Google is reportedly ready to drop $40B on Anthropic — mostly in compute

The AI infrastructure spending spree isn’t slowing down. According to a report from TechCrunch, Google is planning to invest up to $40 billion in Anthropic, the maker of Claude. The kicker? A huge chunk of that isn’t cash — it’s compute credits on Google Cloud.

This is the kind of deal that makes you stop and think about where the real bottleneck in AI actually is. For years, everyone focused on model architectures, training techniques, and data quality. Now the conversation has shifted squarely to compute capacity. You can have the best research team in the world, but without enough GPUs and TPUs to run experiments at scale, you’re stuck.

$40 billion is a staggering number, even by Google’s standards. To put it in perspective, that’s more than the entire GDP of some small countries. But it’s not all going into Anthropic’s bank account. The structure reportedly leans heavily on Google Cloud credits, which means Anthropic gets the hardware it needs to train and run its models, while Google locks in one of the most promising AI labs as a long-term cloud customer. It’s a smart play on both sides.

This news comes right after Anthropic quietly released Mythos, a new model focused on cybersecurity. I say “quietly” because the launch was deliberately limited — no big press tour, no flashy demos. Just a model that, from what I’ve heard, is surprisingly good at finding vulnerabilities and writing secure code. It’s a niche play, but a smart one. Cybersecurity is one of those areas where the ROI on AI is immediately clear, and where mistakes are costly. If Mythos can actually reduce false positives in threat detection or speed up patch development, enterprises will pay a premium.

The timing of the investment announcement relative to Mythos’s release isn’t a coincidence. Anthropic is signaling that it’s not just competing on general-purpose chatbots anymore. It’s building specialized models for high-value verticals. And Google is signaling that it wants to be the infrastructure provider for that entire ecosystem, not just a competitor with Gemini.

I’ve been watching this space long enough to remember when cloud providers used to fight over startups with $100K credits. Now we’re talking about tens of billions. The scale has shifted so dramatically that it’s hard to wrap your head around. But the underlying dynamic is the same: whoever controls the compute, controls the future of AI.

That said, I do wonder about the concentration risk here. If Anthropic is essentially funded by and running on Google’s infrastructure, how independent can it really be? Google has a history of investing in AI companies and then slowly pulling them closer. DeepMind started as a separate entity, and now it’s fully integrated. I’m not saying the same will happen to Anthropic, but the pattern is worth noting.

For now, though, this deal — if it goes through — cements Google’s position as the landlord of the AI industry. Microsoft has OpenAI, Amazon has Anthropic (wait, no, Google has Anthropic now), and everyone else is scrambling for scraps. The consolidation is real, and it’s only going to accelerate.

Mythos itself is interesting but not revolutionary. I’ve tested a few cybersecurity AI tools before, and the hype usually outpaces the reality. But the fact that Anthropic is willing to release a specialized model at all suggests they’re thinking beyond the Claude brand. That’s a good sign for anyone who wants to see more diversity in AI products.

I’ll be watching to see how Mythos performs in real enterprise deployments. If it delivers, this $40B bet starts to look like a bargain. If not, well, Google has deeper pockets than almost anyone else on the planet. They can afford a few expensive lessons.