S&P Global reduced Oracle's credit score to one notch above junk on July 9. The company that has quietly become the financial engine behind Larry Ellison's expanding media ambitions found itself explained not by its film studios or news networks but by the scale of its borrowing against artificial intelligence infrastructure it has not yet finished building. Oracle's shares have shed roughly $230 billion in value since September. Its bonds are trading like the debt of a company under strain, because in the market's estimation, it is one.

The story reads, at first, as a piece of American corporate drama, playing out somewhere between Silicon Valley balance sheets and a takeover battle over Warner Bros. Discovery. Beneath that, however, is a pattern that extends far beyond Oracle and entertainment. The largest infrastructure investment programme in a generation is being financed, in increasing measure, through debt rather than the free cash flow that once made these companies unusually resilient. Bank of America now classifies the big technology groups building AI data centres as capital intensive in the manner of the old oil majors, a description almost nobody would have applied to them three years ago.

That matters for the NHS for reasons that are not immediately obvious, though they are not remote either.

The Federated Data Platform, the health service's most important data infrastructure initiative, was given to a consortium led by Palantir in November 2023. The contract has an outer term of 2031 and a maximum value of £330 million. The initial three-year commitment ends in March 2027, and a break clause review is scheduled for the spring, at a point when adoption evidence and the case for continuation will be weighed by ministers already sensitive to the criticism the contract has attracted in Parliament. MPs have raised the prospect, credibly, that the NHS could reach the end of a subscription arrangement worth hundreds of millions of pounds with no software and no durable intellectual property to show for it.

That question, of what the NHS is actually left holding when a contract with a fast-growing technology supplier ends, has usually been framed around governance and the discomfort of placing sensitive health records with a company that has defence and security origins. Those concerns remain legitimate. But Oracle's downgrade adds a dimension that has had far less attention inside health policy circles. The financial stability of the companies now embedded in critical national infrastructure is not fixed. It moves, sometimes quickly, and it moves in response to spending decisions made well outside health policy altogether.

Palantir is not Oracle, and its balance sheet does not carry comparable debt. But the wider environment in which AI-adjacent suppliers now operate has shifted in ways that should inform how NHS England and the Department of Health and Social Care assess vendor risk, not only at the point of procurement but across the life of a contract. A supplier's credit profile and its dependence on capital markets that have become considerably less forgiving are now legitimate lines of inquiry for a body committing public money over a seven-year horizon.

There is a second, more diffuse implication. AI-related capital spending has been responsible for a significant portion of recent economic expansion in the United States and, thus, some of the budgetary headroom supporting UK projections. JPMorgan Asset Management puts the contribution at just over one percentage point of American growth this year. If that spending slows because the debt financing comes under pressure, growth assumptions built into Office for Budget Responsibility forecasting, and by extension the funding settlements the NHS depends on, are not insulated from a correction playing out in bond markets thousands of miles from any hospital trust.

None of this suggests the Federated Data Platform was wrongly procured or that ministers should panic. It suggests something more modest and rather overdue. Vendor solvency, not only vendor conduct, belongs on the risk register when the NHS signs infrastructure contracts of this scale. Oracle's downgrade is not an NHS story. But the assumptions that produced it are increasingly the assumptions the NHS is relying on when it decides who builds the systems it cannot function without.