Why Nvidia Stock (NVDA) Is Not an AI Bubble

The last few weeks for Nvidia (NVDA) have brought a sharp slowdown in the stock—not because of any real weakness in the business, but mainly due to a mix of macro-driven worries. On the one hand, investors fear that the AI hardware market may no longer rely exclusively on Nvidia, with Meta (META) and Alphabet (GOOGL) testing their own chips instead of Nvidia GPUs (even though, in reality, both architectures end up coexisting).

On the other hand, you have natural profit-taking after massive appreciation in tech, sector rotations that hit the more stretched names, and the usual fear of an “AI bubble”—a trio of factors that have weighed on NVDA in the short term.

But zooming out and focusing on three key points, Nvidia’s long-term thesis still looks structurally intact, with clear micro differences compared to anything we saw during the dot-com bubble. In other words, there’s no real reason for structural panic—quite the opposite. I believe the compounder effect in NVDA will ultimately win out, and taking advantage of periods of weakness to build a long-term position remains the smarter move, which is why I maintain a Buy rating on Nvidia shares.

Nvidia Powers Every Dollar Spent on AI

To begin with, we can highlight Nvidia as having the highest “implicit rate” across the entire AI economy. Given that the company does not depend on a single model, app, or company, Nvidia’s product risk is low, with only volume risk. After all, every dollar invested in AI by any player (from Microsoft (MSFT) to OpenAI and Grok) ends up going to Nvidia GPUs.

As a result of its unique moat, Nvidia has managed to expand its EPS by 2,220% and its free cash flow by 1,930% over the last three years. The strong appreciation in Nvidia’s market cap over the same period was 792% as of the moment I write this, which is more modest than the EPS and FCF expansion, showing that the company’s valuation movement was not speculative. 

Drawing a parallel with the dot-com period, this scenario was just the opposite of the 2000s. At that time, revenues were thin, profits were negative, and multiples expanded without fundamentals. Now, the supercycle of demand is real, cash generation is monumental, and above all, profits are keeping pace with multiples.

Nvidia Becomes the Full AI Stack

But Nvidia’s advantage goes well beyond being the default destination for AI spending. Volume risk tends to shrink as the company moves toward a strategic verticalization of the AI stack, layering hardware, software, and full systems into a single offering.

In the trendy AI space, what really creates lock-in is CUDA (the programming ecosystem), the inference engine, Nvidia’s proprietary libraries, and the broader toolchain that makes it easy to build on Nvidia and almost impossible to migrate away from it. When an enterprise chooses Nvidia, it’s not choosing a chip—it’s selecting an entire ecosystem. No company in the dot-com era had anything close to this level of dominance, this depth of moat, or this kind of irreplicable software advantage.

Even if the AI supercycle slows, Nvidia is still positioned to capture the future digestion phase. The risk of structural decline is lower for Nvidia than for the hyperscalers, and if cloud players end up building capacity to extreme levels, Nvidia benefits twice, from the initial build-out and from the refresh cycle, since GPUs naturally become obsolete much faster than fiber. Back in the dot-com era, fiber optics lasted ~20 years. GPUs today last five to six years, as some sources claim. In other words, as soon as GPUs age, Nvidia gets to sell everything all over again.

Nvidia Is Built to Outlast Any AI Bubble

It’s entirely plausible that, amid the AI boom, players are taking excessive risks, piling on too much debt, or relying on overly aggressive projections—whether from OpenAI, Oracle, or other big names. But the key insight is simple: the boom is financed by others, while the upside is captured by Nvidia.

Nvidia doesn’t need to fully believe in OpenAI’s business plan, nor does it need to bet on Oracle’s (ORCL) massive RPO, or buy into the latest hyperscaler ambition cycle. If everything works, Nvidia wins. If things fail, Nvidia has already won. In practically every scenario—whether the winner ends up being Trinity/Anthropic, OpenAI, Meta, or anyone else—Nvidia still comes out on top. And that’s exactly what makes Nvidia the most asymmetric way to play AI without needing to guess who the eventual champion will be.

Is NVDA a Buy, Hold, or Sell?

Optimism on the Street toward NVDA is practically unanimous. Over the past three months, 39 out of 41 analyst ratings have been Buys, with just one Hold and one Sell. The average price target sits at $257.72, which implies a potential upside of about 43% from the latest price.

See more NVDA analyst ratings

The Case for Staying Long Nvidia

Linking Nvidia to a bubble similar to the dot-com era may sound intuitive at first glance, but a closer look at the microdynamics shows that the comparison falls apart. Beyond delivering earnings and cash flow growth that not only keep up with its valuation but also exceed it, Nvidia carries far less structural risk than the companies building the AI infrastructure. The real risk sits with those taking on massive CapEx and aggressive assumptions—not with the company providing the “compute” that powers everything.

Unlike in 2000, when the market was pricing in a distant dream with no operational backbone, Nvidia’s growth today is anchored in real demand, strong margins, and a software moat that remains practically untouchable. And even if certain parts of the AI ecosystem end up forming a bubble—whether through excessive spending or overly optimistic forecasts—Nvidia is not at the center of that risk. It’s the one selling the shovels, not the one digging the hole. Historically, companies in that position tend to navigate cycles with far more resilience.

All of this reinforces a simple point: Nvidia remains a long-term compounder whose fundamentals overpower any short-term noise, making a Buy stance not just reasonable, but rational.