TL;DR
S&P cut Oracle to one notch above junk as its AI data-centre bet burns cash and bond investors price in the risk.
Oracle's ten-year bonds already yield closer to junk than investment grade, its free cash flow has turned deeply negative, and roughly half of its contracted revenue is tied to a single customer
S&P cut Oracle to one notch above junk as its AI data-centre bet burns cash and bond investors price in the risk.
S&P downgraded Oracle to BBB- on July 9, placing the company one notch above junk status, as a $250 billion data-centre expansion plan burns through cash faster than revenue can replace it. Oracle is now the second-largest non-financial debt issuer in the Bloomberg US Corporate Bond Index after Amazon, with $117 billion outstanding. Shares fell nearly six percent on Thursday as bond investors began pricing Oracle’s debt closer to speculative grade.
Oracle’s free cash flow turned deeply negative in its fiscal year ended May 31, with the company burning through nearly $24 billion after capital expenditure. S&P estimates that deficit could widen to $42 billion as Oracle continues building data centres at an unprecedented pace. Moody’s has also placed a negative outlook on the company, signaling that a second ratings agency sees material risk in Oracle’s trajectory.
The bond market is already acting as if the downgrade has further to go. Oracle’s ten-year bonds yield roughly six and a half percent, well above the BBB index average and closer to the BB range that marks junk territory, according to Bloomberg. George Catrambone, head of fixed income at DWS Americas, told Bloomberg the gap reflects investors demanding a premium for the uncertainty around whether AI revenue will justify the debt.
The bet is heavily concentrated. S&P estimates that roughly half of Oracle’s $638 billion in remaining performance obligations, its measure of contracted future revenue, is tied to OpenAI. Oracle spent more than $55 billion on data centres in its last fiscal year and expects to raise another $40 billion through debt and equity this year, including $20 billion in stock sales at market prices.
Oracle is not alone in loading up on debt for AI infrastructure. Hyperscalers collectively plan to spend up to $725 billion on AI this year, and Big Tech’s combined AI debt has already hit $350 billion, according to Bloomberg. But Oracle lacks the cash-flow cushion that protects its peers: Google posted about $73 billion in free cash flow last year, while Oracle’s cash generation has collapsed under the weight of its capital spending.
Oracle has asked customers to pre-pay for computing components to help offset the financing burden, and says prepaid and customer-supplied hardware for large AI contracts now totals $75 billion. Whether that is enough to prevent a further downgrade depends on how quickly its cloud revenue, which grew 93 percent last quarter, can catch up with its borrowing. The company is betting everything on AI demand sustaining at levels that justify turning one of the most leveraged balance sheets in tech into something even more leveraged.
Get the most important tech news in your inbox each week.