Google is pouring up to $40 billion into Anthropic. Here’s what that actually buys.

Google is reportedly planning to sink up to $40 billion into Anthropic, the AI company behind Claude. That’s not a typo. Forty billion. In cash and, more interestingly, in compute credits.

Let’s be real for a second. This isn’t just another investment round. This is Google essentially buying a guaranteed seat at the table for the next generation of frontier models. And it tells you everything about where the AI arms race is right now: it’s not about talent or algorithms anymore. It’s about who gets the GPUs.

The compute angle is the real story

Everyone focuses on the dollar figure, but the split matters. A significant portion of that $40B is in Google Cloud credits. That means Anthropic is locked into Google’s infrastructure for the foreseeable future. For Google, it’s a way to keep a major AI player off AWS and Azure. For Anthropic, it’s access to compute that they couldn’t buy on the open market anyway—NVIDIA’s H100s and the next-gen Blackwell chips are still absurdly hard to get in bulk.

This is the same play Microsoft pulled with OpenAI. Give them the cloud credits, get the models, and tie their success to your platform. Google is doing exactly that, just with a bigger check and a company that’s been more cautious about safety.

Mythos: the model that changes the conversation

The timing isn’t random. Anthropic just released Mythos in limited preview. It’s their first model designed specifically for cybersecurity—think automated vulnerability discovery, real-time threat analysis, and defensive code generation. I’ve been testing it, and honestly, it’s the first time I’ve seen an AI that doesn’t just talk about security but actually does something useful with it.

But here’s the catch: Mythos is compute-hungry. Like, really hungry. Anthropic has been open about the fact that running these inference workloads at scale requires infrastructure that doesn’t really exist yet outside of a few hyperscalers. The Google deal directly addresses that bottleneck.

Why $40B? That’s a lot of zeroes

Let’s put this in perspective. Anthropic has raised about $7.6B prior to this. Now Google is throwing nearly six times that at them. The valuation is reportedly north of $60B post-money. That’s more than most publicly traded companies.

I think Google is betting that Anthropic’s safety-first approach will pay off as regulation tightens. The EU AI Act is coming. The US is starting to have real conversations about licensing. If you’re the company that can credibly claim to have done alignment research and red-teaming from day one, you might get preferential treatment. Anthropic has that narrative locked down.

But there’s also a darker angle. Google has its own models—Gemini, PaLM—and they’re good. But they haven’t had the same cultural cachet as ChatGPT or Claude. Buying into Anthropic gives Google access to a brand that resonates with developers and enterprises in a way that Google’s own AI products sometimes don’t. It’s a hedge against their own internal teams not moving fast enough.

What this means for the rest of us

If you’re building on top of AI models, this deal should make you nervous. The concentration of compute power in a few hands is already a problem. Now Google and Anthropic together control a massive chunk of the world’s AI compute capacity. That’s not great for competition.

On the other hand, if Mythos delivers on its cybersecurity promise, we might finally see AI being used for defense rather than just generating marketing copy. That’s a win. But it’s a win that comes with a very expensive price tag, and we’re all going to feel the effects in terms of pricing and access down the line.

I’m not saying this is good or bad. It’s just the reality of where we are. The AI industry is consolidating, and the companies with the deepest pockets are making sure they don’t get left behind. Google just wrote a $40B check to prove it.