Amazon's $5B Bet on Anthropic: Is It Worth It?

On April 20th, Amazon and Anthropic jointly announced something significant: Amazon is investing another $5 billion.

Combined with the previous $4 billion, Amazon’s total bet on Anthropic has now exceeded $9 billion. That’s not pocket change—even for a company of Amazon’s size, it’s enough to make the board discuss it seriously for several rounds.

After the news broke, I saw two completely different reactions. One camp said “Amazon is crazy, Anthropic isn’t worth this price.” The other said “This investment is incredibly smart—AWS finally has a chip to compete against OpenAI.”

My personal take? Both perspectives have merit, but both are also somewhat one-sided.

Let’s start with the “not worth it” argument. Anthropic’s current valuation is around $60 billion, but annual revenue (based on my estimates from public information) is probably in the $1-2 billion range. By traditional valuation models, that multiple is absurdly high. More critically, Anthropic is still losing money—and losing a lot. Anyone who understands large model training and inference costs knows what I’m talking about.

But the “worth it” camp uses a different calculation. They’re not measuring how much money Anthropic itself can make, but how much this investment can help AWS make.

Here’s some context: Microsoft’s deep partnership with OpenAI has given Azure a massive advantage in AI cloud services. Many companies choose Azure not because Azure is inherently better, but because it’s the only place to run the latest GPT-4. This kind of “model-as-a-service” exclusivity is something traditional cloud providers can’t buy with pricing or service quality alone.

Amazon is now replicating this strategy. Through its deep investment in Anthropic, AWS gains priority access to Claude models, customization capabilities, and joint sales rights. In other words, when customers say “we want to use Claude,” AWS can confidently respond “come to us for the best experience.”

I noticed one detail: the investment agreement mentions “specific commercial milestones.” This means the $5 billion isn’t a one-time transfer—it’s tied to Anthropic’s performance. This structure is risk hedging for Amazon. If Anthropic performs well, Amazon is happy to invest more. If growth disappoints, future commitments can be reduced.

Another point many people miss: Amazon isn’t “saving” Anthropic—it’s “locking in” Anthropic.

In today’s AI industry, independent foundation model companies are in an awkward position. OpenAI has Microsoft’s backing, Anthropic has Amazon and Google’s dual investment, DeepSeek has ByteDance behind it, and smaller model companies are either being acquired or slowly marginalized. Pure independent development paths in capital-intensive foundation models are essentially non-viable.

So the essence of this investment is a “political marriage” between Amazon and Anthropic. Both sides get what they need—Anthropic gets money and cloud resources, Amazon gets model capabilities and market leverage.

As for whether it’s worth it? I think it depends on your time horizon. If you’re looking at ROI next year, definitely not. If you’re looking at AWS’s AI cloud market share five years from now, this might be one of Amazon’s best investments ever.

Of course, the risks are obvious. Can Anthropic maintain technical leadership? Can Claude’s user experience catch up to ChatGPT? These are unknowns. But that’s how business works—there are no certain bets, only trade-offs between risk and reward.

Amazon is placing a pretty big bet this time.