Semiconductor stocks were among the biggest beneficiaries of the artificial intelligence rally, with Nvidia NVDA going from being known mainly by gamers to a household name. From May 2023 (when Nvidia unveiled a massive jump in earnings thanks to AI interest in its newest, most powerful chips) through Jan. 22 of this year, the Morningstar Global Semiconductors Index rose some 161%, more than three times the overall US stock market.
However, semiconductor stocks have stumbled recently, with the index falling 17% in the span of roughly two months. They’ve been caught up in the broad selloff in technology stocks, as lofty valuations have met an uncertain economic outlook, thanks largely to US President Donald Trump’s trade wars.
Amid the enthusiasm for semiconductor stocks, investors may have lost sight of the highly cyclical nature of other uses for semis, such as in autos, household goods, and manufacturing. To see what may lie ahead, we spoke with Morningstar strategist Brian Colello, who follows many of the key semiconductor stocks. This interview has been edited for length and clarity.
Q: Since AI has been so dominant, lets start with the state of play for chip stocks and how you see the arc of sentiment.
Colello: We’re bullish on the AI opportunity as a trend. We still see plenty of companies investing in AI in all walks of life: startups, software companies, drug discovery, robotics, autonomous driving. That spending is still going on. We were pleased with the hyperscaler capex plans for 2025 from Alphabet GOOG, Meta META, Amazon AMZN, Microsoft MSFT. All gave guidance at or above what I would have expected.
We’re tremendously optimistic about Nvidia’s growth. We have started to see some risk and some caution factored into Nvidia shares, which we didn’t see much of for most of the past two years. While 2025 spending might be great, 2026 is becoming a bit fuzzy. While Nvidia has reported earnings the past two quarters that beat their guidance and consensus, and they guide the next quarter’s revenue above consensus, they’re not doing so at an accelerating pace. The beats are not nearly as large as they were at the dawn of the AI era.
We see a lot of fast money in Nvidia. For a $3 trillion stock, it can move significantly in a day. A lot of investors are likely making big bets on Nvidia beats, and those haven’t come to fruition the past couple quarters. So I think that’s probably the headwind.
Q: How much of that is the law of large numbers, in that as Nvidia gets larger, it takes much more to move the needle, and how much is more of a change in the landscape?
Colello: I think it’s more law of large numbers. Nvidia is consistently growing its data center revenue by $4 billion a quarter. The first time they grew by $4 billion, that was 280% year-over-year revenue growth. And now it’s “only” 78% growth. So those percentages are going to naturally decelerate.
I think investors have caught up to the size of the gold rush and have a better grasp of what this business will look like, at least for the next 12 months. Then it becomes debatable. How long will the Microsofts of the world keep spending? Does that capex get cut? What does the growth trajectory look like?
Q: What about Broadcom, another significant player?
Colello: My colleague Will Kerwin covers Broadcom AVGO. It’s likely to be the other big winner in AI semis. They’re basically building the custom in-house chips for most everyone else. There’s potential for OpenAI, Apple AAPL, and Bytedance to bring custom chips online and gain share. We think Nvidia will almost certainly lose share over the next few years, but in a much bigger market. They’d rather have 75% of a market that’s three times bigger than 95% of a relatively smaller market.
So Broadcom and those customers are going to switch to in-house chips. It makes all the economic and technological sense for them to do so and reduce their reliance on Nvidia. The company we haven’t talked about yet is Advanced Micro Devices AMD, and we still think they will capture a sliver of AI pie. The size of that sliver is up for debate. The market, I think, has sold off a bit too much around pessimism about the size of that pie. Let’s say 2025 growth isn’t as strong as anybody expected a year ago. That doesn’t mean they don’t have great opportunities to be relevant. In 2026/27/28, meanwhile, we think they have plenty of opportunities to gain share from Intel INTC.
Q: What do you think the story is for Intel going forward?
Colello: The story is a lot of uncertainty. I rarely mention Intel when discussing the semiconductor sector because so many of its problems, and perhaps opportunities, are company-specific. At this point, I think the best-case scenario is some sort of breakup where the chip design business either becomes independent or gets sold to a Broadcom or Qualcomm QCOM at a decent multiple, because that business is profitable. And then the foundry business either gets bailed out or gets some sort of support to be propped up, whether that’s a Taiwan Semiconductor TSM semi-joint venture or a consortium of customers coming in to buy it to ensure that there is a second foundry to compete against Taiwan Semiconductor. That said, Taiwan Semiconductor said they’re going to spend $100 billion in the US recently, on top of the $65 billion they’re already going to spend in Arizona for new fabs. This implies they don’t have a lot of room to also buy the Intel fabs, or that would certainly be redundant.
Q: Let’s talk about the macro situation. Investors are increasingly worried about the potential for a US economic slowdown. How much of the outlook for semis is dependent on cyclical demand, or do we think AI demand is going to be less cyclical? For the parts of the semi industry that are clearly affected by the macro economy, where do we stand?
Colello: I’ll start with the last part. Companies with industrial exposure, like Texas Instruments TXN and Analog Devices ADI, faced a severe downturn in 2023 and 2024. It was basically a reckoning from a lot of the chip sales during the global chip shortage. Microchip Technology MCHP might be the worst example of this. Their customers ordered far too many parts than they could possibly use out of fear of not being able to get anything else, and now they’re working through that inventory. For the past nine months, some of these firms were expecting a stronger bounce back in industrial demand (factory automation, medical equipment) and other parts of the economy, such as appliances. However, it wasn’t bouncing back in the second half of 2024, and given tariff concerns and uncertainty, it doesn’t seem like it will bounce back all that great at the start of 2025.
I think inventory is low enough that the chip makers are springloaded to see a recovery in demand whenever production picks back up, but tariffs, macro concerns, and a recession could extend the bottom of this downturn longer than anybody would expect.
Automotive is a similar end market. The downturn there wasn’t as severe because there’s still the favorable trend of rising chip content per car. All these electric vehicles just use more chips for dashboard displays, infotainment, and safety features. Tariffs might hit the auto market especially hard, or at least the US market. So that could be another, more severe headwind that stunts the rebound in the automotive market.
Q: How about the AI side of the equation?
Colello: Our contention has been that companies are going to spend on AI in good times or in bad. In good times, it’s the future and a place for development. In bad times, companies need to become more efficient and look for synergies. So maybe they deploy AI more and reduce their hiring, or use it as a supplement if there’s a layoff going on.
We still haven’t seen signs that the hyperscalers are cutting capex. We haven’t seen too many CEOs say they’re going to scale back AI spending or investment. There’s very much a race, and there are rewards for the company that can get the biggest, best, fastest model integrated most seamlessly into their existing products. So I do think AI is going to be independent of any recession that might come.
Q: To wrap things up, how would you summarize the state of play? Historically, semis are a very volatile part of the market, and that doesn’t seem to have changed. So how should investors approach these stocks.
Colello: We still like buying on downturns, because we think the upturn is inevitable. Semiconductors are not going away as a technology at all. We still like the trend of more chip content per car, more chips in industrial equipment, the smart factory, smart cities, smart gadgets. Even in a recession, we doubt that engineers are going to start building dumber devices to save on chips. There are efficiencies to having sensors and devices to do more with data.
Names that look cheap to us right now are NXP Semiconductors NXP and Infineon Technology IFNNF, although they have automotive exposure and are a bit more cyclical. Infineon actually has a bit of an AI play to it. They might fare better than most for that reason.
In the AI world, Marvell Technology MRVL sold off significantly. They had a tough quarter, but we think their prospects are underestimated. AMD as well. I still think the GPU business has been a bit too beat up. And the potential for share is underestimated in the CPU business against Intel and in PCs and servers. If Nvidia starts trading below $100 per share, I’d start to like the risk/reward. Analog Devices is rarely cheap (it usually trades at a premium to the space), and it’s probably the most defensible out of all the names with cyclical exposure, so that seems like a potential baby in the bathwater from the selloff.
The author or authors do not own shares in any securities mentioned in this article. Find out about Morningstar’s editorial policies.
