OpenAI halts "Stargate" expansion, billion-dollar debt expansion model faces setbacks, Oracle(ORCL.US) faces AI infrastructure "outdated" dilemma

The pace of artificial intelligence chip updates has far surpassed the construction cycle of data centers. This market reality is exposing the core risks of AI trading and the fragility of Oracle’s (ORCL.US) debt-driven expansion model.

According to sources, OpenAI no longer plans to expand its partnership with Oracle in Abilene, Texas. This is the site of the Stargate data center project, and OpenAI’s decision is driven by its desire to acquire a new generation of NVIDIA GPU clusters.

It is understood that the current facilities in Abilene are expected to use NVIDIA’s Blackwell processors, with power access anticipated to take another year. The sources say that by then, OpenAI hopes to secure larger-scale NVIDIA next-generation chip clusters elsewhere.

Oracle posted on its X platform on Sunday that the reports are “false and incorrect,” but the post only stated that the existing project is progressing smoothly and did not mention any expansion plans.

It is reported that Oracle had previously acquired the land, ordered hardware, and invested billions of dollars in construction and staffing, originally expecting to further expand.

For OpenAI, this is a reasonable decision; they certainly do not want to use outdated chips. NVIDIA has traditionally released a new data center processor every two years, but now CEO Jensen Huang has accelerated this to once a year, with each generation showing significant performance leaps. The Vera Rubin chip, unveiled at CES in January and now in production, has inference performance five times that of Blackwell.

For companies building cutting-edge models, even the smallest performance improvements can lead to huge differences in benchmarks and rankings. Developers pay close attention to these metrics, which directly impact user numbers, revenue, and valuation.

All of this points to a deeper structural issue. For infrastructure companies, site selection, power access, and facility construction take at least 12 to 24 months. But customers are chasing the latest and greatest technology, closely watching NVIDIA’s annual chip upgrades.

Oracle faces additional challenges: it is the only major cloud service provider heavily relying on debt financing to support expansion, with debt reaching $100 billion and continuing to grow. In contrast, Google, Amazon, and Microsoft rely on their own massive cash cow businesses.

Meanwhile, Oracle’s partner Blue Owl has refused to finance another facility and plans to lay off up to 30,000 employees.

Oracle will release its Q3 earnings on Tuesday, and investors will be watching closely how management addresses the contradiction between its $50 billion capital expenditure plan and negative free cash flow, as well as whether its financing channels can sustain.

The stock has fallen 23% so far this year, with its market value halved since peaking last September.

Looking across the entire market, the risk of GPU depreciation extends far beyond Oracle. It could trigger a chain reaction across the AI sector, as every infrastructure deal signed today may be destined to deploy outdated hardware before power is even connected.