Blue Owl expands AI data centre lending
private credit shifts from software loans to power-hungry infrastructure, public backlash over electricity bills becomes a balance-sheet variable
Images
Blue Owl Capital has stepped deeper into the business of financing AI-era data centres, placing at least three large loans tied to facilities in the US as demand for power-hungry computing accelerates. Business Insider reports the alternative asset manager recently agreed to provide $240 million for a Minneapolis data centre bought by Cloud Capital and Arcapita Group, and is also set to both invest in and lend against a new large campus in Wichita Falls, Texas being developed by Skybox Datacenters.
The loans sit alongside a previously disclosed $500 million credit exposure connected to a data centre being built in Lancaster, Pennsylvania for the AI cloud provider CoreWeave. In aggregate, they show how a firm that built its reputation in private credit is now using debt as a second channel into “digital infrastructure” — a category that has become Wall Street shorthand for data centres, grid connections, and the land and equipment that turn electricity into rentable compute.
The shift is not happening in a vacuum. Blue Owl raised nearly $10 billion last year for digital infrastructure, much of it for equity stakes, and has talked publicly about scaling its data-centre footprint. But the same quarter that saw the firm expand in data centres also brought higher client withdrawals from two of its credit funds, Business Insider notes, partly because those funds hold loans tied to software companies. As investors reassess which parts of the tech stack are durable, the balance is moving from “apps” to the physical assets that run them.
That rotation comes with a political and operational constraint that the loan documents cannot solve: power. OpenAI chief executive Sam Altman told BlackRock’s US Infrastructure Summit that “AI is not very popular in the US right now,” citing public anger about electricity prices and layoffs that companies attribute to automation. When data centres are framed as the reason household bills rise, grid upgrades and new generation become permitting fights, not just engineering projects — and the timeline risk lands on the same lenders who rely on predictable cashflows.
For financiers, the attraction is clear. Data centres can sign long-term contracts with large tenants, and the build-out is large enough to absorb billions in capital quickly. For communities and ratepayers, the costs often appear indirectly: grid reinforcement, congestion management, and the political pressure to treat power supply as an emergency public good when private projects outpace local capacity.
Altman’s warning about “headwinds” points to the next pricing problem. If public opposition slows interconnection and construction, lenders must decide whether to demand higher spreads, tighter covenants, or more equity — or whether to assume that utilities and regulators will find ways to keep projects energised because the broader economy now depends on them.
Blue Owl declined to comment on the loans. The financing, however, is already public enough to show the direction of travel: the AI boom is being underwritten not only by chipmakers and cloud providers, but by private-credit desks betting that electricity and concrete will outlast software cycles.