Oracle Collects Tomorrow's Cash

If You Only Read One Thing

Nearly half of Oracle’s operating cash came from customer advances for future services. Oracle Collects Tomorrow’s Cash reads the bargain in its earnings tables; Europe Charges for Resilience follows the ECB’s rate increase. They meet in capital markets: Lagarde identifies AI financing demand as a driver of bond yields. The financing burden can reach businesses that never buy a GPU.

Oracle Collects Tomorrow’s Cash

Oracle’s customers are becoming part of its financing system. That makes the cloud expansion easier to fund, but it also makes operating cash flow a less reliable shorthand for cash earned from work already delivered.

The September 10 results are strong on actual output. Cloud infrastructure revenue rose 121% to $7.4 billion. Oracle added 850 megawatts of data-center capacity during the quarter. The company’s quarterly release reports $664 billion of contracted future revenue, including more than $30 billion of additional AI contracts that it says require no change to its capital-raising plans.

The revealing detail sits in the cash-flow statement. Customer prepayments with a significant financing component contributed $11.363 billion to $23.103 billion of operating cash flow: 49.2% of the total. Capital expenditure was $28.499 billion. Subtracting that expenditure leaves negative $5.396 billion of free cash flow, before any attempt to normalize customer advances.

Think of a customer paying a builder before the house is finished. The payment funds construction today, but the builder still owes the house. Oracle’s customer-financed expansion similarly exchanges future service for current liquidity. The money is real; treating all of it as repeatable cash earnings would count tomorrow’s economics too early.

Customer advances supplement capital markets; they do not replace them. Oracle’s comparative tables show these prepayments rising from $4.592 billion in the preceding quarter. Oracle also completed $20 billion of common-stock sales before commissions this quarter. Customers and shareholders are both funding the construction interval.

The strongest bullish reading is that sophisticated customers are willing to finance supply they cannot obtain elsewhere. Advances can reduce Oracle’s need to borrow and make a signed contract more credible than an unfunded reservation. The corresponding risk is concentration: an upstream cloud supplier may depend on the financing capacity of the same labs whose demand justifies its expansion.

My medium-confidence investment read favors equipment and power suppliers with firm payment terms over treating Oracle’s cash-flow growth as a finished equity thesis over the next 12–24 months. This extends the standing pattern of scarce physical inputs capturing value early. Oracle can still win through delivery scale; the current evidence establishes that advantage more clearly than recurring cash profitability. Cancellations reaching equipment suppliers would break that relative preference.

The next quarterly test is operating cash flow after separately identifying financing-component customer prepayments. Growth in that remainder would strengthen the case that delivered services are catching up with construction.

Europe Charges for Resilience

Europe’s economic resilience has become a reason to make borrowing more expensive. The ECB is acting against the risk that an energy shock spreads, even though its preferred evidence of broad price pressure has recently softened.

On September 10, the bank raised its deposit rate by a quarter point to 2.50%, effective September 16. It forecasts 2026 growth of 0.9%, revised upward, while inflation is projected at 3.0% this year and 2.5% next year. Stronger activity gives policymakers more room to tolerate the immediate cost of restraint.

The split inside inflation explains the wager. The ECB’s press-conference statement puts August energy inflation at 14.3%. Yet inflation excluding food and energy fell from 2.5% to 2.4%, and services inflation fell from 3.3% to 3.0%. Wages, the bank says, have not materially responded to the energy shock.

The feared mechanism is second-round inflation. Imagine a delivery company absorbing expensive diesel, then raising its annual shipping prices; workers subsequently seek pay increases to cover the higher cost of goods. An initially narrow fuel increase becomes a broader pattern of wages and prices. Higher rates cannot produce diesel, but they can weaken the spending that allows those increases to stick.

The connection to Oracle runs through capital markets. In the same press conference, Lagarde identified AI companies’ financing demand as a driver of rising bond yields. Yesterday’s oil story concerned the war’s duration; today, energy inflation and investment demand are both pressuring borrowers. Customer advances shift who supplies capital, not whether the buildout needs it.

The counterargument is substantial. Kiel Institute economist Lena Dräger supports this increase but argues for a subsequent pause: additional tightening could arrive just as the direct energy effect fades. That is a warning about policy lag, not evidence that fuel inflation is harmless.

My medium-confidence read for the next 6–12 months favors cash-generative businesses with limited refinancing needs over highly indebted European property and discretionary businesses. The macro shock can raise operating costs and financing costs together. Banks are not automatic winners: better loan yields can be offset by expensive deposits and weaker borrowers. Broad disinflation followed by rate cuts would weaken that relative preference.

The decisive next signal is negotiated wage growth against the ECB tracker’s 2.7% expectation for early 2027: acceleration would support the feared transmission; continued restraint would favor a pause.

The Contrarian Take

Everyone says: An energy-driven rate increase is futile because a central bank cannot reopen a shipping route.

Here’s why that’s wrong (or at least incomplete): Monetary policy targets the propagation of the shock, not the missing barrel. The awkward fact is that European services inflation and wage growth have softened, so the strongest argument supports a precautionary increase rather than an automatic sequence. Kiel’s analysis makes the distinction: insurance can be reasonable while repeated tightening becomes excessive. A forecast of persistent inflation is a reason to test transmission carefully, not proof that transmission has already occurred.

Under the Radar

Shopify is withdrawing a subsidy to the software ecosystem. Its September 10 return to native mobile development includes a quieter funding consequence: React Native Skia sponsorship runs through year-end, FlashList needs long-term stewardship, and Restyle maintenance ends after 2026. The company says agents reduced the cost of maintaining separate Apple and Android implementations; when that makes shared code less valuable internally, outside developers inherit the maintenance bill. The strategic winner may be the platform owner whose native tools no longer require the same duplication of human effort.
The cargo investigation now contains a response failure to examine. The NTSB’s September 9 recorder update describes repeated speed concerns without a consistent verbal response before the 21 Air crash in Miami. This advances September 7’s contractor-accountability story from questions about the carrier relationship to specific cockpit evidence. The findings are preliminary and establish no probable cause; their business significance is that an audit must examine how warnings change operations, not simply whether a warning system exists.

Quick Takes

Anthropic’s report makes private monitoring a public institution. Its new report spans seven misuse areas and alleges Alibaba extracted over 151 million Claude exchanges for model training. Anthropic also describes potential biological misuse but explicitly does not assert malicious intent by the scientists. These are the lab’s findings, not independently established facts. The structural question is who audits a vendor that sees evidence governments and customers cannot; selected disclosures do not measure undetected abuse. (Source)
Trump promises a distribution without an enacted funding mechanism. The White House reiterated a proposed $5,000 payment to every adult citizen, conditional on Republicans retaining both congressional chambers. This is a campaign promise, not an authorized payment. Its economic significance is the policy direction: broad cash transfers would support demand while central bankers worry about inflation, leaving fiscal and monetary policy potentially working against each other. (Source)
Cognition is internalizing more of its supplier’s margin. After the funding covered on September 9, it introduced SWE-2, derived from Kimi K3. Cognition reports performance within one point of Fable 5.1 on its FrontierCode benchmark at 64% lower cost. Those are vendor results, not established customer savings. The business implication is clearer: an agent distributor can refine an open model rather than surrender every incremental dollar to a frontier supplier. (Source)

The Thread

The cloud buildout can make financing harder for businesses outside technology. Oracle collects customer cash to secure capacity, while Lagarde identifies AI financing demand as one force pushing up bond yields; Europe’s energy shock adds a separate reason for rates to rise. If enacted without offsetting restraint, Trump’s promised dividend would add demand to that mix. This extends the scarce-input pattern: owning equipment or power is only part of the advantage when the money needed to build them is becoming expensive. Firms able to secure customer advances can keep investing while borrowers dependent on refinancing absorb tighter terms. The next separation is therefore between companies whose customers fund expansion and companies whose customers need cheaper credit before they can spend.

Predictions

New predictions:

I predict: Oracle will report cloud-infrastructure revenue above $10 billion in at least one quarter by the end of fiscal 2027. (Confidence: medium; Check by: 2027-06-30)

Coming Next Week

Next week’s question is how much of the cloud investment cycle customers are financing themselves. The distinction between an order, an advance payment and earned cash is becoming as consequential as revenue growth.

September 11, 2026 · 03:19 AM ET