In a major strategic shift away from total reliance on Nvidia, OpenAI has signed a multi-year agreement with Cerebras Systems to use computing hardware equivalent to 750 megawatts of power consumption, in a deal valued at more than $10 billion.
Article contents:
- Introduction
- Massive deal details
- Speed and inference
- Unique Cerebras technology
- OpenAI’s diversification strategy
- Market impact and competition
- FAQs
Introduction
OpenAI is striving to enhance its artificial intelligence infrastructure like never before, and it has found a new partner to achieve this. The company announced on Wednesday that it has signed a massive deal with Cerebras Systems, a semiconductor startup famous for its giant chips, to provide massive computing capabilities that will support the next generation of AI models.
Massive deal details
According to a joint statement, OpenAI will use Cerebras hardware as a computing resource to achieve faster response times when running AI models. The infrastructure will be built in multiple phases “through 2028” and hosted by Cerebras. While terms were not officially disclosed, people familiar with the matter estimated the size of the deal at more than $10 billion.
Speed and inference
Greg Brockman, co-founder and president of OpenAI, said: “This partnership will make ChatGPT not only the most capable AI platform but also the fastest in the world.” He added that this speed will help unlock “the next generation of use cases and bring the next billion users to AI.” This collaboration focuses specifically on the “inference” phase, the process by which the model responds to user queries, where Cerebras products excel.
Unique Cerebras technology
Cerebras takes a unique approach to processing information using massive wafer-scale chips, which differ radically from traditional graphics processing units. Recent work done by Cerebras to support an OpenAI model showed that it operates 15 times faster than traditional hardware.
OpenAI’s diversification strategy
This deal comes as part of an unprecedented bet by the tech industry that insatiable demand for power-hungry AI tools will continue unabated. It is also an attempt to diversify suppliers away from complete reliance on Nvidia. In October, AMD said it would deploy 6 gigawatts of GPUs for OpenAI, and the company is also developing its own custom chip with Broadcom.
Market impact and competition
Andrew Feldman, founder and CEO of Cerebras, said: “They are choosing a new and different architecture because it is faster and delivers value for them. This deal launches us into the big leagues and brings high-speed inference to the mainstream.” The news comes as Cerebras is in talks to raise funds ahead of a potential initial public offering, at a valuation that could reach $22 billion.
FAQs
Question: Why is the Cerebras deal important?
Answer: Because it represents a strong alternative to Nvidia chips, relying on “giant wafer” technology that provides tremendous response speed for AI models.
Question: What is the amount of power consumed in the deal?
Answer: The deal includes hardware consuming about 750 megawatts, a massive figure that reflects the scale of the infrastructure required to run modern AI.
Question: Will OpenAI stop using Nvidia?
Answer: No, this deal is part of a diversification strategy. OpenAI still relies on Nvidia and other companies and is working on developing its own chips as well.
