Google Bets $40B on Anthropic's Mythos Moment

welcome to Briefly, AI. A daily podcast made by AI, about AI, using a grumpy AI voice of a real human called Harry. Let's get into it.

Google just committed forty billion dollars to Anthropic. In cash and compute. That's not a partnership — that's a takeover bid without the paperwork. Right, let's get into it.

So, Google is writing Anthropic a cheque for up to forty billion dollars. That's billion, with a B. A mix of cash and cloud credits, which in practice means unlimited access to Google's compute infrastructure. Now, if you've been following along, you'll remember Anthropic's had a rather embarrassing fortnight. Their ultra-secure Mythos model — the one they said was too dangerous to release — got accessed by someone who wasn't supposed to have it. Bloomberg broke that story last week, and Anthropic's "responsible AI lab" brand took a bit of a kicking.

Here's the thing. Google isn't rescuing them. Google's locking them in. Forty billion buys you more than compute — it buys exclusivity, roadmap influence, and a very long leash. For Anthropic, it means they can stop worrying about where the next training run's coming from. For Google, it means they've just made sure Claude doesn't end up powering a competitor's entire product stack. If you're using Claude at work, this matters. Your AI assistant now has a very well-funded landlord, and landlords tend to have opinions about how the property gets used.

Now, speaking of Anthropic. They published something this week that's either brilliant or deeply unnerving, depending on your tolerance for letting robots handle your eBay account. It's called Project Deal. They gave Claude access to an internal marketplace where Anthropic employees could buy and sell personal belongings. And then they let the AI... negotiate. On behalf of humans. Autonomously.

So, Claude's browsing listings, making offers, counter-offering, closing deals. All without a human in the loop. According to Anthropic, it worked. People sold their old furniture, bought someone's bike, haggled over a desk lamp. The AI handled the whole transaction. Now, they're framing this as a research experiment. A controlled test of autonomous commercial agents. But let's be honest — this is a product demo dressed up as a paper. They're not asking "can Claude do this?" They're showing you it already does.

Why does that matter? Because if an AI can negotiate the price of your sofa, it can negotiate your mobile phone contract. Your insurance renewal. Your salary. And once that's normalised, the economics of consumer choice start to look very different. You won't be comparing prices anymore. Your AI will. And the businesses on the other side? They'll have AIs too. It's agent versus agent, all the way down. We're not there yet, but Anthropic just moved the goalposts a few metres closer.

And finally, a quick one that sounds boring until you realise what it actually means. Oracle just closed sixteen billion dollars in financing for a data centre in Michigan. The bonds sold out. Bank of America handled it. And the reason Oracle's building this enormous campus in the American midwest is to power applications for OpenAI.

Sixteen billion. That's more than some countries spend on infrastructure in a year. And it's going into a single facility, purpose-built to run AI workloads for one customer. Now, Oracle's been in the enterprise game for decades, but this isn't about databases anymore. This is about becoming the physical backbone of the AI economy. If you're OpenAI, you can't just rent cloud capacity and hope Amazon doesn't decide to compete with you next quarter. You need dedicated infrastructure. Long-term commitments. Oracle's betting that the future of AI isn't just about who builds the best model — it's about who controls the power and the racks.

For the rest of us, it's a reminder that AI's getting expensive in ways that don't show up in your ChatGPT subscription. Sixteen billion for one data centre. Forty billion for one partnership. These numbers aren't hype. They're capital allocation at a scale that means the people writing the cheques think this is the future. And they're not hedging.

That's your lot. Three stories, five minutes, and not a single use of the phrase "paradigm shift." If this was useful, pass it on. If not, blame the machine. See you next time.