“Many existing generative-AI-powered chat assistants are great for consumers, and you can do really cool things with them, but they lack a lot of the features that you would need to make them truly useful at work,” Selipsky says.

Amazon is making other moves to catch up in generative AI. In September, it said it would introduce ChatGPT-like capabilities to Alexa, its voice assistant. Earlier this month, Reuters reported that Amazon has dedicated a team to building a language model, code-named Olympus, that will be larger than OpenAI’s GPT-4, which powers ChatGPT, perhaps promising better capabilities. That suggests OpenAI will see more intense competition next year. Google is also working on a more powerful AI model, known as Gemini.

AWS also announced two new AI chips available to its customers today. Access to powerful processors is an essential component for building large language models, and Amazon, Google, and a number of startups are competing with Nvidia, whose GPUs are standard in AI projects.

Amazon says its new fourth generation of silicon for running AI models, Graviton 4, offers 30 percent better performance than the previous model, and that its second generation of chip for training AI models, Trainium 2, is four times as fast and twice as energy-efficient as its predecessor. As part of Amazon’s $4 billion investment in Anthropic, whose Claude chatbot competes with ChatGPT, the startup has agreed to use AWS silicon to train its models.