AWS will power OpenAI’s next-generation models in a $38 billion deal as the AI firm moves beyond Microsoft.
Summary
- OpenAI signed a $38 billion deal with Amazon to get access to its computing
- AWS will use its UltraServers, powered by Nvidia GPUs, for OpenAI’s models
- Skeptics warn that similar deals show increasing signs of an AI bubble
As demand for powerful infrastructure explodes, tech giants are forging multibillion-dollar deals to power AI models. On Monday, November 3, OpenAI signed a seven-year, $38 billion agreement with Amazon Web Services. The deal gives OpenAI access to AWS’s massive compute capacity.
“Scaling frontier AI requires massive, reliable compute,” said OpenAI co-founder and CEO Sam Altman. “Our partnership with AWS strengthens the broad compute ecosystem that will power this next era and bring advanced AI to everyone.”
OpenAI will have access to thousands of Nvidia GPUs. These include the GB200 and GB300 series Amazon EC2 UltraServers, optimized for AI compute. According to Amazon, this infrastructure can scale to tens of millions of CPUs, both for training and running the models.
“As OpenAI continues to push the boundaries of what’s possible, AWS’s best-in-class infrastructure will serve as a backbone for their AI ambitions,” said Matt Garman, CEO of AWS. “The breadth and immediate availability of optimized compute demonstrates why AWS is uniquely positioned to support OpenAI’s vast AI workloads.”
You might also like: Interview | Big tech is training AI on junk data: Intuition
OpenAI deals raise questions about an AI bubble
The deal is part of OpenAI’s $1.4 trillion investment to expand its infrastructure capacity. According to CEO Sam Altman, the firm plans to develop 30 gigawatts of power capacity. For comparison, a single nuclear power plant has about 1 gigawatt in capacity, while 30 gigawatts could power 26 million homes for a year.
OpenAI’s previous deals have raised questions about a potential AI bubble. Critics warned that increased AI firm valuations, as well as their capital expenditure, are not in line with AI demand and profitability.
Read more: Are SLMs the future of AI? Nvidia researchers think so. Here’s why
