Oracle is spending more on building data centres than it earns in a quarter

Oracle disclosed $664 billion in remaining performance obligations, representing contracted customer work yet to be delivered. The backlog provides substantial forward revenue visibility, likely reflecting demand for Oracle's cloud and AI infrastructure offerings, while shifting execution focus to the company's ability to build and fulfill those commitments.

Analysis

The key underwriting issue is not demand visibility but conversion capacity. A backlog of this scale can support a multi-year revenue-growth and multiple-expansion narrative only if Oracle can secure enough GPUs, power, data-center capacity and implementation labor without sacrificing cloud gross margin. The market is likely to capitalize the headline immediately; the more investable catalyst over the next 1-3 months is whether management quantifies annualized OCI revenue conversion, capex, committed power capacity and customer concentration at the next earnings call.

Oracle’s positioning is increasingly that of an AI-infrastructure financing and deployment vehicle rather than a mature software vendor. That raises upside to OCI growth but also increases sensitivity to hyperscaler-style risks: capex running ahead of recognized revenue, depreciation pressure, customer-credit exposure and delayed go-lives. The second-order beneficiary is NVIDIA (NVDA), where Oracle’s buildout reinforces enterprise GPU demand; however, Oracle’s incremental demand may be less profitable than its software base, creating a risk that earnings beat while free-cash-flow conversion disappoints over the next 6-18 months.

Consensus may be underestimating concentration and execution risk embedded in contracted demand. Large AI commitments can be renegotiated if model economics weaken or power delivery slips, and backlog is not equivalent to near-term revenue or cash collection. A material increase in OCI capex without a commensurate increase in cloud gross-margin guidance, or disclosure that a small number of customers represent an outsized share of commitments, would challenge the rerating thesis.

Key Decisions for Investors

• Maintain a tactical long ORCL into the next earnings call only if the stock has not already rerated materially on the backlog disclosure; target a 1-3 month catalyst from OCI revenue and guidance detail, with risk defined by a failure to raise FY cloud-growth or capex-conversion outlook.
• Prefer a pair trade long ORCL / short IBM over 3-6 months: Oracle has a clearer AI-infrastructure demand vector, while IBM remains more exposed to slower legacy-services spending. Exit if Oracle indicates cloud gross-margin dilution or IBM materially accelerates GenAI bookings conversion.
• Use NVDA as the cleaner second-order expression of Oracle capacity additions, but avoid chasing spot strength; accumulate on AI-capex pullbacks with a 6-12 month horizon. Falsifier: evidence that Oracle is constrained by power or financing rather than GPU availability, which would defer hardware revenue recognition.
• Set a diligence alert for Oracle’s next filing/call: quantify remaining-performance-obligation duration, top-customer concentration, OCI capex and free-cash-flow guidance. Without these data, treat the backlog figure as directional demand evidence rather than a standalone valuation anchor.