Data-Center Delays Put Highly Leveraged AI Firms Under Pressure
Anna Rathbun, founder and CEO of Grenadilla Advisory, argues that political resistance to data-center construction could impose financial discipline on a buildout driven by companies borrowing and building to secure capacity. New oversight or limits could slow expansion and reduce the risk of overbuilding, she says, though she expects construction to continue at a slower pace. In an interview with Bloomberg Tech’s Tim Stenovec, Rathbun said the market selloff may create opportunities in financially stronger AI-related companies, while firms dependent on data centers becoming operational face greater risk from delayed revenue.
Political resistance could slow a race to build at any cost
Anna Rathbun sees a possible benefit in political pushback against data-center construction: it could bring financial discipline to a buildout that, in her account, had few hurdles. The political shift Stenovec described as a “blue tsunami” raised the prospect of new oversight and regulation. Rathbun named possible legislation, moratoriums, and the loss of tax breaks as hurdles that could disrupt an expansion investors had treated as uninterrupted.
Before that possibility emerged, she said, companies were borrowing and building as fast as they could to outpace competitors and secure more capacity. The concern was that an unrestricted race could eventually lead to overbuilding. Political hurdles might therefore have a positive side: they could make companies more deliberate about how much capital they put to work and potentially mitigate that risk.
That is a conditional argument, not a claim that new rules will necessarily arrive or that construction will stop. Rathbun characterized the moment as a “hiccup” in the broader cycle. Her long-term view was that the buildout would slow, not end; what changes is the pace at which capital is deployed and returns arrive.
The worry was always that, okay, if you build it as fast as you can, at some point you're going to overbuild.
The selloff puts the timing of cash flows in focus
Rathbun described the market decline as a readjustment to the possibility of disruption and to questions about when data-center investments will pay off. Stenovec said the Nasdaq 100 was down close to 6% during the session, after falling 6.2% earlier. The scale of the drop was immediate; Rathbun’s point was that the consequences would not be uniform across companies exposed to AI infrastructure.
She separated companies that earn money while data centers are being built from those whose revenue depends on the facilities being completed and brought online. Hyperscalers, she said, had been borrowing and were expected to spend on capital expenditures. Chipmakers such as Nvidia, construction firms, and other suppliers to data centers could still get paid as spending and construction continued, even if the buildout slowed.
For companies waiting for data centers to go live, the issue is different. Their expected cash inflows may arrive later than planned. A slowdown can therefore leave the prospect of payment intact while changing when a company receives it. Rathbun said stocks were already in correction, but expected the market eventually to price companies differently according to those different revenue timelines.
A longer wait makes balance-sheet strength decisive
For firms that must wait for facilities to go live, time is also a financing problem. Rathbun said companies need enough cash to keep operating while revenue is deferred. Highly leveraged firms face more pressure because a longer delay extends the period they must remain a going concern before expected inflows arrive.
She singled out “neo clouds” as facing a double exposure: in her account, they are highly leveraged and get paid when data centers go live. If construction or activation takes longer, they must endure the delay while waiting for the revenue their business depends on. Rathbun called this a quality game: can a company afford to wait?
That distinction informs her investment view, but she was not recommending a broad purchase of AI-related stocks. She said the selloff could create a buying opportunity for some high-quality names, particularly for investors with a long horizon who still expect the AI data-center buildout to happen over time. Her clients are mostly long-term investors, and she said she was “dipping in” because some prices looked very cheap to her. She also acknowledged the risk of buying into a continuing decline: “sometimes you don't want to be catching a falling knife.”
That changes the profile of the internal rate of return, not necessarily that the companies are going to get paid.
Rathbun’s position is that the buildout is likely to continue, but at a slower pace, with returns arriving on a different timetable. For investors, the practical questions are whether a company can withstand that delay and whether it earns revenue during construction or only after a data center is operating.