The largest borrower in Asia’s data center build-out is not a technology company. It is a private equity firm’s real estate arm.
Bridge Data Centres, owned by Bain Capital, is in talks to expand an existing $2.8 billion loan to roughly $6.6 billion, Bloomberg reported on October 12. The money would fund more machine capacity across the region. If it closes at that size, the borrowing would rank among the largest ever raised for data centers in Asia.
The structure tells the story of where this build-out is being paid for. The companies training and serving AI models no longer carry all of the construction bill on their own balance sheets. The debt now runs through banks, private credit funds and specialist landlords, and the entities signing the papers often have names that would not appear in an AI press release.
Bridge is that kind of entity. It builds and runs the shells — the land, the power, the cooling and the racks — and leases them to customers who fill the space with servers. Bain’s ownership means the capital behind the expansion comes from a fund that expects a return, not from a cloud operator’s cash flow. Once a single loan reaches $6.6 billion, the terms matter as much as the capacity it buys, because the interest bill has to be covered before any of the underlying tenants turn a profit.
That arithmetic is what makes the deal worth watching. Data centers are long-lived assets financed on shorter paper. A lender who commits for a decade is betting that demand for computing power stays strong for at least that long, that power contracts hold, and that the tenants signed to fill the halls remain able to pay. Each of those assumptions has held so far. None is guaranteed by the technology itself.
Private credit has become the natural home for this kind of borrowing. Banks have limits on how much exposure they will take to a single sector, especially one whose revenue depends on capital spending that has not yet shown a stable return. Credit funds face fewer such caps and can price the risk higher. The trade-off is that the cost of the money goes up, and the total repayments that a data center landlord must clear in the first years grow with it.
The scale of the regional build-out gives the talks their edge. Asia’s demand for AI computing has risen on the same curve as the rest of the world, but the region’s power grids, land markets and permitting systems move at different speeds. Building capacity in a market where electricity is scarce and approvals are slow costs more than building where they are not, and the financing has to absorb that gap. A lender underwriting $6.6 billion is underwriting the completion schedule as much as the site plan.
For Bain, the expansion is a vote on the durability of the AI spending cycle. The firm bought into Bridge because it believed demand would outrun supply for years. Doubling the borrowing on that bet raises the return if the belief holds and the loss if it does not. That is the ordinary shape of private equity, applied now to a sector whose customers are spending at a pace with no precedent to measure against.
The signal for the wider market is the direction of the money. When AI infrastructure was a story about chip orders, it showed up in the earnings of a handful of listed companies. Now it shows up in syndicated loans and credit funds, in the filings of builders most investors have never heard of, and in the borrowing costs those builders pay. The financing has moved downstream from the labs to the landlords, and the landlords are only as safe as the tenants they can sign.
The regional picture sharpens the point. Data center borrowing in Asia has grown quickly, but it has grown from a small base, and the lenders entering now are not the lenders who funded the last cycle. Insurance companies, credit funds and sovereign investors have replaced or joined the banks, drawn by yields that look generous next to government paper. Whether those yields are actually generous depends on defaults that have not happened yet, in a sector that has not yet been tested by a downturn.
There is a limit to how far the debt can run ahead of the demand. A data center financed before a tenant is signed is a bet, and a data center financed at $6.6 billion before its tenants are signed is a large one. The industry has so far avoided the empty-hall problem because compute has been scarce. If capacity catches up with demand, the borrowers who expanded fastest will be the ones who feel it first.
Bloomberg reported that the talks are still under way, and a loan of this size can change shape or stall before it is signed. What is less uncertain is the pattern the talks confirm. The build-out that began as a technology story has become a credit story, told through the balance sheets of firms that own the buildings, not the models. The next phase of the AI boom will be judged in part by whether those balance sheets can carry the weight.
