This article is from Big Technology, a newsletter by Alex Kantrowitz. Amazon's absence from the recent surge in generative AI has raised some questions. Despite being the mastermind behind Alexa, it seemed to relinquish its position when OpenAI launched ChatGPT. However, rather than sitting on the sidelines, Amazon is carefully strategizing its approach. Instead of focusing on just one AI product, it aims to have a stake in all of them. Furthermore, it is not shy about its ambitions. During an interview at the recent cloud service's summit, Matt Wood, the Vice President of Amazon Web Services, expressed his belief that the AI part of their cloud-computing business could surpass the rest of AWS combined in a couple of years. Amazon's plan is not centered around releasing a single product or a large language model. Their goal is to empower developers using generative AI to create a wide range of products utilizing any model they choose. They will provide model access, customization, and computing power, while profiting from the developers' creations. Andrew Lipsman, a principal analyst at Insider Intelligence, sees this as a lucrative business opportunity. Amazon is strategically focusing on where the profits lie. At the core of their efforts is their new product, Bedrock. Available within AWS, Bedrock allows developers to choose from various AI models, including those from Anthropic, AI21 Labs, and Stability AI. With these models, developers can create their own products, such as AI chatbots, and run them on AWS' infrastructure. For instance, Bloomberg built its financial information bot, BloombergGPT, using a precursor to Bedrock called SageMaker. To achieve this, Bloomberg loaded decades of unstructured financial data and analytics into AWS, incorporated additional training material, and fine-tuned the model. Bedrock aims to streamline this process by offering preloaded models in a catalog.

Once a product is developed, Amazon continues to provide support. When users engage with BloombergGPT, Amazon's storage comes into play, granting them a share of the process. Wood explains, "We get paid by providing compute capacity to actually do the model training, and for providing the access to the large amounts of storage that are needed. You may train a model once a month, once a week, but you're going to be running predictions and inference and chatting with that model hundreds, thousands, tens of thousands of times a day. " While competitors like Microsoft and Google offer similar capabilities, Amazon possesses several advantages. The absence of its own consumer chatbot, coupled with its neutrality and pragmatism, distinguishes it from its peers. This may be appealing to developers seeking more customization or guarantees regarding data privacy, which is a pressing issue for many. Additionally, Amazon benefits from the fact that numerous internet companies already store their data on its cloud platform. "You'd be surprised how many customers have exabytes of data on AWS, " says Wood. An exabyte equates to one billion gigabytes. Despite Amazon's own AI model named Titan, offered alongside other options, the company positions itself as being relatively neutral. Moreover, Amazon develops its own AI-specific chips for improved computing performance, though they are not sold like Nvidia's. These efforts are aimed at enhancing the core service offering. Amazon learned the value of being a frontrunner early on. It established its dominance in cloud services and remains a market leader. However, in the field of AI, it is playing catch-up, with Microsoft seemingly in the lead through its partnership with OpenAI, which boasts 11, 000 customers utilizing its generative AI service. Nonetheless, Amazon is entering the race with a plan that could prove successful regardless of which model or product eventually comes out on top. Wood notes, "We are three steps into a marathon race. And I don't think anybody without a smile on their face could call a winner three steps into a marathon, right?"