Key takeaways

  • First-quarter losses for AI stocks were caused by a combination of macroeconomic uncertainty and DeepSeek’s debut.
  • AI stocks remain volatile, but they have recovered their losses from the first quarter.
  • Hardware companies remain the easier investment.

2025 has been nothing short of dramatic for artificial intelligence stocks. After a blistering two-year run, fueled by surging investor enthusiasm and rapid technological breakthroughs, the momentum came to an abrupt halt in early January, followed by a steep collapse. But then came the massive rebound.

The selloff began when the Chinese AI model DeepSeek launched. It offers performance on par with leading market models at a fraction of the cost, raising serious doubts about the long-term need for expensive infrastructure. Morningstar director of equity research Eric Compton says that more broadly, concerns grew that data center investment could slow, particularly at Microsoft MSFT.

Tech stocks took another blow in early April after President Donald Trump announced aggressive tariffs, with semiconductor stocks and other big tech names (most notably Apple AAPL) potentially in the crosshairs. However, after Trump walked back the worst of the levies, stocks promptly rebounded, with AI names far outpacing the rest of the market. At the same time, new data and corporate commentary confirmed that demand for AI services remained strong and data center buildouts were continuing as planned.

By the time the dust settled, a basket of 38 AI stocks selected by Morningstar analysts had risen 27.3% in the second quarter, or up 15% year-to-date. That compared with an 11% gain in the Morningstar US Market Index and a nearly 22% rise in the Morningstar US Technology Index. Stocks such as Broadcom AVGO, with a 65.0% return, and Palantir PLTR, which rose 61.5%, led the recovery.

AI stocks have been more volatile than the overall technology sector, and “the uncertainty is legitimately higher,” Compton says. “For many companies, the cone of possibilities is something like growing 2%-5% for the next three years. For some younger software companies, it might be ‘Do you grow at a 10% CAGR or 15% for the next decade?’”

But when it comes to AI stocks, “Everyone is fighting over ‘Is this the next big thing that takes over more market value than anything before it, or will it stall out?’ So there is a much larger cone of uncertainty, and every incremental data point feeds into which path it is headed down.”

The AI stock basket

The Morningstar equal-weighted basket of 38 stocks aims to represent a cross section of the AI universe. The largest portion is allocated to hardware firms involved in building or supporting AI server infrastructure, such as Taiwan Semiconductor Manufacturing TSM. Next are the hyperscalers, included for their cloud businesses, product-level AI integration, and internal infrastructure development which includes names like Amazon AMZN. Finally, there are select software firms that could potentially benefit from advancements in AI, such as Salesforce CRM.

The best and worst performers in the Morningstar AI basket

There were clear winners and losers in the year to date.

Among the top five best-performing stocks so far in 2025 are SK Hynix 000660, up 74.54%, and Micron Technology MU, up 44.79%. Compton says these stocks “benefit from good belief in the HBM story related to AI,” referring to growing optimism around high-bandwidth memory and its role in powering advanced AI workloads.

Others had less fundamental support for their rallies, such as Palantir PLTR, which was up 74.69% over the same period. “Palantir is more pure ‘meme/market beta’-related,” says Compton, who attributes its gains more to investor sentiment and speculative momentum than business fundamentals.

Several high-profile names are at the bottom of the list. The five worst-performing stocks so far in 2025 include Marvell Technology MRVL, down 33% after facing headwinds due to investor concerns about potentially losing a major custom chip contract with a hyperscaler.

Meanwhile, Adobe’s ADBE fall was driven by growing fears that generative AI could disrupt demand among its core creative user base. As for Apple, Compton points to slowing sales growth and lingering uncertainty around tariffs as leading to the stock’s underperformance. Both of these household names have shed nearly 15% since the start of 2025.

AI stock valuations

Along with prices, valuations have swung widely. At the start of this year, the AI basket was valued at 1.12 times its fair value estimate. This plummeted to 0.71 times by April, then recovered to 1.01 by June.

Now, with the AI basket trading close to its aggregated fair value estimate, “if you combine solid structural tailwinds with fairly valued stocks, it’s not a bad spot to be,” Compton says. But he also suggests that opportunities still exist.

Compton points to the hardware industry as the “easier and more direct play” for investors, with software still in the early stages of integrating AI capabilities at scale. While long-term potential exists across both ends of the value chain, from chipmakers to enterprise AI users, investors looking for near-term clarity may be better served by focusing on the infrastructure backbone of the AI ecosystem.

“We think Palantir, Arista ANET, Amphenol APH, Vertiv VRT, KLA KLAC, Broadcom, and AMD AMD are looking pricey today,” Compton says. Instead, Morningstar analysts favor Taiwan Semi, Alphabet GOOG, ASML ASML and Marvell.

Get Morningstar insights to your inbox