Amazon Cloud Launches New AI Chip to Challenge Nvidia & Google
Amazon.com Inc.’s cloud unit swiftly advanced to launch the latest iteration of its artificial intelligence chip, revitalizing its initiatives to offer hardware that can compete with offerings from Nvidia Corp. and Google. In a recent interview, Dave Brown, a vice president with Amazon Web Services, announced that the accelerator, known as Trainium3, has been installed in several data centers and will be accessible to customers starting Tuesday. “As we get into early next year we’ll start to scale out very, very quickly,” he stated. The chip push represents a crucial aspect of Amazon’s strategy to differentiate itself in the realm of AI. AWS stands as the foremost provider of rented computing power and data storage. However, it has faced challenges in achieving that level of dominance among top AI tool developers, as certain companies choose to partner with Microsoft Corp., which maintains strong connections with ChatGPT creator OpenAI, or Alphabet Inc.’s Google. At 11:09 a.m. in New York, Amazon shares experienced an increase of 1.6 percent, reaching a price of $237.71, while Nvidia shares reduced their gains and AI chip competitor Advanced Micro Devices Inc. fell to a session low.
Amazon aims to attract businesses in search of a deal. According to the company, Trainium chips are capable of powering the intensive calculations behind AI models more cheaply and efficiently than Nvidia’s market-leading graphics processing units. “We’ve been very pleased with our ability to get the right price performance with Trainium,” Brown stated. Amazon is set to release Trainium3 approximately one year following the deployment of its predecessor accelerator, a rapid pace by chip industry standards. “The main thing we’re gonna be hoping for here is just that we don’t see any kind of smoke or fire,” an AWS engineer remarked humorously when the chip was first activated in August. The swift turnaround aligns with Nvidia’s commitment to unveil a new chip annually. There’s a catch, however: Amazon’s chips lack the extensive software libraries that facilitate the quick deployment of Nvidia’s graphics processing units. Bedrock Robotics, a company utilizing artificial intelligence models to facilitate the autonomous operation of construction equipment, operates its infrastructure on AWS servers, but relies on Nvidia chips for model development, according to Kevin Peterson, who stated, “We need it to be performant and easy to use. That’s Nvidia.”
Numerous Trainium chips currently in operation are allocated to Anthropic, located within data centers in Indiana, Mississippi, and Pennsylvania. AWS stated earlier this year that it had connected over 500,000 of them to assist the AI startup in training its latest models and plans to allocate 1 million of the chips to Anthropic by the year’s end. Amazon is wagering that Anthropic’s success, in conjunction with its own AI services, can attract other companies, though Amazon has revealed a limited number of significant customers for the chip, prompting analysts to evaluate Trainium’s effectiveness. Anthropic is utilizing Google’s Tensor Processing Units and has secured an agreement earlier this year with the Alphabet unit, granting the startup access to tens of billions of dollars in computing power. During an interview Matt Garman stated that the company’s relationship with Anthropic was “incredibly strong,” noting that the startup’s computing demand requires it to use multiple providers. Amazon made the chip announcement at re:Invent, its annual user conference, which in recent years has become a rolling advertisement for AI services where Amazon courts builders of cutting-edge tools and companies that might want to pay for access to them.
On Tuesday, Amazon announced updates to its primary line of AI models, known as Nova, with new Nova 2 products featuring a variant named Omni, capable of receiving text, image, speech, or video inputs and responding with both text and images. Similar to its chips, Amazon has endeavored to persuade customers regarding the performance of its models relative to their price, although Nova models have not historically been frontrunners in industry benchmarks assessing model performance. “The real benchmark is the real world,” said Rohit Prasad, who leads much of Amazon’s model development and its Artificial General Intelligence team, expressing confidence that the new models will be competitive. The company also intends to allow customers to incorporate additional data when personalizing Amazon’s models. Nova Forge, an innovative product, enables advanced users to access versions of Amazon’s Nova models prior to the completion of their training and tailor them with their own data. Reddit Inc. is employing Nova Forge to develop a model capable of evaluating whether a post on its platform breaches safety policies, with CTO Chris Slowe noting that “the fact that we can make it an expert in our specific area is where the value comes from.”
