The calls began months ago, before either company had filed for an offering, according to people familiar with the efforts. Investment banks advising Anthropic and OpenAI have been making the rounds at Standard & Poor’s, Moody’s and Fitch, arguing that the two artificial intelligence companies deserve investment-grade credit ratings when they eventually sell shares to the public.
The lobbying reflects a quiet front in the AI industry’s march toward public markets. Both companies are widely expected to pursue initial public offerings in the coming months, and their advisers have concluded that the first impression they make on the bond market matters as much as the one they make on equity investors. A top-tier rating, the bankers argue, would let the two companies borrow at far lower cost once listed, funding the data-center buildouts that have become the defining expense of the sector.
The pitch to rating agencies rests on a straightforward claim: whatever the risks of fast-moving technology and enormous capital spending, the companies’ revenue growth, large cash balances and strong demand justify high credit quality. Bankers point to the scale of contracted commitments from customers and the pace at which both firms have grown sales, figures that have made them two of the fastest-growing software businesses on record.
Rating agencies are not easily swayed by courtship. Analysts at the three firms have spent the past year building analytical frameworks for AI companies, wrestling with questions the industry has not yet answered: how durable are the revenue streams, how much of the enormous compute spending is locked in by contract, and what happens to credit quality if the pace of model improvement slows. The banks’ arguments land in a debate those analysts are already having internally.
The stakes are unusually high because of the size of the financing that follows. Anthropic has disclosed agreements for computing capacity that could cost hundreds of billions of dollars over their terms, and OpenAI has described plans to secure tens of gigawatts of power. Neither company can fund that trajectory from operating cash flow alone. The bond market, with its capacity for long-dated, low-cost debt, is the natural next source, but only for borrowers whose ratings keep their interest bills manageable.
Investment-grade status matters in ways that go beyond borrowing costs. A rating of BBB- or above from at least two agencies would widen the pool of buyers for the companies’ debt, since many institutional investors are barred from holding speculative-grade securities. It would also signal to equity investors that the rating agencies, after months of scrutiny, found the business models could support the debt they planned to carry. For companies preparing to sell stock for the first time, that validation carries its own price.
The bankers face structural obstacles. Rating agencies typically require a track record of audited financial statements and a demonstrated ability to generate cash, and both companies have spent their existence prioritizing growth over profit. The technology itself complicates the analysis: models are replaced every year or two, and a competitor’s advance can render expensive infrastructure less valuable. Agencies have signaled that they will treat compute contracts as real liabilities, not optional commitments.
There is also the question of corporate structure. Neither company has issued public debt, and both have layered financing from private investors and strategic partners, including the cloud providers that also supply their computing capacity. Rating analysts must decide whether those relationships strengthen or complicate the credit picture, particularly where suppliers and customers are the same entities.
The precedent is thin but not absent. Technology companies with heavy capital needs and thin margins have navigated the rating process before, and the cloud giants that dominate the industry carry investment-grade ratings built on cash flows far larger than anything Anthropic or OpenAI can yet show. The question for the agencies is whether rapid growth in artificial intelligence earnings power justifies an early upgrade, or whether ratings should wait for proof.
The courtship runs into the agencies’ institutional memory. Rating firms emerged from the last financial crisis determined to demand evidence over optimism, and their analysts have spent months debating how to treat contracts that commit AI companies to billions of dollars in future spending against revenue that is growing quickly but has not been tested through a downturn. People familiar with the agencies’ thinking say the analytical focus has settled on contract structure: how much of each company’s revenue is committed by customers, whether the compute agreements carry obligations the company cannot shed, and how fast spending could be adjusted if demand stalls.
The arithmetic of the difference is not abstract. On the scale of borrowing both companies contemplate, even a one-notch gap in rating can translate into tens of millions of dollars in annual interest costs, and across the long tenors favored for infrastructure debt the spread compounds into a sum large enough to fund additional computing capacity. That is why the bankers keep calling.
People familiar with the discussions say the agencies have made no commitments, and none are expected before the offerings take shape. The banks’ goal in the interim is more modest: to make sure the first formal presentations are met with frameworks that already account for the industry’s peculiar economics, rather than with skepticism born of unfamiliarity. In a sector where perception can move financing costs by whole percentage points, that preparation is itself a form of return. The rating decisions, when they come, will tell investors how Wall Street’s most cautious institutions read the AI boom.
