A short YouTube recap by Matt Wolfe summarizing major AI announcements from AWS re:Invent: new high-memory/high-bandwidth servers (including systems using Nvidia GB300), AWS tooling for building custom frontier models (Amazon Nova Forge, Bedrock reinforcement fine-tuning, SageMaker AI customization), and AWS's agent core platform focused on building safer AI agents. Overall theme: infrastructure (bigger chips), bespoke models, and agent safety are the leading trends going into 2026.
Chapters
So here are all the top news headlines you missed from AWS Re Invent first, the biggest AI companies are going all in on hardware. AWS revealed two new ultra servers. The first is more efficient and has way more memory bandwidth than before. And the second is powered by Nvidia's GB300 chips, their most advanced architecture yet. These are built for massive scale and are basically going to be the backbone of AI for the next decade. But hardware is only half the story. Every company is trying to build their own custom AI model and big tech companies are responding responding to the demand with tools made to help with the exact purpose. So for example, Amazon just announced the new Amazon Nova Forge system for building custom frontier models, reinforcement fine tuning through Bedrock and serverless model customization in SageMaker AI. And lastly, AI agents are still the biggest thing in tech right now. We've been talking about AI agents all year and we're still talking about them at the end of the year, and we're probably going to be talking about them in 2026. Everybody wants to build an AI agent and AWS just announced that that their agent core platform can not only help you build agents, but build them safely. So in short, bigger chips, more customized models, and safer agents. Put it all together and you've got a pretty exciting recap of what are the most important topics in AI right now. We're just going to have to wait and see what 2026 has in store for us. It's going to be exciting.
