OpenAI recently signed tens of billions of dollars in cooperation agreements with chip giants such as AMD and NVIDIA, which has attracted industry attention. When NVIDIA CEO Jen-Hsun Huang unexpectedly learned of OpenAI’s multi-billion dollar deal with rival AMD, and his company had just committed to investing up to $100 billion in OpenAI, OpenAI CEO Sam Altman said that similar large-scale transactions are still under negotiation.

It is reported that the cooperation between AMD and OpenAI is quite novel. AMD agreed to grant OpenAI up to 10% of the company's shares in batches based on future stock price growth and other performance conditions, in exchange for OpenAI's widespread adoption and assistance in developing its next-generation AI GPU chips in the next few years. This means OpenAI will become a shareholder of AMD. NVIDIA reversely invested in OpenAI, and NVIDIA became a shareholder of OpenAI.

Huang Renxun admitted in an interview with CNBC that this is the first time OpenAI has directly purchased AI hardware equipment from NVIDIA. In the past, OpenAI has been purchasing indirectly through cloud service platforms such as Microsoft Azure and Oracle OCI. In addition to GPUs, this direct sale also covers AI systems and networks, in order to help OpenAI realize its vision of building its own hyperscaler data center. But Huang Renxun also admitted that OpenAI currently does not have enough funds to undertake such a large-scale equipment purchase: "Each 1GW of AI infrastructure costs approximately US$50 to US$60 billion, including land, electricity, servers and equipment."

From 2025 to now, OpenAI has built a total of 10GW of data centers in the United States through the Stargate plan in cooperation with Oracle and SoftBank, with a cooperation amount of up to US$500 billion, and has signed an additional US$300 billion cloud service agreement with Oracle. At the same time, the data center construction scale signed by OpenAI and NVIDIA is at least 10GW, and the cooperation with AMD reaches 6GW. In addition, OpenAI’s “Stargate UK” plan in the UK and other European cooperation are also advancing. Some analysts say that the total amount of OpenAI’s infrastructure-related agreements this year has approached US$1 trillion.

Although these agreements enable OpenAI to acquire huge amounts of equipment in a "others pay the bill" model, some experts have questioned that the "revolving investment" structure between NVIDIA and OpenAI may cause financial risks: NVIDIA actually guarantees OpenAI's purchase of equipment in disguise by purchasing shares in OpenAI.

At the same time, Altman said in a podcast interview with venture capital firm Andreessen Horowitz that OpenAI is expected to have more large-scale cooperation in the next few months. Horowitz praised OpenAI's "impressive" innovation in transaction structures - OpenAI has repeatedly successfully acquired billions of dollars of equipment at low cost through innovative financial models. Altman is confident about future development: "We decided to bet on aggressive infrastructure expansion because we believe that future models and product capabilities will greatly promote market demand." According to reports, although OpenAI's revenue reached US$4.5 billion in the first half of this year, it is far from reaching a trillion-dollar scale. However, Altman firmly believes that all investments will eventually bring huge returns: "I feel more confident than ever about the future R&D route and economic value prospects."

But Altman also admitted that OpenAI cannot achieve an industry breakthrough of this scale on its own: "If you want to bet at this level, you need the joint support of the entire industry, or at least the major players, including upstream infrastructure, distribution and many links. We will carry out more cooperation and there will be more transactions in the future."

To sum up, OpenAI’s industry layout continues to expand, more blockbuster transactions will be announced soon, and the global technology industry is paying close attention to its next developments.

Related articles:

Trillions of AI infrastructure NVIDIA and OpenAI are frantically adding fuel