Amazon (AMZN) shares just ripped 25% higher in 30 days. The rally has been fast, vertical, and impossible to ignore. Now comes the uncomfortable question: Is the easy money already made?

Late last week, Meta revealed it will buy “tens of millions of cores” of Amazon’s Graviton5 CPUs in a multi-year deal. The target: AI inference and agentic workloads.

For two years, the AI hardware story has been almost entirely a GPU story. Meta’s order doesn’t kill that narrative — but it complicates it, powerfully. High-volume AI deployment is hungry for cost-efficient, custom-designed CPUs, and Amazon is moving fast to fill that slot. Meta’s signature on the contract is the clearest outside validation yet that Amazon’s homegrown chip strategy is a winning one.

Then came the Anthropic bombshell.

Almost simultaneously, the company behind Claude committed to spending more than $100 billion over the next decade on AWS capacity. The deal spans Graviton, Trainium2, Trainium3, and even future Trainium4 chips.

Amazon immediately deepened its financial ties, injecting another $5 billion. CEO Andy Jassy kept it simple: custom silicon offers high performance at materially lower cost. “That’s why it’s in such hot demand.”

The numbers are getting too big to ignore

Jassy recently peeled back the curtain on what this chip push adds up to. The internal chip business — Graviton, Trainium, Nitro — has already reached a $20 billion annualized revenue run rate, growing at a triple-digit pace. Trainium2 is effectively sold out. Trainium3 is nearly fully allocated. Trainium4, still in the future, has already locked in a substantial share of its capacity.

If Amazon chose to sell these chips to third parties, Jassy says that run rate would leap to roughly $50 billion. And that’s revenue alone. On the cost side, scaling in-house silicon is expected to save “tens of billions of dollars” a year in capex.

No wonder the market is re-rating the stock.

But this is where the math gets trickier.

A 25% rally means a lot of good news is already priced in. Amazon now trades at around 37 times trailing earnings.

At the same time, the company is guiding for roughly $200 billion in total capital expenditures in 2026. Even if every chip bet pays off, that level of spending will weigh on margins. Any slip in execution — or any delay in the payoff — could quickly deflate the multiple.

So, buy or wait?

The long-term logic is intact, and the Meta and Anthropic deals make it more tangible than ever. Amazon is turning its chip division from an internal utility into something that looks suspiciously like a profit engine.

But after a 25% surge in a single month, the stock is no longer offering much margin of safety. This is starting to look less like a short-term trade and more like a patient, multi-year holding. For investors weighing a fresh entry, the smartest move may be to keep it on the watchlist and wait for the next bout of market nervousness to offer a better price.