When Nvidia trades cheaper than it did in 2018 and Amazon looks like a worse bet than Walmart, the market is telling you something about AI's economics.

The Summary

  • Nvidia shares hit their lowest forward earnings multiple in seven years, while Amazon trades at 2008 crisis-level valuations despite 12% revenue growth versus Walmart's 5%
  • Microsoft and Oracle valuations have converged for the first time in a decade, suggesting the market no longer prices in AI infrastructure premium
  • This isn't general tech pessimism. It's selective repricing of companies closest to AI infrastructure spend

The Signal

The market is doing something it almost never does: it's pricing in that the people selling shovels during a gold rush might not get rich. Amazon trading at a discount to Walmart while growing 2.4x faster isn't a blip. It's a statement about cloud margins in an AI-first world.

Here's what the valuation compression means. When Nvidia's forward P/E drops to 2018 levels, investors are saying one of two things: either AI infrastructure demand will crater, or the margins on that demand are about to get squeezed hard. Given that every enterprise on Earth is still trying to figure out their AI strategy, cratering demand seems unlikely. That leaves margin compression.

Look at the pattern. The companies getting repriced are the ones monetizing AI compute, storage, and chips. The companies not getting repriced are the ones using AI to build actual products people pay for. The market is starting to understand that infrastructure commoditizes fast. Building and selling picks and shovels worked in 1849 because there was one way to mine gold. In 2025, there are seventeen ways to train a model, twelve chip architectures, and three hyperscalers racing each other to zero margin.

Microsoft converging with Oracle's valuation tells you the market no longer believes Azure's AI advantage is structural. Oracle has been quietly building AI infrastructure without the hype tax. The valuation gap closing means investors think the moat was temporary.

The Implication

If you're building in AI, this valuation reset is a map. The infrastructure layer is getting commoditized in real time. The value is moving up the stack to companies building agents that do actual work, not companies renting the GPUs those agents run on. Watch where the smart money goes next. It won't be into pure infrastructure plays. It will be into vertical AI products with pricing power and companies that own the relationship with the end user, not the data center.

For OpenAI and Anthropic's IPO prospects, this is a warning shot. If the market won't pay a premium for Nvidia, it's not paying 50x revenue for a foundation model company burning $5 billion a year.

Source: The Information