The Latest AI Developments In 60 Seconds

As the tempo of AI newsflow approached the frenzied rollercoaster pace of geopolitical headlines during the biggest oil shock in decades, it’s becoming easy to get lost in all the latest developments and drama surrounding OpenAI, Anthropic, Nvidia, government blacklists, the AI circle jerk, sentinent killer robots, and so on…

To help readers keep on top of things, we are launching a brief AI news roundup, which should help you get up to speed in under 60 seconds. 

Here are the four main things you need to know: 

And just like that, Anthropic goes from Pentagon supply chain risk to $13B anchor tenant of AWS: Amazon’s fresh $5B investment brings its total Anthropic commitment to $13B – with Anthropic pledging $100B+ in AWS cloud spend over 10 years in return, securing 5GW of compute capacity across Tranium2 through Tranium4. On the government track, NSA is reportedly deploying Anthropic’s Mythos model despite the DoD designation – a contradiction that speaks to how deeply embedded Claude has become in mission critical workflows. And perhaps the clearest signal of where employees think this is going: Anthropic’s recent tender offer fell short of the $5-6B investors had lined up – not because demand was weak, but employees choosing to hold, perhaps betting the public listing will price meaningfully higher.
 
OpenAI, meanwhile, is cutting… not expanding. Kevin Weil and Bill Peebles both departed as OpenAI pivots away from compute heavy side quests towards enterprise monetization and a forthcoming superapp. The Codex revamp signals the same thesis: agentic workflow ownership over model novelty. Both companies are refining narrative and product surface, and capital structure simultaneously – the ARR accounting dispute where OpenAI internally accused Anthropic of overstating revenue metrics signals the positioning war is intensifying.
 
But the most consequential bet of the week may not be in software at all. Jeff Bezos is close to finalizing a $10B funding round for Project Prometheus – his physical AI lab valued at $28B, with JPM and Blackrock among investors per the Financial Times. While Anthropic and OpenAI race to own the enterprise workflow layer, Bezos is making a different wager: that the next frontier of AI Value Creation is in the physical world – manufacturing, aerospace, robotics, logistics – where the training data isn’t scraped from the internet but locked inside the factory floor. Not to mention, this is the first time Bezos has held an operational role since leaving Amazon in 2021.
 
And zooming out, the Private Capital machine isn’t slowing. Sequoia raised $7B under new co-stewards Alfred Lin and Pat Grady, nearly double its prior $3.4B comparable fund – for late stage AI expansion. Accel followed with $5B, deploying $4B into a Leaders Fund targeting at least 20 checks averaging $200M each, explicitly naming robotics and defense alongside AI software. Taken together: $12B+ of late stage conviction in a single week, with physical AI now sitting at the center of both mandates. With Capital is concentrating, Manger Selection now matters more than vintage year timing.

Source: UBS