A Amazon is expanding its presence in the field of artificial intelligence, launching powerful new chipsets and a robust language model that, according to the company, can compete with the main competitors in the market.
The Seattle giant is connecting hundreds of thousands of your Trainium2 semiconductors em clusters designed to facilitate the training of broad language models used in generative AI and other machine learning tasks. The new infrastructure will quintuple the processing capacity of partner startup Anthropic, according to Amazon.
The company’s cloud services division, Amazon Web Services (AWS), began making the new chips available to customers this Tuesday, during the annual re:Invent conference.
READ MORE: Amazon invests another $4 billion in Anthropic, OpenAI’s rival
Andy Jassy, in his first participation at the event since taking over as CEO of Amazon in 2021, presented the new models called Nova. Capable of generating text, images and videos, the models represent Amazon’s latest effort to compete with companies like OpenAI and other developers of advanced language models that power chatbots and generative AI tools.
AWS, the world’s largest cloud computing provider, operates many of the servers rented by other companies to train artificial intelligence applications. Additionally, AWS makes templates created by third parties, such as Anthropic’s Claude and Meta’s Llama, available to its customers. However, to date, Amazon has not yet produced a broad language model that is widely recognized as competitive with OpenAI’s most advanced GPT models.
The models previously developed by Amazon, called Titan and released in the last two years, were generally smaller in scope. The new Nova models, some available immediately and others scheduled for next year, include a “multimodal to multimodal” version, capable of processing inputs in text, speech, images and video, generating responses in any of these formats.
READ ALSO: ChatGPT’s Rival: How to Use Anthropic’s Claude AI
Jassy highlighted that the company will continue to both develop its own models and offer third-party solutions. “Our goal is to provide the broadest, most complete functionality you can find anywhere,” he said.
Last month, Amazon announced an additional $4 billion investment in Anthropic. As part of the agreement, the startup committed to using Amazon’s cloud infrastructure and chips to develop its most advanced models.
The new chip cluster, called Project Rainierwill have “well over” 100,000 units, according to Gadi Hutt, executive at Annapurna Labs, Amazon’s chip division. The company says it expects the cluster to become the largest set of dedicated AI hardware in the world.
With the new chips – Amazon’s third generation of AI semiconductors – the company seeks to compete with Nvidia’s products, offering AWS customers an alternative for developing generative AI products. Currently, Nvidia graphics processors are the market standard for these tasks, although they are expensive and often in short supply.
Amazon also announced that it will offer customers computing power based on Nvidia’s new Blackwell chip starting early next year, further expanding its options for those looking for artificial intelligence solutions.
