In order to make a bigger calculation, Altman and Yellow Commander reunited again.
On September 22, local time, OpenAI announced the signing of a letter of intent and a strategic partnership with NVIDIA, allowing it to use NVIDIA systems (meaning millions of GPUs) to build and deploy at least 10GW (GWh) of AI data centers to train and run the next generation of models.
OpenAI will become the preferred computing and networking strategic partner of the NVIDIA AI plant growth plan, ensuring that NVIDIA systems are always able to the first and best optimization effects for OpenAI models and infrastructure software.
Not only that, Nvidia also plans to gradually invest US $100 billion in OpenAI.
According to the phase-by-phase investment plan, when deploying the first GW computing power, NVIDIA will invest 10%. NVIDIA’s equity investment in OpenAI does not give it control, and the non-profit parent company of OpenAI will retain majority governance control.
Sam Altman, CEO of OpenAI, said: “Everything begins with computing, and computing infrastructure will be the foundation of the economy of the future.”
Nvidia founder and CEO Huang Yonglong said that from the first DGX supercomputer to the breakthrough of ChatGPT, Nvidia and OpenAI have pushed each other, jointly developing for a decade, this investment and infrastructure partnership marks the next leap, driving the next smart era.
According to reports, the first GW computing power target will be put into operation in the second half of 2026 using Nvidia's Vera Rubin platform. Nvidia and OpenAI will determine the details of this new phase of strategic partnership in the coming weeks.
At present, about 700 million people use ChatGPT every week, and Outman has recently repeatedly said that OpenAI needs larger GPUs to scale its business, and a shortage of computing power has led to better models that could not be provided.
Thanks to the strong growth of ChatGPT, OpenAI significantly increased its expected cash consumption to $115 billion by 2029, an increase of $80 billion over previously expected.To control the surge in costs, OpenAI seeks to develop its own data center server chips and facilities to support its technology.
In July, OpenAI deepened its cooperation with Oracle. Based on the "Stargate" plan, it added a new data center with a power consumption of 4.5 GW. Combined with the first "Stargate" base under construction in Abilene, Texas, OpenAI will develop The total capacity of the AI data center in China exceeds 5GW and can support the operation of more than 2 million chips.
It is worth mentioning here that in the previously announced first-quarter financial report, Oracle Bone Inscriptions predicted that its revenue would double in the next three years. Its remaining performance obligations (RPO, contracted but not yet recognised revenue) more than tripled in three months to $455 billion, while billions of dollars of deals are being negotiated.
In September, it was revealed that OpenAI planned to cooperate with American semiconductor company Broadcom to produce its first artificial intelligence chip next year, and OpenAI planned to use the chip internally rather than sell it to customers. At the same time, according to people familiar with the matter, OpenAI also plans to build an artificial intelligence data center with a power consumption of more than 1GW in India, and has conducted initial negotiations with potential partners. Ultraman, who is expected to visit India later this month, may disclose more details during his trip.
OpenAI is also working in the field of AI hardware.
According to foreign media reports, the company has obtained assembly contracts for at least one OpenAI device.In addition, OpenAI has also entered into negotiations with Goal’s stake in Apple products such as AirPods, hoping to supply components such as speaker modules for its future products.
OpenAI filed a new trademark application to the U.S. Patent and Trademark Office in January this year, involving human-made robots, XR, smartwatches, wearable devices and other hardware devices.OpenAI has also announced the acquisition of the AI hardware company of former Apple design director Johnny Ive for $6.5 billion and the establishment of a division for the company that will focus on developing AI-driven devices.
According to people familiar with the matter, OpenAI's planned products include similar display-less smart speakers, as well as products such as glasses, digital recording pens and wearable brooches. The first batch of devices may be released in late 2026 or early 2027.