GoKawiilStephen Schenck / Android Authority
TL;DR Amazon is reportedly developing Moonraker, an advanced AI agent that could make Alexa much more capable.
Unlike Alexa Plus, Moonraker can complete multiple connected tasks from a single request.
The project could cost over $100 million in GPUs in 2026, sparking internal debate over whether to delay or scale it back.
Amazon is launching a hotly contested race to develop AI agents for its Alexa platform, but internal disputes over a huge $100 million hardware bill could see the project scaled back before it even launches.
Business Insider reviewed internal planning documents that indicate the unreleased initiative is codenamed Moonraker. It’s an extension of Alexa Plus, Amazon’s generative AI assistant that already lets users complete actions such as hailing an Uber or buying tickets through Ticketmaster.
Moonraker is said to be able to understand compound requests and to perform several related actions in sequence, rather than stopping after one task. Let’s say you want Alexa to “order me a ride and text my friend.” Alexa can handle both, and you don’t have to tell the AI to do each one individually.
Companies including OpenAI, Anthropic, and Google are already racing to build so-called agentic AI systems that can reason through multi-step workflows, rather than just answering questions. If Amazon delivers on Moonraker, then Alexa would be much closer to functioning as a real digital assistant.
But getting there will not be a cheap trick. According to the internal documents, Moonraker alone could cost more than $100 million in GPUs in 2026, making it the costliest new effort tied to the Alexa Plus overhaul. The report also said some Amazon leaders have expressed alarm over the rising cost of the AI models that power the assistant, with one proposal saying the project could be delayed or scaled back to ease spending.
Amazon is testing the system with more advanced AI models from Anthropic and is preparing hundreds of Nvidia GPUs to enable more advanced reasoning and visual capabilities.
... continue reading