Mega-cap tech firms such as Nvidia Corp. (ticker: NVDA) and Broadcom Inc. (AVGO) became Wall Street darlings during the recent AI buildout. Now, memory stocks are in the spotlight.
That’s because artificial intelligence systems don’t run on processors alone. They also need massive amounts of memory and storage to hold, move and feed data into graphics processing units (GPUs).
Intense demand has made memory stocks some of 2026’s biggest AI winners, with Micron Technology Inc. (MU) up 154% year to date as of May 15 and Sandisk Corp. (SNDK) soaring by 493%.
[Sign up for stock news with our Invested newsletter.]
“The entire industry is benefiting from a really steep supply shortage that’s driving up prices,” says William Kerwin, CFA, senior technology equity analyst at Morningstar.
Still, memory stocks aren’t suddenly safe, wide-moat companies. Instead, investors are buying into one of the hottest — and most cyclical — corners of the AI trade.
What Is AI Memory, and Why Invest?
AI memory refers to the chips and storage technologies used by artificial intelligence systems to process, move and retain data.
The basic idea is simple: AI chips do the computing, while memory helps data flow quickly enough for the system to run at full speed.
AI has pushed memory demand so high, supply can’t keep up.
“Memory is getting more expensive and harder to source, but demand is still strong enough that companies are spending through it,” says Jake Behan, head of capital markets at Direxion, an ETF issuer specializing in leveraged and inverse funds.
That shortage is the core investment thesis driving names like Micron and Sandisk higher.
Memory’s Role in the AI Revolution
Memory isn’t one single product. In AI systems, it generally falls into a few buckets:
— DRAM, or dynamic random-access memory, acts like short-term working memory.
— HBM, or high-bandwidth memory, is a high-performance DRAM technology designed to move large amounts of data quickly, making it especially important for AI workloads.
— NAND flash stores information after a device is powered off.
While all forms of memory are in demand right now, HBM is garnering investor attention because it’s critical to advanced AI accelerators from companies like Nvidia and AMD. “I think the shortage started as HBM-specific, but it’s now clearly spreading into conventional DRAM and storage,” says Shay Boloor, chief market strategist at Futurum, a technology research and advisory firm.
SemiAnalysis, an independent research and analysis firm specializing in the semiconductor and AI industries, has described HBM as central to AI workloads.
“All leading AI accelerators deployed for GenAI training and inference use HBM,” SemiAnalysis authors wrote in an August 2025 research note, adding that relying on other forms of memory results in suboptimal performance.
Why the Memory Boom Might Not Last Forever
Memory has historically been a brutal boom-and-bust industry. As recently as March 2023, Micron — a recent AI rock star — experienced a revenue drop of about 53% in its fiscal second quarter and the company lost $2.3 billion.
Kerwin says he doesn’t think this time is different.
“When we look at the current cycle, we still think it’s a cycle,” he says.
New fabs, or chip manufacturing facilities, can take billions of dollars and years to build out. That lag is helping memory suppliers’ valuations right now because demand is outpacing supply by a country mile.
But eventually, that long-awaited supply will come online, resulting in surpluses, says Kerwin.
“It’s going to likely dampen the growth and profit expansion we’re seeing today,” says Kerwin, who adds that the bulk of new production capacity will likely come online in late 2027 and 2028.
That doesn’t mean the boom is coming to an end anytime soon, though. In fact, Kerwin says Morningstar still sees more upside in the current cycle.
“We think this up-cycle is going to be the strongest we’ve ever seen,” he says. “We think it’s going to be particularly strong in terms of how high the peak goes and how long it lasts.”
The challenge for investors, then, is anticipating when supply will finally catch up.
“You’re effectively needing to predict where that demand is going to land in two to three years,” he says. “Especially in this age of AI, where it’s just skyrocketing, that’s very hard to do.”
AI memory stocks will likely peak well before new production fully ramps up because Wall Street is forward-looking and typically bakes in expectations for a year or two down the road, says Kerwin.
“It can be a minefield for investors because you don’t want to be caught holding the bag,” he says.
Here are five stocks to keep an eye on while the AI memory trend remains on the upswing:
| Stock | Market value | Year-to-date return as of May 15 |
| Micron Technology Inc. (MU) | $843 billion | 154% |
| Sandisk Corp. (SNDK) | $196 billion | 493% |
| Western Digital Corp. (WDC) | $156 billion | 180% |
| Rambus Inc. (RMBS) | $13 billion | 38% |
| Seagate Technology Holdings PLC (STX) | $165 billion | 189% |
Micron Technology Inc. (MU)
Micron Technology is the most direct bet on the AI memory boom among U.S.-listed stocks. The company makes DRAM, NAND and HBM, giving it exposure to both traditional memory demand and the higher-value chips used in AI servers.
Micron’s fiscal second-quarter results show cloud memory revenue climbed to $7.7 billion from $2.9 billion a year earlier, while core data center revenue rose to $5.7 billion. Both segments posted gross margins of 74%, a clear signal that AI-related demand isn’t simply lifting sales, but improving profitability, too.
Investors remain bullish on Micron, with shares of the memory maker up 641% over the past 12 months as of May 15.
Analysts are also forecasting a rosy outlook: TipRanks indicates a “strong buy” consensus from 30 analysts, including 27 “buys” and no “sell” ratings.
Still, Morningstar doesn’t view Micron or its peers as having a clear competitive advantage because customers can generally swap between memory chips from different suppliers.
That makes Micron compelling, but not low-risk. If HBM supply ramps up faster than expected or Big Tech taps the brakes on AI spending, the stock could stumble.
Sandisk Corp. (SNDK)
Sandisk, once known for digital camera memory cards and other consumer storage products, now finds itself at the center of the AI revolution.
Sandisk’s stock has rocketed by 3,380% over the last 12 months. That historic run has made the company one of the top-performing stocks in the S&P 500 during the first quarter. A single share of SNDK was trading at just over $1,400 on May 15.
Its most recent earnings report was jaw-dropping. Fiscal third-quarter revenue at Sandisk hit $6 billion, up 97% from the prior quarter. Data center revenue rose 233%, and gross margin hit 78.4%.
“It’s not just that price increases are remaining high — they’re accelerating,” says Kerwin.
Sandisk also highlighted $42 billion in long-term revenue agreements with business partners in its latest earnings call on April 30, though only $11 billion was fully guaranteed. That gives the company more visibility than in past cycles, but risk remains.
“It creates a higher lock-in, but I don’t think it changes the overarching dynamic that this demand could still turn down,” says Kerwin.
Western Digital Corp. (WDC)
After spinning off Sandisk in early 2025, Western Digital became a more focused bet on hard disk drives and data-center storage infrastructure.
“I think the most undervalued layer is storage,” says Boloor. “The market still tends to treat storage like an old commodity cycle, but AI inference creates durable storage demand in a way I don’t think is fully priced yet.”
Western Digital’s latest results show why investors are paying attention. The company reported fiscal third-quarter 2026 revenue of $3.3 billion, up 45% year over year, with free cash flow of $978 million. It also guided for fiscal fourth-quarter revenue to rise about 36% to 44% year over year.
The company is a clear beneficiary of the memory trade’s blistering rally, with shares up 180% since the start of the year.
Still, Western Digital is a more concentrated hard-drive company after the Sandisk spinoff, which leaves less room for error. If hyperscalers shift more resources toward flash storage, Western Digital could lose leverage that’s helping the stock today.
Rambus Inc. (RMBS)
Rambus is an under-the-radar pick for investors looking for a company with some potential room to run, despite the flood of momentum into the space.
Rambus doesn’t make memory chips the way Micron does. Instead, it makes memory interface chips and IP that help processors communicate with memory faster and more efficiently.
The company recently expanded its lineup of chips and technology used in next-generation AI servers, including products tied to advanced memory systems. Rambus also reported first-quarter 2026 product revenue of $88 million, up 15% from a year earlier.
However, Rambus isn’t a pure AI memory stock, and its story is less obvious than Micron or Sandisk. Investors need to understand they’re buying a picks-and-shovels supplier tied to memory performance, not a company selling massive volumes of HBM.
Its average 12-month price target of $140.40 implies roughly 11% upside to its May 15 close of $127.05, according to TipRanks, suggesting analysts are positive but see less room to run than in some larger memory-linked names.
Seagate Technology Holdings PLC (STX)
Seagate Technology Holdings is an old-school storage name caught up in a modern AI story. The company is best known for hard disk drives, which may sound sleepy next to HBM and GPUs. But AI systems need more than fast memory. They also need massive amounts of storage for training data, model outputs and enterprise workloads.
Seagate is already benefiting from that momentum: In fiscal third-quarter 2026, the company’s revenue rose 44% year over year to $3.1 billion, while free cash flow reached $953 million.
Wall Street is still leaning in. TipRanks shows 14 “buy” ratings and no “sells” on the stock as of May 18.
The risk? Seagate is riding the AI storage wave, but it’s still a mature hardware business. When supply finally catches up, the stock could lose altitude fast.
Can You Invest In an AI Memory ETF?
Investors who don’t want to hand-pick AI memory stocks can now buy a more targeted basket of companies through the Roundhill Memory ETF (DRAM), which launched April 2 and focuses on companies tied to HBM, DRAM and NAND.
Investor demand was immediate and fierce, with the ETF gathering $1 billion in its first 10 trading days. Roundhill CEO Dave Mazza noted in a press release that the memory sector sits at the “critical intersection of AI demand and constrained supply.” Investors seem to agree.
DRAM had about $9.7 billion in assets as of May 18, an eye-catching accumulation in just over six weeks.
Its top holdings include Micron, SK hynix Inc. (000660.KS), Samsung Electronics Co. Ltd. (005930.KS), Kioxia Holdings Corp. (285A.T) and Sandisk. The fund maintains an expense ratio of 0.65%, a relatively affordable price point for a niche actively managed ETF.
DRAM also gives U.S. investors an easier way to access SK hynix, Samsung Electronics and Kioxia — three of the biggest memory companies that are largely inaccessible for American investors because they aren’t listed on U.S. exchanges.
DRAM can help reduce the single-stock risk that comes with a name like Sandisk, though it won’t eliminate volatility entirely.
“You end up running some concentration risk when you carve out such a narrow piece of the AI stack that’s dominated primarily by just a few key players,” says Behan of Direxion.
Broader semiconductor ETFs can also work. The VanEck Semiconductor ETF (SMH), for example, held Micron at 6.3% of assets as of May 15, but its largest positions were Nvidia and Taiwan Semiconductor Manufacturing Co. Ltd. (TSM). That gives investors AI chip exposure, but not a concentrated memory bet.
Behan says that might be a good thing.
“A broader semiconductor ETF gives you more diversified exposure across the AI buildout — not just memory, but compute, networking and the full stack,” says Behan.
More from U.S. News
8 Best Quantum Computing Stocks to Buy in 2026
How to Invest in Digital Twin Technology and Industrial Automation
5 Top Laser Photonics Stocks for 2026
5 Best AI Memory Stocks to Buy for 2026 originally appeared on usnews.com