Nvidia (NVDA 2.70%) has been one of the best performing stocks on the market over the past two years, and the catalysts that pushed it higher are still in place. But after its sharp rise, is Nvidia stock still a wise buy at its current levels, or would those holding shares be advised to sell and take some of the profits?

There are valid arguments for both views.

The selling point: How long will this wave of demand last?

Nvidia’s rise is directly linked to the artificial intelligence (AI) arms race. Its main products are graphics processing units (GPUs) – parallel processors that are excellent for handling large and complex computing tasks that can be easily broken down into many smaller tasks that can be completed independently and simultaneously. When you connect GPUs in clusters, you get a computing platform that can handle certain types of incredibly complex workloads at blazing speeds – and those are exactly the types of workloads that AI systems generate.

As AI companies and cloud computing providers rushed to meet growing demand for computing power, Nvidia’s revenue went through the roof. In recent years, quarterly sales have often tripled year-over-year. However, due to more difficult year-on-year comparisons, its stellar growth is starting to slow slightly. This slowdown in growth makes sense, but the bigger question is: Can Nvidia maintain its overall revenue at this level?

As companies purchase these GPUs to rapidly expand their AI computing capacity, there will come a time when demand will be met. At that point, Nvidia’s sales could collapse as companies only buy replacement GPUs or gradually increase capacity. This could be a big problem for Nvidia, as revenue levels in recent quarters are well above those of the past.

NVDA sales data (quarterly) from YCharts.

This also illustrates the cyclical nature of the chip business. Nvidia has gone through several boom-and-bust cycles in its life as a company. If demand for AI wanes, investors could be in trouble.

But has Nvidia built a sufficient sales base to offset this cyclicality?

The selling point: New technologies will continue to boost demand beyond 2025

GPUs don’t last forever. They usually need to be replaced after about three to five years. This means that companies that have recently expanded their computing infrastructure and want to maintain computing power in the long term regularly have to spend large sums on new hardware.

AI expansion has been underway for two years and many companies are still ramping up their AI computing power, meaning 2025 will be another year of strong demand. That brings investors to 2026, where the natural replacement cycle begins for the GPUs purchased at the start of the generative AI era. But there could also be other reasons for companies to upgrade.

First, the semiconductor chips in these Nvidia GPUs are manufactured by Taiwan semiconductor manufacturing (TSM 0.59%). Taiwan Semi continues to innovate in the process node space, enabling chip designers like Nvidia to create denser, more powerful chips.

TSMC expects its chips built with its next-generation N2 process node to be 25% to 30% more power efficient than previous-generation chips at the same speed. Energy costs represent a huge operating expense for server farms, which is why some customers choose to upgrade regardless of whether they need more computing power or not. Production on the N2 assembly lines is not expected to begin until 2025, which likely means Nvidia GPUs based on them won’t reach customers in large quantities until 2026.

Meanwhile, Nvidia is currently rolling out its Blackwell architecture GPUs, which will replace the Hopper architecture that the current top-of-the-line chips are built on, and the improvements are astounding. Blackwell’s architecture is four times faster than Hopper’s, allowing AI companies to build more complex models faster.

The combination of all of these factors suggests that demand will remain strong well beyond 2026. In other words, the market is unlikely to peak anytime soon. This is crucial because Nvidia’s forward price-to-earnings ratio has already reached a level that appears reasonable, at least relative to its pace of growth.

NVDA PE Ratio (Forward) data from YCharts.

Trading at 45 times forward earnings, Nvidia stock is far from cheap, but it is experiencing strong growth, making this valuation acceptable.

Investors’ decisions about whether to buy or sell Nvidia stock today should depend on how they expect the company’s business to perform in 2026 and beyond. There are enough catalysts that Nvidia’s growth should continue well beyond 2026, and with the upgrade cycle, the company should be able to maintain its newfound revenue levels.

So I think Nvidia’s buying arguments are greater than its selling points today.