Bank of America is pushing back on the idea that CoreWeave's aggressive data-center expansion is a financing trap. In a research note, the bank argued the AI cloud provider has a clearer path to funding its power build-out than the market has assumed — a view that lands after the stock fell 15.5% on Wednesday amid broader anxiety about debt across the AI sector.
The core of the argument is structural. CoreWeave doesn't build most of its data centers itself. Instead, it leases capacity from third-party developers, then signs matching agreements with suppliers and customers so that costs and revenue line up. Bank of America says that approach, combined with hefty prepayments from customers, reduces the amount of capital CoreWeave has to front on its own.
Why the market got nervous
CoreWeave rents out access to high-end graphics chips — the hardware that powers AI models — to companies that need enormous amounts of computing power but don't want to own the infrastructure. That business is growing fast, but it is also capital-hungry. Adding electricity, cooling and servers at scale takes billions of dollars, and much of that money has historically come from debt markets.
That reliance on borrowing has made CoreWeave a lightning rod for investors worried that the AI boom is being financed with too much leverage. When bond yields for AI-linked companies spike, as they have recently, the market starts asking whether the build-out can keep going at the same pace. Those concerns have shown up in CoreWeave's share price and in the wider AI trade, with related names selling off alongside it.
Wednesday's 15.5% drop fits that pattern. It wasn't a reaction to a specific piece of bad company news so much as a repricing of risk across the AI infrastructure complex. Investors have been rotating out of the most leveraged names and demanding more evidence that the spending is sustainable.
The prepayment angle
Bank of America's note tries to answer that demand. Customer prepayments are essentially cash paid upfront for future computing capacity. For a company like CoreWeave, that money can be used to fund construction and equipment without immediately adding to the debt pile. It also signals that customers are committed — they've put money down, which makes the revenue stream more predictable.
The back-to-back contract structure matters for a similar reason. If CoreWeave signs a lease for data-center space at the same time it signs a contract to sell the computing capacity that will run there, the economics are locked in before the spending happens. That's different from building speculatively and hoping to fill the space later. In theory, it turns a capital-intensive gamble into something closer to a matched book.
Bank of America's estimate that this model could support 4.8 gigawatts of capacity by the end of 2028 is the headline number. For context, a gigawatt is a billion watts — roughly the output of a large power plant. Building that much AI-ready capacity is an enormous undertaking, and the bank's point is that the financing may be more manageable than the raw number suggests.
What it means for investors
For everyday investors, the CoreWeave story is a useful window into how the AI trade is evolving. The first phase was about chip makers and software companies with obvious earnings power. The current phase is about the infrastructure layer — the data centers, power contracts and financing structures that make AI compute possible. That layer is more leveraged, more complex and more sensitive to interest rates.
When a major bank publishes a note defending a beaten-down AI name, it doesn't mean the concerns were wrong. It means there's a debate. Bulls will point to the prepayments and matched contracts as evidence that CoreWeave's growth is funded prudently. Bears will note that leasing still creates long-term obligations, that customers could renegotiate, and that the entire model depends on AI demand continuing to grow at its current pace.
Investors watching this space should focus on a few things. First, the pace of new customer prepayments — if they slow, the funding math gets harder. Second, the cost of debt for AI infrastructure companies, which has been rising. Third, whether the power and data-center capacity actually gets built on schedule, since delays would push revenue further out.
It's also worth remembering that CoreWeave is not a typical utility or REIT. It sits at the intersection of cloud computing, semiconductors and energy, and its fortunes are tied to a technology cycle that is still young. That makes it a higher-risk holding than a diversified tech fund, and one where sentiment can swing sharply in either direction.
The broader takeaway is that the AI build-out is increasingly a financing story as much as a technology story. Companies that can lock in customers and funding early will be better positioned than those that can't. Bank of America thinks CoreWeave is in the first camp. The market, at least on Wednesday, wasn't so sure.
