With bellwethers of the artificial intelligence trade down big so far in 2025, Deepwater co-founder and managing partner Gene Munster says he is at a key inflection point after years of high conviction in the AI trade. “This has made me a little nervous that I’m too bought into the power of AI,” Munster told CNBC. “It’s a test of our conviction [and] it’s something that if you get it wrong, people will ask ‘how on earth did you get that one wrong?’ Shares of chipmaker Nvidia have plummeted more than 20% in 2025. Microsoft, which is a key investor in ChatGPT creator OpenAI, has seen its stock pull back 7%, while Google-parent Alphabet stock has lost about 17%. All three have underperformed the S & P 500 , which is down 6.3% in 2025. Valuations have sharply come down as well. The forward price-to-earnings ratio of the Global X Artificial Intelligence & Technology ETF (AIQ) was sitting at roughly 17 times on Tuesday, compared with about 27 times in 2023 amid the height of the AI boom. The ETF’s biggest holdings include Nvidia, Apple, Microsoft and Alphabet. NVDA .SPX YTD mountain Nvidia stock is still sharply underperforming the S & P 500 in 2025. There’s no shortage of reasons as to why stocks so closely tied to the AI boom have been so hard hit. The broader market is contending with recession worry spurred by President Donald Trump’s tariffs, which has stoked volatility and dampened the appetite for risk. It came as the AI trade was already under scrutiny due to questionably high spending levels and efficiency concerns. And if an economic slowdown does occur, investors realize the tech businesses that generate the funds that support this innovation could become constrained. “Historically, a recession has snuffed out almost every bull market, and there are very few segments of the market that can power through a recession,” Munster said. Most investors that have been riding the AI wave are now playing it safe to see how the AI picture develops over the next three to 12 months before jumping back in, he explained. The DeepSeek moment Even before Trump unveiled his trade policy, there were cracks in the AI trade. The big shot across the bow was the ability of DeepSeek’s R1 model to outperform Open AI’s ChatGPT in several tests. It was allegedly trained using older Nvidia AI chips for $6 million . Although that figure has been disputed, it fanned fears that U.S. tech giants weren’t getting enough return on their investments. For comparison, U.S. megacap tech companies plan to spend more than $300 billion on AI this year alone. While companies like Meta Platforms have been able to show some of the advantages it’s gained from the AI it’s deployed across its businesses, “it’s been hard to quantify those benefits,” according to Truist Securities . The risk-to-reward equation has been on Wall Street’s mind for a while, and reports that both Microsoft and Amazon have quietly paused data center expansion plans, according to channel checks by TD Cowen and Wells Fargo, haven’t helped. Data centers are an essential component of power-hungry AI chips that need large amounts of energy to train large language models. “One thing that is challenging for investors is when there’s a new market that’s created like this, you just don’t know how big this market is going to be long term,” said John Belton, portfolio manager at GAMCO Investors. “As a result, every near-term data point can just get overly extrapolated … so I think it’s a risk off market where it’s easier to take the bearish side of the argument.” And events like Apple’s delayed rollout of its own AI play, Apple Intelligence, don’t help the case. First announced last summer, Apple Intelligence was expected to be incorporated in the iPhone 16 launch in the fall of 2024. Since then, it’s been delayed several times, with the company pushing back AI integration with Siri to 2026. Belton also believes that concern over cloud growth — the segment that drives much of the revenue for megacaps like Amazon and Microsoft — is also weighing on sentiment. “There is a different dynamic happening where some of these macroeconomic indicators have slowed, I think there’s a fear the some of these core public cloud businesses could slow, which I think is a well founded fear,” Belton said. He pointed to the pullback in cloud stocks in 2022 that followed growth concerns caused by higher interest rates. Amazon, for example, lost half its value in 2022 . That’s because cloud growth is tied closely to economic cycles as it’s fueled by business spending. A ‘massive multiyear driver’ To be sure, Munster remains confident that his view on the transformative potential of AI will come to fruition, despite the near-term choppiness. He said that he expects the broader macroeconomic picture to improve over the next 12 months, while global governments also spend on AI projects in concert with hyperscalers. “At the end of the day, even if AI is a small percentage of company revenue for [hyperscalers], it’s still a potentially massive multiyear driver,” he said. “If they’re wrong and it costs them $60 billion, which is chump change, it doesn’t really matter [when] the reward could be a trillion dollar business.” Belton cautioned that recent headlines tied to both Amazon and Microsoft pausing future data center leases perhaps isn’t indicative of weakness of the future of AI or demand in general. Instead, he said he’s still optimistic that demand will remain robust. “These are generally plans that are made years ahead of demand, so I don’t even think activity in these pipelines would really be that reflective of near-term demand in the first place, I just think these checks are coming against an already nervous market,” Belton said. UBS expects the next iteration of the AI trade will broaden out beyond the industry’s biggest players, fueling a new source of growth. “Notably, we expect the share of AI spend beyond the Big 4 to rise to over 40% in 2025 from below 20% in 2023, thanks to the emergence of new AI players in China, the rise of neoclouds, and other developments,” UBS strategist Sundeep Gantori wrote in a note last week. “Hence, we believe AI industry fundamentals are more resilient than what the recent market volatility implies.” — CNBC’s Christopher Hayes contributed to this report.
