Google is ready to drop $40B on Anthropic — cash and compute

Google is reportedly planning to sink up to $40 billion into Anthropic — and it’s not just cash. A good chunk of that will be in compute credits, which is basically Google saying “use our infrastructure, please.”

This is a huge number. For context, that’s more than Microsoft has committed to OpenAI across multiple rounds, and it shows just how desperate the big tech companies are to lock down access to the best AI talent and models.

Let’s be real: Anthropic has been the darling of the safety-conscious AI crowd, but they’re also burning through cash at an alarming rate. Training frontier models is absurdly expensive, and running inference at scale is even worse. Google knows this, and they’re betting that Anthropic’s approach — especially with their new Mythos model — will pay off.

Speaking of Mythos, it’s been released in a very limited capacity, and it’s focused on cybersecurity. That’s an interesting play. Most AI labs are chasing general-purpose chatbots or code generators, but Anthropic is going after a niche that’s both high-stakes and potentially lucrative. If Mythos can actually detect and respond to zero-day exploits in real time, that’s a game-changer. But the limited release suggests they’re still ironing out kinks — or they’re being extra cautious, which is on brand for them.

This investment also tells me that Google is feeling the heat. They have their own models — Gemini, PaLM, whatever they’re calling it this week — but they haven’t been able to match the hype or performance of GPT-4 or Claude. Throwing money at Anthropic is a hedge. It’s Google saying “we can’t beat ’em, so we’ll buy a piece of ’em.”

The compute component is the part that interests me most. These credits are essentially a way for Google to lock Anthropic into their cloud ecosystem. If Anthropic trains all their future models on TPUs and Google Cloud, they’re not going to switch to AWS or Azure overnight. It’s a classic vendor lock-in play, but with a $40B price tag.

I’m curious how regulators will view this. The FTC has been sniffing around big AI deals, and a $40B investment in a competitor — even if it’s not a full acquisition — might raise eyebrows. But Google has deep pockets and good lawyers, so I expect this to go through with some concessions.

For now, Anthropic gets a war chest, Google gets a seat at the table, and the rest of us get to watch the AI arms race escalate even further. I just hope Mythos lives up to the hype, because if it doesn’t, that’s a lot of money down the drain.