By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panelists generally agreed that Sandisk's current valuation is precarious, resting on the assumption of 'this time is different' in a cyclical industry. They expressed skepticism about the durability of high margins and AI-driven demand, with concerns about potential supply glut, margin compression, and capex intensity.
Risk: Margin compression due to supply normalization and potential renegotiation of hyperscaler contracts.
Opportunity: Sustained high demand for HBM driven by AI, potentially leading to elevated margins.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Sandisk (NASDAQ: SNDK) stock closed last week at an all-time high near $990 a share -- and it's starting this week right, too.
Shares of the computer memory-maker jumped 4.2% through 10 a.m. this morning after Melius Research initiated coverage with a buy rating and a $1,350 price target. The analyst believes Sandisk shares will rise another 36% over the next 12 months.
Will AI create the world's first trillionaire? Our team just released a report on the one little-known company, called an "Indispensable Monopoly" providing the critical technology Nvidia and Intel both need. Continue »
Sandisk makes computer memory, which works in tandem with fast artificial intelligence chips to power the AI revolution. Melius argues in a note covered on TheFly.com this morning that this revolution is still in early innings -- and may last "through the end of the decade."
As Melius explains, demand for high-bandwidth memory and the high profit margins Sandisk is earning from it are both "unusual." Stock market investors are having a hard time believing demand will last long, or profit margins remain high long -- anticipating instead that the semiconductor market will do what it always does, and turn from an up-cycle into a down-cycle relatively quickly.
Once investors realize this time is different, however, they'll pay higher multiples to earnings for Sandisk -- and the stock will go higher.
But what if Melius is wrong? What if the semiconductor industry remains as cyclical as it always has been? What if this time is not different?
Sandisk spent three years losing money before finally turning profitable this year. Most analysts see earnings hitting $41.75 in 2026, growing past $107 in 2027 -- then starting to fall in the traditional cyclical pattern. And yet, even in 2030, forecasts have Sandisk earning $43 a share, which is more than it might make this year.
With the stock trading under 25 times earnings both now and then, Sandisk looks cheap.
Before you buy stock in Sandisk, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Sandisk wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $498,522! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,276,807!
Now, it’s worth noting Stock Advisor’s total average return is 983% — a market-crushing outperformance compared to 200% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of April 27, 2026. *
Rich Smith has no position in any of the stocks mentioned. The Motley Fool has no position in any of the stocks mentioned. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"Sandisk is being priced at peak-cycle earnings under the unproven 'this time is different' narrative, leaving significant downside when the inevitable memory downturn arrives."
Melius' $1,350 PT on SNDK (36% upside) rests on the thesis that AI-driven HBM demand and 60%+ margins will persist through decade-end, defying semiconductor cyclicality. The article correctly notes current ~24x 2026E P/E looks reasonable against $41.75 EPS, yet forward multiples already embed much of the optimism. Missing context: Sandisk's three straight loss-making years pre-2025, extreme cyclical history, and consensus forecasts showing EPS peaking in 2027 then collapsing to $43 by 2030. Valuation at peak earnings is not cheap; it's priced for permanent structural change that has rarely materialized in memory.
If AI capex remains robust and HBM supply stays tight longer than historical cycles, Sandisk could re-rate to 30-35x trough earnings, sending shares well above $1,350; the article's bear case may be the higher-probability outcome only if past cycles repeat exactly.
"The market is erroneously pricing a cyclical commodity producer as a secular growth monopoly, ignoring the inevitable supply-side response that historically crushes semiconductor margins."
The premise that 'this time is different' for SanDisk is a dangerous narrative in a sector defined by extreme capital intensity and commodity-like supply gluts. While AI demand for high-bandwidth memory (HBM) is real, the article ignores the massive capacity expansion by incumbents like SK Hynix and Samsung, which historically leads to margin compression. Trading at ~24x forward earnings for a cyclical player is a premium that assumes perfect execution and no supply-side overreaction. If SanDisk's HBM yields falter or if hyperscalers diversify their supply chains, that $1,350 target will evaporate. Investors are currently pricing in a structural shift that the underlying manufacturing economics have yet to prove sustainable.
If AI infrastructure spending is truly a secular shift rather than a standard capex cycle, the current supply constraints could create a multi-year pricing power moat that justifies a permanent valuation re-rating.
"SNDK's valuation appears cheap only if you believe memory demand remains elevated through 2030, but the article provides zero evidence that margin expansion is structural rather than cyclical supply tightness."
Melius's $1,350 target implies 36% upside from ~$990, but the thesis rests entirely on 'this time is different'—a phrase that should trigger skepticism in cyclical industries. The article itself hedges by noting 2027 earnings forecasts then decline, yet SNDK trades at <25x 2026E ($41.75), which is genuinely cheap IF those numbers hold. But here's the trap: memory chip margins are already elevated due to supply constraints, not structural AI demand. When supply normalizes—and it will—margins compress hard. The article also omits SNDK's balance sheet, competitive position vs. SK Hynix/Micron, and capex requirements. A single analyst initiation, however bullish, doesn't move conviction without independent verification of the 'unusual' demand thesis.
If AI truly does drive sustained high-bandwidth memory demand through 2030 and SNDK maintains 25x+ multiples while growing EPS 2.5x, the stock could easily exceed $1,350. The article's own numbers—$43 EPS in 2030 vs. ~$20-25 today—suggest the market is pricing in cyclical pessimism, not AI upside.
"The stock's upside hinges on a durable AI memory cycle and aggressive multiple expansion, both of which are uncertain."
Sandisk (SNDK) is pitched as a secular AI memory winner, with a 12-month target of $1,350 implying ~36% upside from ~ $990. The piece relies on Melius's view that demand for high-bandwidth memory will stay strong through the decade and that margins stay elevated, justifying a higher multiple. What it misses: whether AI demand proves durable vs. a late-cycle pullback, how pricing pressure from Samsung/Micron evolves, and Sandisk's capital needs to expand HBMs and NAND without derailing profitability. The math also glosses over the inconsistency between 2030 earnings forecasts and the implied valuation, which could be a red flag.
Bull case: AI-driven workloads could sustain HBMs and pricing power, potentially supporting a higher multiple and a $1,350 target if demand proves durable and the cycle stays constructive. However, that hinges on a best-case growth path that may not materialize and ignores cyclical risk.
"HBM's contract-driven pricing and superior growth trajectory differentiate this from past memory cycles, supporting sustained margins beyond Claude's normalization assumption."
Claude's claim that margins are 'already elevated due to supply constraints, not structural AI demand' misses that HBM bit growth is running 3-4x faster than traditional DRAM with 60%+ gross margins locked via long-term hyperscaler contracts. This isn't a standard cycle; it's allocation economics. The 2030 EPS collapse forecast assumes supply glut that current capex discipline from the big three makes improbable.
"Hyperscaler pricing power will eventually erode the margin protection currently provided by long-term HBM contracts."
Grok, your 'allocation economics' argument ignores the massive, non-linear capex ramp-up currently underway. You cite long-term contracts as a moat, but hyperscalers like Google and Amazon are notoriously ruthless at renegotiating terms once supply parity is reached. If bit growth outpaces demand by even 10%, those 'locked' margins will evaporate as pricing power shifts back to the buyer. We are seeing a classic supply-side reaction to high prices that you are dismissing as impossible.
"Long-term contracts don't eliminate margin compression risk; they defer it to 2027–28 when renegotiations occur, which consensus already prices in via the EPS cliff."
Gemini's renegotiation risk is real, but both panelists are treating hyperscaler contracts as binary—either locked or broken. Reality: tiered pricing. SNDK likely has volume commitments at fixed prices through 2026–27, then reset clauses. The 2030 EPS cliff in consensus forecasts may reflect exactly this: margin compression post-2027 when renegotiations hit. That's not a supply glut; it's contractual structure. The question is whether SNDK's cost curve improves faster than pricing falls. Neither side has addressed capex intensity required to maintain HBM share.
"Long-term contracts won't guarantee durable margins; true resilience depends on demand durability and capex dynamics, not just contract terms."
Challenging Grok's allocation economics view: locked 60%+ margins via hyperscaler contracts are not an unassailable moat. If SK Hynix/Micron expand capacity or if demand softens, renegotiation or price resets could erode those margins faster than the model assumes. The article's math also ignores potential capex shocks that could lift unit costs. A durable secular upcycle requires more than long-term contracts—it's a real, wide-margin pricing power story.
The panelists generally agreed that Sandisk's current valuation is precarious, resting on the assumption of 'this time is different' in a cyclical industry. They expressed skepticism about the durability of high margins and AI-driven demand, with concerns about potential supply glut, margin compression, and capex intensity.
Sustained high demand for HBM driven by AI, potentially leading to elevated margins.
Margin compression due to supply normalization and potential renegotiation of hyperscaler contracts.