Nothing moves in a straight line, especially in an industry known for dramatic booms and busts. The key question is whether the recent sell-off is a healthy reset in a powerful growth cycle, or an early warning that the good news has already peaked.
Memory stocks have been one of the hottest corners of the artificial intelligence (AI) trade, with soaring demand and tight supply driving prices, profits and investor expectations sharply higher. But after a blistering rally, questions are mounting, and some of the memory names have pulled back from recent highs.
This article isn’t personal advice. If you’re not sure an investment is right for you, seek advice. Investments and any income from them will rise and fall in value, so you could get back less than you invest. Ratios also shouldn’t be looked at on their own.
Investing in an individual company isn’t right for everyone because if that company fails, you could lose your whole investment. If you cannot afford this, investing in a single company might not be right for you. You should make sure you understand the companies you’re investing in and their specific risks. You should also make sure any shares you own are part of a diversified portfolio.
What is memory and who are the key players?
AI does not run on processors alone. It also needs memory – chips that hold data. Without enough memory, even the most powerful AI processor can't work at full capacity.
The two main types are DRAM and NAND. DRAM provides fast working memory, while NAND stores data for longer periods in products like smartphones, laptops and data centres. High-bandwidth memory (HBM) is an advanced form of DRAM that feeds data to AI processors at extremely high speeds.
Complex AI systems need vast amounts of information delivered quickly, making memory a potential bottleneck for the wider AI buildout.
Micron, Samsung and SK Hynix dominate the broader memory market. Their exposure to HBM, conventional DRAM and NAND varies, so while similar, they shouldn't be treated as identical investments.
Why has memory become such a hot topic?
AI demand has collided with an industry that can't quickly create new supply. Semiconductor factories take years to build and bring into production, leaving manufacturers unable to respond quickly when demand accelerates.
HBM adds further pressure because it uses considerably more manufacturing capacity than conventional memory. Shifting production towards HBM can therefore reduce the amount available for traditional servers, smartphones and computers.
Supply has struggled to keep pace, driving sharp price rises, profit margins above previous cycle peaks, and repeated upgrades to earnings forecasts. But memory is traditionally a boom-and-bust industry, so the key question is whether AI demand has changed the cycle or simply extended it.
Why have memory stocks sold off?
After exceptional performance, memory stocks became vulnerable to any news that challenged expectations of continuously rising prices and profits.
Recent price increases can't continue indefinitely without making AI infrastructure increasingly expensive. Slower price growth isn't the same as falling demand, but it can still disappoint investors expecting constant upgrades.
Concerns around higher spending on new factories, export restrictions, Chinese competition and customers potentially using less memory have added to the pressure. None point to a clear collapse in data centre demand, but together they gave investors a reason to take profits from a crowded trade.
Not all memory stocks are the same
Micron offers broad exposure to DRAM and NAND through a major US-listed company and looks to have structurally higher margins in this latest cycle. But significant factory expansion means spending levels and execution will be important.
SK Hynix has a leadership position in HBM, giving it direct exposure to demand for AI processors. Key questions are whether HBM leadership can be sustained and whether a broader investor base, with its new US listing, can help close the valuation gap vs Micron.
Samsung combines memory with smartphones, displays and other semiconductor operations. This provides greater diversification but means its investment case isn't driven by memory alone.
Smaller names like SanDisk offer greater exposure to NAND. They could benefit as advanced AI systems require more storage but face a different supply picture and potentially greater competitive pressure from China.
Why this cycle could have further to run
Memory remains a genuine bottleneck for AI infrastructure, and there is not a quick supply fix. New factories are being built, but construction and production ramps take time, while HBM's greater manufacturing demands limit the benefit of additional capacity.
New and improved long-term supply agreements offer another layer of support and are a key differentiator for this cycle. Customers who provide deposits or financial guarantees face a real cost if they walk away, giving manufacturers greater protection than in previous cycles.
These contracts can't guarantee today's earnings or remove the industry's ups and downs. But they could improve visibility and soften the next downturn. We think future cycles will see higher peaks, lower troughs, and run for longer.
Chinese competition also looks more like a longer-term risk than an immediate answer to current shortages. Near-term supply still depends more heavily on the investment decisions made by Samsung, SK Hynix and Micron.
What do the numbers say?
The following scenario matrices show the potential upside/downside under different earnings and valuation assumptions. The columns reflect a range of next-12-month earnings outcomes, from 20% below to 20% above the current consensus. The rows show different price-to-earnings (P/E) multiples investors might be willing to pay, with the current P/E highlighted and the 90-day average used as an anchor (the middle number).
Each cell displays the potential upside or downside from a given scenario. It is not a price forecast, but a way to illustrate how changes in earnings expectations and multiples could affect potential returns, and nothing is guaranteed.
Micron:
Earnings Scenario
Price to Earnings Multiple | Bear -20% | Downside -10% | Base 0% | Upside +10% | Bull +20% |
|---|---|---|---|---|---|
4.2 | -33% | -25% | -17% | -8% | 0% |
4.6 | -26% | -17% | -8% | 2% | 11% |
5.1 | -18% | -8% | 3% | 13% | 23% |
5.7 (current) | 9% | 3% | 14% | 26% | 37% |
6.3 (90-day av.) | 1% | 14% | 27% | 39% | 52% |
7.0 | 12% | 26% | 39% | 53% | 67% |
7.7 | 23% | 38% | 53% | 69% | 84% |
8.4 | 35% | 52% | 69% | 86% | 103% |
9.3 | 49% | 67% | 86% | 104% | 123% |
SK Hynix:
Earnings Scenario
Price to Earnings Multiple | Bear -20% | Downside -10% | Base 0% | Upside +10% | Bull +20% |
|---|---|---|---|---|---|
3.4 | -25% | -16% | -6% | 3% | 13% |
3.8 | -17% | -6% | 4% | 15% | 25% |
4.2 (current) | -7% | 4% | 16% | 27% | 39% |
4.6 | 3% | 16% | 29% | 42% | 54% |
5.2 (90-day av.) | 14% | 29% | 43% | 57% | 72% |
5.7 | 26% | 42% | 57% | 73% | 89% |
6.2 | 38% | 56% | 73% | 90% | 108% |
6.9 | 52% | 71% | 90% | 109% | 128% |
7.6 | 67% | 88% | 109% | 130% | 151% |
Samsung:
Earnings Scenario
Price to Earnings Multiple | Bear -20% | Downside -10% | Base 0% | Upside +10% | Bull +20% |
|---|---|---|---|---|---|
3.3 | -30% | -21% | -12% | -4% | 5% |
3.7 | -22% | -12% | -3% | 7% | 17% |
4.1 (current) | -14% | -3% | 8% | 19% | 30% |
4.6 | -4% | 8% | 20% | 32% | 44% |
5.1 (90-day av.) | 7% | 20% | 33% | 47% | 60% |
5.6 | 17% | 32% | 47% | 61% | 76% |
6.2 | 29% | 45% | 61% | 78% | 94% |
6.8 | 42% | 60% | 78% | 95% | 113% |
7.4 | 56% | 76% | 95% | 115% | 134% |
This scenario analysis can help investors frame the valuation debate. We believe that structural shifts in the memory market could allow these companies to sustain higher multiples than we see today. But each investor will need to make their own judgment.
Cash flow as a weapon
We think the market is overlooking the power of the expected incoming cash flow from the memory names, and this is a major source of strength, even if our view that this cycle is different does not pan out.
Current estimates suggest these three companies could generate more than $1.3tn of free cash flow over the next three financial years. That would give them substantial financial firepower, potentially supporting a new wave of shareholder returns that could materially impact the investment case. Although nothing is guaranteed.
What could go wrong?
High profits encourage investment, and memory manufacturers are already spending heavily to expand. If Samsung, SK Hynix and Micron add capacity too aggressively, today's shortage could eventually become an oversupply, putting pressure on prices and margins.
Long-term contracts have limits. Customers might still walk away if market prices fall far enough for the savings to outweigh any deposit or penalty, with protection potentially weaker in the earlier years.
Demand could also moderate. AI spending can't rise indefinitely unless it generates attractive returns, while expensive memory could push customers to redesign equipment, use less memory per server or delay deployments.
How to position from here?
We think AI hardware could capture a larger share of the economic value created by AI than in past technology cycles. That leaves memory companies well placed structurally, although expectations are now high.
Pullbacks are likely along the way and can help remove some of the excess from a crowded trade. But investors should focus on the strength of the underlying cycle rather than trying to time short-term moves.
Rapid upgrades to earnings forecasts have been a key catalyst for the sector, and the easiest progress may now be behind it. From here, the case rests more on whether greater confidence in the cycle's durability can support structurally higher valuation multiples – we believe it can, though there are no guarantees.
The author holds shares in Micron.
This article is original Hargreaves Lansdown content, published by Hargreaves Lansdown. It was correct as at the date of publication, and our views may have changed since then. Unless otherwise stated estimates, including prospective yields, are a consensus of analyst forecasts provided by LSEG. These estimates are not a reliable indicator of future performance. Past performance is not a guide to the future. Investments rise and fall in value so investors could make a loss. Yields are variable and not guaranteed.
This article is not advice or a recommendation to buy, sell or hold any investment. No view is given on the present or future value or price of any investment, and investors should form their own view on any proposed investment.