Oracle’s $90–95B AI Expansion: 97.5% GPU Utilization but Margin Risks
ORCL•Oracle plans fiscal 2027 capital expenditures of $90–95 billion (net cash outlay ~$70 billion), funded by $40 billion in debt and equity, shifting its margin profile toward capital-intensive AI infrastructure. At 97.5% GPU utilization, robust AI demand clashes with operational risks highlighted by Microsoft’s scrapped $3 billion data-center lease.
1. Strategic Pivot to AI Infrastructure
Oracle plans fiscal 2027 capital expenditures of $90–95 billion, with a net cash outlay of approximately $70 billion, to build a global AI data-center network. The company will raise about $40 billion in debt and equity to fund this shift from high-margin software to capital-intensive infrastructure.
2. Profitability and Margin Outlook
Management expects fiscal 2027 gross margin to step down as new data centers come online, introducing near-term pressure on profitability. The long-term margin profile of this AI infrastructure business remains untested, creating valuation risk if returns trail expectations.
3. Demand Indicators and Partnership Setback
Data center GPUs are running at 97.5% utilization, reflecting strong enterprise AI demand and a boost for accelerator suppliers. However, Microsoft’s decision to walk away from a potential $3 billion Oracle data-center lease over security framework disputes underscores execution and compliance challenges.
