Anthropic has developed an AI model called 'Claude' that many think is as good, or even better, than the other current leading AI models. So they have some credibility in their pitch to Amazon about 'Claude-Next', the AI they talk about beginning to automate large portions of the economy.
Anthropic estimates Claude-Next will require on the order of 1025 FLOPs, or floating point operations — several orders of magnitude larger than even the biggest models today. Amazon AWS computing resources will presumably be a big help with that. They are not just scaling up current AI, but say in addition they want to make their AI self-teaching to solve some of current AI's problems with mistakes/hallucinations.
Given all the concerns around Amazon and its monopolization of online retail, I'm surprised more in the media haven't cottoned on to it investing so heavily in a firm whose stated goal is automating large portions of the economy.