitations that prevent them from funding large projects outright – or through corporate debt – without risking credit downgrades, Nunes said.

“Utilities are all struggling with how to fund what are three-to-four-year construction periods during which time they run the risk of being downgraded if they make investments,” Nunes said.

IDF is also focused on distributed power solutions that bridge interconnection delays for data centers and other medical or manufacturing facilities, Nunes added, “where the grid just can’t get there.”

“ They don’t have the generation or the transmission to get to some of these places,” he said.

Fundless sponsor

IDF employs a fundless sponsor model, enabling it to structure bespoke financing solutions without raising traditional closed-end funds. Instead, IDF partners with infrastructure credit and equity platforms, such as Blackstone and HPS (now part of BlackRock), and financial institutions like MUFG and Mizuho.

For power generation deals, IDF targets returns of 6% to 8%, leveraging the investment-grade creditworthiness of utilities and IPPs. GPU-related deals, by contrast, demand higher returns due to inherent risks, with structured financings often yielding mid-teens and occasionally exceeding 20% when profit-sharing is included.

IDF collaborates with Mitsubishi Power to structure financing deals that enable utilities and IPPs to acquire turbine equipment under long-term payment plans. It has done over USD 800m of these deals for Capital Power, J-Power, and Iberdrola, the parent company of Avangrid Renewables, and is focused on expanding its deal flow with additional utilities, Nunes said.

The firm receives a retainer from Mitsubishi and Blackstone for ongoing deal structuring and origination, and takes a fee tied to completion, effectively “being levered to the back end” of the transaction, he added.

‘Structural arbitrage’

In collaboration with HPS, IDF recently acquired nearly USD 500m of Bloom Energy assets, including 60 MW of distributed power generation systems used by data centers and industrial facilities. The assets are supported by investment tax credits as well as long-term power purchase agreements (PPAs) and investment-grade customers like Intel and Equinix.

While the acquisition of distributed power generation systems might appear to be a departure from the fundless sponsor model, for Nunes it represents the flexibility that can be built into IDF’s financing structures.

Bloom was looking to sell the portfolio, while HPS was interested in lending to it, so IDF stepped in as the owner. “The lender wants to lend us a loan, the seller wants to sell us a receivable. We fill the gap,” he said. “It’s a structural arbitrage.”

IDF plans to execute more of these deals with Bloom in the future, he said.

More cautious on GPUs

Central to IDF’s strategy in the digital infrastructure space is that it can provide capital that bridges the gap for newer entrants in the GPU space that do not qualify for the traditional bank market and may not want to raise dilutive, and therefore expensive, venture equity.

But while IDF sees substantial opportunities in GPU financing, it remains cautious due to technological obsolescence risks. Nunes noted the rapid pace of innovation, particularly with NVIDIA, which releases new chips every six months.

Even so, IDF has participated in GPU financing, notably through its deal with Lambda, which involved a Macquarie Group-led special purpose GPU financing vehicle of up to USD 500m to fund the expansion of its cloud offering.

At the time of the deal, Lambda had only raised USD 40m of debt financing from a lessor of some of its GPUs, would not qualify for the traditional bank lending market, and so likely needed to raise capital in the form of venture equity, according to Nunes. By structuring asset-backed financing, IDF bridged the gap between venture equity and traditional bank loans.

“Our thesis is that there’s something between venture equity and bank debt that can get comfortable with at least the underlying asset, which is an NVIDIA GPU that’s very valuable,” he said. “It may not have 10 years of life, but it’s got at least four or five years of life.”

Additionally, not all of the new entrants’ contracts are with hyperscalers or investment-grade counterparties, and many of the commercial arrangements are shorter term and on demand, he said. So how did they get comfortable with it?

“You’re really taking a view on cash flow-based lending for the use of those assets that banks will never get comfortable with,” he said.

Nunes said he is starting to observe some frothiness in the market for GPU lending: Lessors are starting to write paper backing GPUs at between 12% – 14% target returns with no profit sharing, and also lending against the potential of the chips having value beyond three to four years, even though NVIDIA recently shortened its product vitality cycle.

“What I’d rather own is the power block that is needed because those GPUs and those data centers are going to need a lot more power, which has some ability to pivot and sell into the grid or whatever else,” he said.

“To go six to 10 years long on a block of GPUs is a little scary,” he said, “versus six, 10 years even 15, 20 years long on a power block for the same GPU data center feels a lot safer.”