An analyst says Micron investors have gotten used to rapid growth but now have to reorient themselves around the idea that AI is more like a marathon

Micron's stock has lost 26% since peaking in June

Micron Technology's stock has seen a tough stretch recently as investors question the sustainability of artificial-intelligence spending and financing.

The fears aren't exclusive to Micron (MU), Stifel analyst Brian Chin told MarketWatch. But they nonetheless have investors wondering what could finally get one of the most notable AI stocks going again.

If demand slows down, memory is affected more than other components, because "it's not just a volume or unit problem; it ends up being a pricing problem," Wedbush analyst Matt Bryson told MarketWatch. In other words, Micron and its fellow memory providers have been able to charge high prices because demand is so strong, but customers won't be willing to pay so much in a slower environment.

In reality, any AI bubble is unlikely to burst immediately. So we asked analysts what they're saying to advise clients during this period of heightened anxiety but limited fundamental changes.

First, it's worth thinking about how much longer the memory market can keep growing rapidly. While some investors are concerned about a negative inflection point in demand two quarters out, Chin said the reality could be more like two to three years. If strong conditions can sustain that long, Micron's earnings per share would be much higher than it is today, he added.

That brings up the valuation point. There's a camp that believes Micron's stock can fetch a much higher multiple than its current mid-single-digit level, but Chin isn't one of those. Micron shares trade at about five times his estimates for forward earnings per share, but he doubts investors will value Micron more favorably for each dollar of profit it brings in.

On the plus side, Micron's notably low multiple doesn't leave much space to compress, making the stock "somewhat defensive if earnings at least grind higher," Chin said.

Ryuta Makino, an analyst at Gabelli Funds, told MarketWatch that Micron's stock stacks up well on a relative basis. Shares are trading at a cheaper multiple than storage stocks, making Micron attractive, he said.

Then there's the question of market supply. Micron investors don't only have to worry about a demand slowdown, but also about what happens if too much supply floods the market. Additional supply from Micron or a rival could impact the industry's ability to charge high prices for memory, and all of the players have some plans to add capacity.

However, Chin said semiconductor equipment makers are having difficulty ramping up the tools that go into chip-fabrication plants. As such, "friction in supply chains will meter some supply growth."

His view is that momentum in the memory market is shifting from a sprint race to a long-distance marathon. "That's the mentality needed in terms of being an investor in Micron at this stage," Chin said.

Now what about the funding side of the equation? Micron's cash position is swelling, but some companies that buy AI chips are increasingly reliant on borrowed money, and that's helping the AI ecosystem stay afloat. So broader concerns over mounting debt in the AI trade are weighing on Micron's stock as well, Makino told MarketWatch.

Frontier AI labs and new-age cloud providers are seeing negative free cash flow, and it doesn't help that data-center server prices are increasing due to rising memory prices, he said. That "further exacerbates financing risks."

Still, for Micron, the supply-demand dynamic that has allowed it to raise prices is still tight, Makino said, and he expects that it will only get tighter next year and into 2028.

Overall, Wedbush's Bryson doesn't see signs that demand is slowing, though he does expect the memory trade to remain volatile.

Case in point: Micron's stock was choppy in early Wednesday trade, rising as much as 3.6% before briefly turning negative and then inching back into the black. The stock has fallen 26% from its June peak, but it's more than tripled on the year.

"The history of compute tells us you need more memory to support applications, and if you believe there will be more AI applications, you need more memory to support those," he said.

-Britney Nguyen