Nvidia (NASDAQ: NVDA) shares have taken quite the nosedive. A 17% drop in a single day grabbed most of the headlines, but the stock has been down 22% at its lowest, after kicked off 2025 at an all-time high. The $116 share price was one some talking heads said we’d never see again. Is it time to buy the dip before new all-time highs? Or could this one have further to crash?

AI dominance

To understand the appeal of Nvidia shares, it’s worth taking a moment to assess just how dependent large language models (LLMs) are on its chips. 

When ChatGPT launched, Nvidia made the best GPUs for it. It had a head start. And that resulted in 90% or so of the chips used being from Nvidia. 

But ChatGPT debuted, if you can believe it, nearly two and a half years ago. Tonnes of rival LLMs have hit the market, like Claude, Grok, and Gemini. Surely Nvidia’s competitors have had a chance to catch up? 

Well, not really. The percentage of Nvidia’s chips is thought to still be around 85%. Nvidia is halfway down the track while its competitors haven’t even finished tying their shoelaces.

It’s the kind of seemingly unassailable lead that easily explains why Nvidia shares have multiplied 11 times in value since LLMs burst onto the scene. AMD shares haven’t even doubled. Intel shares are down 38%. Crikey.

Why did the stock drop?

So what’s this drop about, then? A 22% fall is nothing to sniff at. Is it a sign that Nvidia’s dominance has an end in sight? 

Well, the basic story is that a Chinese startup named Deepseek made an LLM for a fraction of the cost of all the others. The relevant point to Nvidia is that it doesn’t need as many chips, which could make a long-term dent in sales.

The counterargument goes that this has kicked the door wide open to widespread adoption of AI. We might see custom models running on the smartphones we all have in our pocket.

If that’s the case, then Nvidia could come out on top again. After all, their chips are still the best in class. The hubbub around Deepseek could end up increasing sales. 

My opinion

Personally, I don’t think the investment case has been harmed much in the last week. An investor lacking exposure to the benefits of AI may want to consider buying in at the discounted price. 

What is informing my decision is the valuations – currently a price-to-earnings ratio of 48 with a forward P/E ratio of 30. Those aren’t cheap, though they aren’t astronomical either for a company with such good growth prospects. 

The issue is that earnings have been elevated thanks to the AI gold rush. It is rumoured that most of the sales come from four or five megacap tech firms. Combined with my exposure in other areas, that seems like too much risk for me.