NVIDIA's (NVDA 3.37%) graphics chips are widely used for artificial intelligence (AI) workloads. While AI will almost certainly prove to be a transformative technology, there's a decent chance that a bubble is brewing. The internet was transformative, after all, and it still produced the mother of all bubbles.

AI hype is one reason why shares of NVIDIA have shot through the roof this year. Despite tumbling revenue and profit in the latest quarter, driven by crashing demand for PCs and the end of the cryptocurrency bubble, NVIDIA stock has more than doubled so far in 2023.

Is AI going to reignite NVIDIA's growth story? It very well could, but you'll be paying a historically high price for the stock to find out.

More expensive than the dot-com bubble

Right now, NVIDIA stock trades for just under 30 times sales. There have only been two other times when the valuation has come close to this level. The first time was the dot-com bubble of the late 1990s. The most recent time was during the pandemic when strong demand from gamers collided with a massive cryptocurrency bubble. In addition to AI, NVIDIA's graphics processing units (GPUs) can be used to mine various cryptocurrencies. 

NVDA PS Ratio Chart

NVDA PS Ratio data by YCharts.

Given that the only times NVIDIA stock has been this expensive on a price-to-sales basis were during historic bubbles, it would be a good idea for investors to take a cautious approach. The potential downside is huge.

NVIDIA is also quite expensive relative to earnings. The stock trades for around 180 times earnings at the moment, a level that has not been seen since the early 2000s. It's true that earnings are depressed and should recover to a degree as gaming demand picks back up, but NVIDIA stock still trades for nearly 70 times the average analyst estimate for fiscal 2024 earnings.

NVDA PE Ratio Chart

NVDA PE Ratio data by YCharts.

Competition in AI

Beyond a sky-high valuation, NVIDIA's products do not stand alone in enabling AI. In the world of semiconductor companies, rival Advanced Micro Devices (AMD) and Intel are both in the mix. AMD now views AI as its "No. 1" strategic priority, and Intel is going after AI on multiple fronts. Intel's latest Sapphire Rapids server CPUs have built-in AI accelerators, and its data-center GPUs provide even more AI firepower.

Outside of CPUs and GPUs, custom AI chips are becoming more common. Social media giant Meta, for example, recently detailed its homegrown AI chips that help power the company's recommendation algorithms. Amazon offers usage of its Inferentia chips through Amazon Web Services (AWS), giving customers an efficient way to deploy complex AI models. And Alphabet's Google has designed multiple generations of its tensor-processing unit.

A custom chip tailor-made for AI has the potential to greatly outperform more general-purpose chips like GPUs. NVIDIA has been investing in AI for a long time, and its products have become the de facto standard. But as AI is deployed more widely, cost efficiency will become increasingly important.

A risky stock

There may be no company better positioned to benefit from a surge in AI use than NVIDIA. But with the stock more than doubling this year, a potential boom in AI-related sales may already be priced in. NVIDIA stock is about as expensive as it's ever been, driven higher by extremely optimistic expectations. If those expectations aren't met, even if AI does drive a surge in sales, the stock could be in for a reckoning.

Long story short, be careful if you're considering investing in NVIDIA stock.