Oracle has turned AI demand into a bridge-financing problem. Its fiscal 2026 release says remaining performance obligations reached $638 billion, up 363 percent year over year, while free cash flow was negative $23.7 billion as the company built out cloud infrastructure. [1] A backlog that large does not end the financing question. It creates one.

The paper's June 17 feature on Oracle backlog as financing structure, not simple demand warned that RPO is not cash. June 18 supplies the bridge terms. Oracle says it raised $43 billion in debt financing and $5 billion in equity financing in fiscal 2026, and expects about $40 billion of combined debt and equity financing in fiscal 2027. [1]

That is the part of the AI boom that polite demand stories tend to bury. Oracle's release is bullish by any ordinary revenue standard: total Q4 revenue up 21 percent to $19.2 billion, total cloud revenue up 47 percent to $9.9 billion, and cloud infrastructure revenue up 93 percent. [1] The same release says the backlog increase is tied to large-scale AI contracts, with prepaid and customer-supplied hardware portions totaling $75 billion. [1]

Customer-supplied hardware and prepayments are not footnotes. They are the mechanism that keeps the bridge from collapsing under its own weight. If a customer prepays for GPUs, Oracle's capital need falls. If a customer supplies hardware, Oracle's capex burden changes. But both facts also tell readers that this is not a clean old-software backlog. It is a physical build-out wrapped in contracts, financing, customer dependence, and delivery risk.

OpenAI sits beside that story, not inside the cited backlog record. OpenAI says Oracle Cloud customers will be able to use eligible Oracle Universal Credits to access OpenAI models and Codex through OCI. [2] Fortune's IPO analysis asks what OpenAI's eventual S-1 will reveal about revenue, costs, and valuation. [3] The two stories meet as separate disclosure problems: Oracle's release names a vast AI backlog without naming the customer mix, while OpenAI's Oracle distribution rail shows why AI demand keeps pressing against private economics.

MSM can celebrate Oracle's record cloud demand and still miss the financing bridge. X can call the bridge a debt bubble and still miss the contract quality behind it. The hard version is less satisfying: Oracle may have captured real AI demand, but real demand for physical AI infrastructure must be financed before it becomes cash.

The financing schedule matters because timing is destiny in infrastructure. If customers prepay, if hardware arrives, if data centers power up, and if usage converts into recognized revenue at attractive margins, debt becomes a rational bridge to contracted cash. If capacity lags, utilization disappoints, margins compress, or customer concentration bites, the same bridge becomes leverage against a delayed future.

Oracle's release tries to answer part of that criticism by saying prepaid and customer-supplied hardware substantially reduce the capital Oracle must raise to build AI data centers. [1] That sentence is both reassuring and revealing. A company does not need that sentence for ordinary software demand. It needs it when demand is welded to metal, land, power, chips, and financing markets.

That is why the backlog should be read less like a trophy and more like collateral. It supports borrowing because it suggests future cash. It also requires borrowing because the future cash needs a physical machine built first. Oracle's AI story now lives between those two facts.

-- DAVID CHEN, Beijing