Topline
Amazon Web Services, Amazon’s cloud computing arm, announced it’s launching a new AI supercomputer built from its own machine learning chips that could be one of the largest used to train AI models—and tries to rival chipmaking giant Nvidia.
Matt Garman, CEO of Amazon Web Services, speaks during the WSJ Tech Live conference at the Montage … [+]
Key Facts
Amazon Web Services’ (AWS) new supercomputer will be an “Ultracluser” made of the firm’s Trainium chips, which are intended for AI programs and which the company hopes could rival the highly-in-demand graphics processing units (GPUs) from chipmaker Nvidia, the Wall Street Journal first reported Tuesday.
Named Project Rainier, the supercomputer will be finalized in 2025 and could break records for its size, Dave Brown, a vice president of compute and networking services, told the Journal.
Following the announcement, Amazon’s share price rose more than 1% to nearly $213.
AWS is also working on Project Ceiba in partnership with Nvidia, another supercomputer for AI programs that hosts more than 20,000 Nvidia Blackwell GPUs, which are chips used by data centers.
The Project Rainier announcement comes alongside news Amazon will release a new Ultraserver, a cloud server used to train AI programs that will use its more advanced Trainium chips, while Apple will start using its Trainium chips for its Apple Intelligence AI technology, Benoit Duplin, the senior director of machine learning and AI at Apple, said Tuesday.
Key Background
AWS has operated in the AI space since the technology’s early days, and it’s looking to capture the boom in demand for the technology. It has made major investments in building out its data center infrastructure and product lines to keep up with demand, and grown its relationship with collaborator Anthropic over two $4 billion investments, as it seeks to compete with generative AI giants such as OpenAI. Amazon also hopes to advance in this space through its Trainium silicon chips by aiming to supply potentially cheaper and more efficient chips.
Big Number
More than $100 billion. That’s how much Amazon—which has a market capitalization of more than $2.2 trillion—told the Journal in June that it will funnel into AI infrastructure over the next decade.
Crucial Quote
“Today, there’s really only one choice on the GPU side, and it’s just Nvidia. We think that customers would appreciate having multiple choices,” Matt Garman, the chief executive officer of Amazon Web Services, told the Wall Street Journal.
Tangent
AI models, the technology used by generative AI companies like ChatGPT or Anthropic, require massive amounts of power and energy to function. The models process large amounts of data using a complicated algorithm that requires the fastest tech available and high-powered GPUs. Chipmaker Nvidia has seen astronomical success in recent years as the AI boom fueled demand for its GPUs and chips over those produced by its rivals. It is now one of the most valuable companies in the world, with a market cap of more than $3.3 billion, and is the clear market leader.
