Burry's Oracle short rests on $261bn of leases that have not started

Oracle's cloud infrastructure segment booked $7.39bn in the quarter it reported on September 10, which annualises to about $29.6bn, so $261bn of lease commitments comes to roughly 8.8 years of the revenue those buildings are meant to produce ^{[5][14][3][15]}.

The 121% came off a small base. A year earlier the same segment did about $3.34bn in the quarter, so the increase added roughly $4.05bn of quarterly revenue ^[16]. Annualised, cloud infrastructure is about a third of the $90bn Oracle has now guided for the year ^[17][7].

Many of the leases have not commenced, according to Burry's reading of the filings ^[4]. Oracle owes payments on facilities that are not operational and not generating revenue ^[4]. Operating leases kept off the balance sheet can carry enormous future cash obligations without drawing the scrutiny that borrowed money draws, and Burry says most analysts skip that part of a filing ^[10]. So the 121% and the $261bn describe two different moments: the revenue comes from buildings that exist, the obligations attach to buildings still to come ^[5][4]. Burry's post argues there may be trillions of such commitments across the five hyperscalers, and its title names $3 trillion ^[11][1].

S&P reached the leverage question in July 2026, before the quarter that printed 121% growth ^[8][5].

Burry, who bet against the housing market before the 2008 crisis, holds short positions against Oracle ^[12][9]. The position and the accounting argument point the same way. The $261bn came out of Oracle's own filing. The durations and counterparties behind Oracle's contracted capacity are not in this account. The case against the short is simple enough: a lease that has not commenced is an obligation on a delivery schedule, and if the capacity behind it is already contracted to customers for a matching term at decent credit, the payments are covered before the facility is operating.