Oracle Plans Major Expansion of AI Compute Capacity to 10GW by FY28

Oracle has announced an ambitious plan to expand its AI compute capacity from approximately 1 GW in fiscal year 2025 to 10 GW by fiscal year 2028. This tenfold increase is part of a broader strategy unveiled during its October 2026 Investor Day, where the company outlined its roadmap for growth in AI infrastructure.

The company reported a backlog of $664 billion in remaining performance obligations (RPO) by the end of the first quarter of fiscal year 2027, a figure that has surged due to increasing demand for AI services. In the first quarter alone, Oracle brought online 850 MW of new capacity, which included over 300,000 GPUs across 38 projects in 11 countries.

Oracle's strategy involves a mix of financing through partnerships, customer pre-payments, and a 'bring-your-own-hardware' model, where customers supply their own chips while Oracle provides the necessary infrastructure. The company anticipates that capital expenditures will peak in fiscal years 2027 and 2028, with free cash flow expected to turn positive in the second half of fiscal year 2028.

During the Investor Day, Oracle also revealed that a significant AI infrastructure contract carries approximately $80 billion in RPO, with an estimated capital expenditure of around $37 billion. This contract is expected to generate roughly $16 billion in annual revenue once fully operational.

As Oracle transitions from a traditional software vendor to a major player in hyperscale data center operations, its success will depend on its ability to deliver on these ambitious plans and convert its substantial backlog into recognized revenue.