Employees of memory chip company SK Hynix attend the company's opening bell ceremony at the Nasdaq market on the day of their IPO in New York on July 10. REUTERS/Angelina Katsanis/File PhotoAngelina Katsanis/Reuters
Memory and storage stocks have been on a blistering rally over the past year, fuelled by booming chip demand from artificial intelligence (AI) data centres and supply shortages.
U.S.-listed Micron Technology Inc. MU-Q has surged by 653 per cent; SanDisk Corp. SNDK-Q, 3,522 per cent; Western Digital Corp. WDC-Q, 742 per cent and Seagate Technology Holdings PLC STX-Q, 490 per cent.
South Korea-based chipmaker SK Hynix Inc. SKHY-Q, whose American depositary receipt began trading last Friday, and Samsung Electronics Co. Ltd. have racked up massive returns as well.
Although these gains also reflect recent sharp pullbacks, some tech-focused fund managers remain bullish. However, they disagree on whether memory and storage stocks are still cyclical with boom-and-bust cycles or have become secular growth plays.
Mickey Ganguly, associate portfolio manager at CIBC Global Asset Management in Toronto, is in the secular-growth camp because of insatiable chip demand from hyperscaler tech giants, such as Amazon.com Inc. AMZN-Q, which operate data centres.
“They cannot buy enough of this product,” says Mr. Ganguly, lead contributor on CIBC Technology Innovation Fund ATL158.CF, which holds the four U.S.-listed memory and storage stocks.
SK Hynix, Samsung Electronics and Micron Technology are the three largest global players making high-bandwidth memory (HBM) chips for AI data centres. They also produce consumer-focused, dynamic random-access memory DRAM chips, whose prices are rising. SanDisk is a flash memory and storage company, while Seagate Technology and Western Digital are data-storage firms.
A shortage of the advanced, power-efficient HBM chips stems from the industry shift from generative AI to agentic AI, which relies on intensive inference to plan, reason and generate answers.
Industrial and automotive companies once drove memory chip demand, but tech giants now sign three-year to-five-year contracts to guarantee supply, he says. “It’s very different than in the past.”
Recently, Apple Inc. AAPL-Q also raised prices on iPads, MacBooks and other devices because of surging memory and storage costs.
“We have gone through many boom cycles for semiconductors but have never seen this sort of action from industry players,” Mr. Ganguly says. “It shows how structurally different this cycle is.”
He was not surprised by the recent drawdown in memory and storage stocks because they’re historically volatile, while some money managers sold to rebalance portfolios at the end of June.
The CIBC fund also holds the so-called Magnificent Seven stocks – Amazon, Alphabet Inc. GOOGL-Q, Apple Inc., Meta Platforms Inc. META-Q, Microsoft Corp. MSFT-Q, Nvidia Corp. NVDA-Q and Tesla Inc. TSLA-Q - which, as a group, have lagged the S&P 500 this year.
Some money managers were likely selling to move into “high torque, memory names,” he says. “But I remain bullish on the hyperscalers. They’re in the best position to monetize AI.”
Mr. Ganguly is also very bullish on Nvidia because its graphics processing units are a key component of AI and its stock is “relatively inexpensive” based on valuation.
Shane Obata, portfolio manager at Toronto-based Middlefield Ltd., is also very positive on storage and memory stocks, particularly the latter, which trade at a cheaper valuation.
“We think the memory cycle is going to last longer and the companies will make more profits than people expect,” says Mr. Obata, who runs Middlefield Innovation Dividend ETF MINN-NE.
The recent pullback in these stocks was partly due to rotation into some beaten-up Magnificent Seven names, he says. “It’s a temporary reversal, but I don’t think it is durable.”
The Middlefield ETF is underweight the Magnificent Seven, but Mr. Obata says he expects they will outperform again when they reduce AI capital spending, which has hurt free cash flow margins.
The ETF holds SK Hynix (purchased before its U.S. listing), Samsung Electronics, Seagate Technology and Western Digital. Both SK Hynix, which has 58 per cent of the HBM market, and Samsung have traded at lower valuations than Micron because of concerns about South Korea’s corporate governance and opaque conglomerates, he says.
Given that SK Hynix’s U.S. listing will attract more investors, he is “optimistic that the valuation discount will close and that could help valuation for the whole memory group.”
However, he sees memory and storage as cyclical businesses, but expects the current cycle to last longer before the shortage leads to capacity overbuilding and falling prices.
Although Samsung and SK Hynix recently announced major plans to expand chip production over the next five years, he doesn’t expect a big impact because of extremely strong demand. However, competition from China’s memory-chip players is a medium-term risk, Mr. Obata adds.
China-based ChangXin Memory Technologies Inc., which specializes in DRAM memory, will soon go public in Shanghai. Apple has been lobbying the U.S. government for approval to buy its chips for devices aimed at the Chinese market.
Nick Mersch, portfolio manager at Toronto-based Purpose Investments Inc., says he expects the bull run for memory and storage stocks to continue for another couple of years.
“We’re going to be extremely tight in terms of supply into 2028,” says Mr. Mersch, who holds the four U.S.-listed memory and storage stocks in his Purpose Global Innovators Fund ETF PINV-T.
The memory producers have become more disciplined after the sector’s severe downturn in 2022 to 2023, “so they’re adding capacity deliberately rather than flooding the market,” he says.
“I don’t think we’re going to see the period in which the fundamentals reverse quickly,” so any stock drawdowns could be a buying opportunity, Mr. Mersch says.
Still, he too sees memory and storage stocks as cyclical plays, but with higher peaks and higher troughs than in previous cycles.
“Every time you have one of these massive supply constraints, it typically leads to a glut as capacity gets overbuilt,” he says. “The main question is when we will see peak earnings.”
One risk is the possibility of an algorithmic breakthrough that reduces the amount of memory needed for AI performance, Mr. Mersch says.
Still, demand should remain high for now and that should provide “pretty good comfort during these periods of volatility,” he adds. “They still have very, very strong earnings growth.”