What Anthropic’s S-1 Reveals About the Next Decade of AI Spending

When a company files to go public, its S-1 shows how the business really runs. Anthropic’s filing does that, and it also shows something about the companies around it. The AI lab is on track for a debut valuation near $2 trillion, but the more useful detail may be what it plans to spend: roughly $518 billion on computing infrastructure over the next decade.

That figure covers only what Anthropic itself will buy. Many businesses prefer to run their own Claude instances on leased servers to keep their data separate, which means even more infrastructure has to be built beyond Anthropic’s own plans. Three companies sit at the center of that demand, each in a different role.

Broadcom (AVGO): Custom Computing Power for AI

Broadcom is one of Anthropic’s primary computing unit providers. As AI workloads grow, so does the demand for the chips that handle them, and Broadcom is positioned to benefit as more computing power is built out over the coming years.

Valuation is a point in its favor. Broadcom trades for less than 20 times next year’s expected earnings, a modest multiple for a business tied to one of the largest spending plans in tech. Anthropic has confirmed that the AI data center build-out is expected to last for many more years, which supports the idea that demand for Broadcom’s chips has room to run.

Alphabet (GOOGL): Earning From Training and From Usage

Anthropic doesn’t operate its own data centers. It runs its workloads on servers hosted by Google Cloud and Amazon Web Services. That gives Alphabet two ways to earn. It charges Anthropic for the computing used to train models, and it earns again each time Claude is used, because Google Cloud is the one running those workloads.

The result is a revenue stream that behaves like a subscription, producing income whenever AI is used and growing as usage grows. The market understands this, so Alphabet trades at a premium to the chip suppliers. In exchange, investors get a dominant, mature business with consistent results, the kind that works well as the core of a portfolio.

Nvidia (NVDA): The Largest Name in the Chain

Nvidia is also among Anthropic’s primary computing unit providers, and it is the largest company in this group, with a market cap near $4.9 trillion and a share price around $205. The stock is roughly flat over the past year and sits below its 52-week high of about $243.

Its gross margin is about 75%, and its forward earnings multiple is below 20. Nvidia benefits from the same dynamic as Broadcom: Anthropic’s own spending is only part of the picture, since businesses hosting Claude on leased servers will need substantial additional computing power.

Choosing Where to Stand in the Chain

Broadcom and Nvidia are paid when computing power is built, and Alphabet is paid when that power is used. The chip names trade at lower earnings multiples, while Alphabet’s recurring revenue earns it a premium. Because all three depend on the same long-term build-out, they are likely to move together if enthusiasm for AI cools. Investors should decide whether they prefer lower valuations tied to construction or steadier income tied to usage.



NEXT: