By developing its own, specialized AI chips, OpenAI hopes to achieve greater independence and enormous cost savings.
• OpenAI enters into several chip partnerships
• Own AI accelerator also planned
• Independence and costs in view
In collaboration with the US semiconductor company Broadcom, OpenAI is working on developing its own AI accelerators. Through this chip partnership, the AI pioneer not only hopes to achieve more independence but also cost savings of 20 to 30 percent compared to NVIDIA hardware, reports the Bloomberg news agency.
Alliance with NVIDIA
In order to cover its huge demand for computing power, OpenAI apparently relies on several pillars: First, the largest AI alliance of all time was announced in September together with NVIDIA: The deal not only includes an investment of up to 100 billion US dollars in the ChatGPT developer but also the delivery of ten gigawatt systems from NVIDIA’s upcoming Vera Rubin generation.
Deal with AMD
Nevertheless, OpenAI obviously wants to remain independent of NVIDIA. For this reason, a long-term supply contract worth billions was concluded with AMD at the beginning of October. Starting in the second half of 2026, AMD will deliver hundreds of thousands of its AI chips or graphics processors (GPUs) with an output of six gigawatts to OpenAI over several years. As part of this agreement, the ChatGPT maker also has the opportunity to invest up to ten percent in AMD.
The world needs “a lot more computing capacity,” said OpenAI boss Sam Altman, commenting on the AMD partnership, according to “The Decoder”. He confirmed that his company relies on several hardware suppliers to meet the growing need for computing resources for future AI models.
OpenAI also relies on its own chips
But OpenAI is taking its quest for independence one step further and has agreed to collaborate with US semiconductor company Broadcom on customized AI accelerators with a power capacity of ten gigawatts. OpenAI designs the chips itself and Broadcom supplies the network technology.
OpenAI intends to use these chips only internally and not to sell them to third parties. This follows the strategy of other tech giants such as Google, Amazon and Meta, which have already developed their own specialized chips for AI applications. Of the usual $50 billion that OpenAI expects to spend to build a 1 gigawatt data center, around $35 billion will go to high-performance chips – this is where the hoped-for cost reduction comes in.
However, this path also brings with it considerable risks: on the one hand, the development of tailor-made chips requires specialized expertise, and on the other hand, investments worth billions are required. In addition, rapidly changing AI technologies threaten rapid obsolescence.
Editorial team finanzen.net
By the way: Alphabet A (ex Google) and other US stocks can even be traded on finanzen.net ZERO until 11 p.m. (without order fees, plus spreads). Open a depot now for free and secure a new customer bonus!
Selected leverage products on Alphabet A (ex Google)
With knock-outs, speculative investors can participate disproportionately in price movements. Simply select the lever you want and we will show you suitable open-end products on Alphabet A (ex Google)
The leverage must be between 2 and 20
Advertising