I come at digital infrastructure from two seats.
As chairman of Elauwit Connection, I spend my time around the people who design and support networks, and I see how a network performs once it is built and people depend on it. As CEO of a credit firm, I spend the rest of it thinking about what happens to a loan when the plan runs into trouble.
Those two views have started to meet in the same place, and I think the credit market is only now catching up to something operators have understood for a long time.
The size of the buildout
JPMorgan estimates that more than $5 trillion will go into data center and AI infrastructure, along with the power to run it, over the next five years. Annual funding needs run around $700 billion this year and are expected to pass $1.4 trillion by 2030.
The bank expects every public capital market to participate and also points to private credit as part of the answer. Numbers of that size cannot be financed by one kind of lender, which is why so many different pools of capital are working on the problem at once.
Structures that lenders already understand
Fiber networks are further down that path than most people realize.
In March, Surf Internet closed its first fiber securitization at $407 million, and the deal drew demand several times the size of the offering, with interest from insurance accounts and credit funds.
In April, Zayo issued $2.37 billion of asset-backed notes secured by its fiber network assets and the contracts attached to them. It was the company's fourth securitization and pushed its total ABS borrowing past $6 billion. Zayo said the proceeds would mainly repay existing debt and fund network investment, so the securitization market is already doing refinancing work alongside new construction.
I read those deals as evidence that the market has learned how to price contracted network revenue. Cash that arrives under long contracts, backed by physical assets that are expensive to replicate, is the kind of cash flow a credit investor can analyze with some confidence.
Where the risk sits
Even a good structure leaves risk in place, and this sector has its own failure points. Cleary Gottlieb's review of data center financing lays several of them out.
Many projects lean on a single anchor tenant, so that tenant's credit and its termination rights carry most of the weight. Construction delays can trigger lease termination rights and put a backstop at risk.
The equipment inside a facility also ages far faster than the building around it. The facility is built to last for decades, while the GPUs it houses last about seven years on average and possibly as little as three. That mismatch can leave a lender holding the consequences of a decision the equity holders thought was theirs.
Power adds a different kind of exposure. The same review names power availability as the main constraint on development, and it notes that projects started in 2022 took five years to reach commercial operation, against less than three in 2008.
A project can have a creditworthy tenant and a fully funded budget and still wait years for a connection.
Why capital preservation comes first
I come to all of this with a bias toward getting the money back. A lender who protects principal can afford to be patient about return, and a market with this much capital chasing it makes that discipline easy to forget.
The question I find most useful is what the lender holds if the schedule slips or the tenant changes its mind. Collateral that holds its value beyond the original borrower gives a credit room to be wrong, and so do counterparties who can pay.
Operating experience helps here. Networks need maintenance and a team that answers when something breaks, and a lender who understands that will ask better questions of the people running the asset than one who is reading a model.
Credit story, not a technology story
I expect the credit markets to keep finding new ways to finance the networks that everything else depends on. Digital infrastructure has moved from a technology story to a credit story, and the lenders who learn the operating side will be better placed than the ones who treat it as another yield product.
This post reflects my own views on market structure. It is not investment advice or an offer of any security.
