Oracle’s tie-up with the San Francisco-based research lab underscores the database vendor’s ambitions in the fast-growing market for high-performance computing services. Although Oracle previously relied chiefly on software licensing and business applications, its cloud-infrastructure segment now supplies more than one-quarter of total revenue. Even with that growth, the firm trails the three largest hyperscalers—Amazon Web Services, Microsoft Azure and Google Cloud Platform—whose capital expenditures on data centers and specialized chips dwarf Oracle’s outlays.
OpenAI, best known for ChatGPT, committed in September to spend more than $300 billion with Oracle over the next five years. Clay Magouyrk, who became one of Oracle’s two chief executive officers in September, described the relationship as strong during an investor meeting the following month. The multiyear arrangement designates Oracle Cloud Infrastructure to host certain large-scale workloads and to supply advanced graphics processing units (GPUs) required for generative-AI training and inference.
The organization led by Sam Altman is also lining up capacity with additional vendors to meet escalating demand for its models. In September, Nvidia disclosed a nonbinding letter of intent to provide at least 10 gigawatts of computing equipment to OpenAI, with the first tranche targeted for the second half of 2026. Nvidia’s management noted in a November filing with the U.S. Securities and Exchange Commission that the arrangement is not yet definitive, stressing that negotiations could still fall through.
OpenAI’s diversification strategy extends beyond procuring off-the-shelf GPUs. The company is collaborating with Broadcom to design custom application-specific integrated circuits tailored for its AI workloads, an initiative revealed in October. Speaking on Broadcom’s earnings call Thursday, Chief Executive Hock Tan outlined a preliminary timetable that forecasts meaningful output between 2027 and 2029, with limited contribution expected in 2026. Tan characterized the understanding with OpenAI as an “alignment” but emphasized that it remains subject to final terms.
Imagem: Internet
The confluence of projects underscores the scramble across the semiconductor, cloud and AI sectors to secure infrastructure able to power large language models. A scarcity of top-end Nvidia chips has prompted providers to add manufacturing capacity and to experiment with proprietary hardware, even as supply chains grapple with continued constraints in both technical talent and component availability.
For Oracle, timely delivery of the promised data-center footprint is pivotal to expanding its share in the cloud market and to reinforcing investor confidence. The company’s Friday clarification aimed to counter suggestions that shortages could impede progress. While Oracle acknowledged the macroeconomic pressures affecting construction and technology deployment globally, it reiterated that project schedules remain aligned with the customer’s requirements.
OpenAI declined to comment on the timeline. Meanwhile, market watchers will look for concrete updates as Oracle reports earnings, Nvidia formalizes agreements, and Broadcom advances its chip-design program. Any confirmed slippage could influence spending plans across the broader AI ecosystem, where capital commitments already span multiple years and tens of billions of dollars.
Crédito da imagem: Kyle Grillot | Bloomberg | Getty Images
