Nvidia (NASDAQ: NVDA) stock has fallen 17% this year. Now at $111, it’s trading at the same price it was back in May. Yet the company continues to put up huge growth rates and reckons the artificial intelligence (AI) revolution is still in its infancy.
This leaves me wondering if there’s a buying opportunity here for my Stocks and Shares ISA.
Attractive valuation, on paper
On the one hand, it’s a bit strange that Nvidia stock is flat after 10 months. After all, the firm looks nailed on to continue growing strongly due to the colossal AI investments that Big Tech’s already confirmed it will make this year.
Amazon Web Services (AWS) plans to invest over $100bn in AI infrastructure to enhance its cloud services. Microsoft has earmarked more than $80bn, Alphabet (Google) $75bn, and Meta Platforms as much as $65bn. Then there’s Oracle, Tesla, OpenAI, and so on.
Wall Street analysts expect Nvidia’s revenue to grow 57% to $204bn this year, with earnings per share (EPS) increasing by 52%. Based on this, the price-to-earnings (P/E) ratio is 24.5. That’s not a daft multiple for a company still at the centre of the AI boom.
By 2028, the firm’s revenue is tipped to reach almost $300bn, with EPS at about $6.70. That translates into a forward P/E ratio of just 17. Seen from this angle, Nvidia stock looks a bit of a no-brainer buy for my portfolio.
But I do have some concerns.
DeepSeek doubts
One is that Nvidia’s growth is heavily reliant on those Big Tech customers previously mentioned. While sales for its latest Blackwell chips are very strong, demand could always taper off next year as some large customers start deploying their own custom-developed chips.
Meanwhile, Chinese AI firm DeepSeek reportedly developed a cost-efficient large language model (LLM) that doesn’t run on a load of high-end chips. In theory, this might eventually mean reduced demand for Nvidia’s AI hardware.
DeepSeek misunderstanding
But am I thinking about things in the wrong way? Nvidia CEO Jensen Huang thinks so. Speaking to CNBC about DeepSeek’s R1 model, he said: “This reasoning AI consumes 100 times more compute than a non-reasoning AI… the exact opposite conclusion that everybody had.”
At its recent annual technology conference, Nvidia reiterated that we’re moving towards AI agents and long-think reasoning models. Put simply, agentic AI refers to models that can understand, plan, and take action. In other words, AI with agency.
The planning part means back-and-forth reasoning to thoughtfully take the best course of action. To do that quickly will need much more computing power, not less.
This kind of intelligence lays the groundwork for physical AI, such as useful humanoid robots.
I’m expecting volatility
Looking forward, I have less concerns about competition. That’s because Nvidia says its next AI superchip, Rubin Ultra in 2027, will have 400 times the performance of its 2022 Hopper architecture!
I suspect Big Tech firms will continue to rely on the best chips (ie Nvidia’s). The stakes appear too high, creating a sort of ‘Prisoner’s Dilemma’, where holding back risks being left behind in the AI race.
I’m expecting further market volatility due to tariffs and worries about restricted Nvidia chip sales to China. If the stock moves near $100, I’ll buy it.
