By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
Despite a recent price cut, Susquehanna still sees 80% upside in the stock, but panelists are bearish due to uncertainty around AI memory architecture, NAND price cycles, and execution risks. The panel agrees that the transition to BiCS8/10 NAND and hyperscaler contracts are key opportunities, but these are offset by significant risks such as vertical integration by hyperscalers and potential substitution of NAND with HBM or DRAM.
Risk: Hyperscalers' pivot to HBM or DRAM for AI inference, leading to reduced demand for enterprise SSDs
Opportunity: Cost reductions through BiCS8/10 NAND and potential growth from hyperscaler contracts
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 (SNDK) stock is in focus on July 23 after a senior Susquehanna analyst trimmed his price target on the flash memory products specialist. In a research note today, analyst Mehdi Hosseini set a downwardly revised $3,050 price target on SNDK, which nonetheless signals potential upside of an exciting 80% from current levels.
SanDisk shares have pulled back sharply in recent weeks amid a broader tech rout. At the time of writing, they're down nearly 30% versus their June high.
According to Hosseini, the reduced price objective reflects updated internal estimates following a correction to the firm's financial model, which previously incorporated inaccurate sales and earnings per share (EPS) estimates.
Moreover, the memory architecture for artificial intelligence (AI) inferencing applications remains "somewhat unsettled."
Specifically, the Susquehanna analyst pointed to uncertainty surrounding how inferencing memory hierarchies would split workloads between DRAM and local SSDs.
If AI firms aggressively adopt KV cache offloading, enterprise SSD demand would increase rather significantly. Conversely, if they continue relying mostly on DRAM/HBM, demand for SanDisk's flagship product will be lower.
Despite lowering the price target, Susquehanna remains strongly bullish on SanDisk shares' longer-term trajectory.
According to Barchart, the company is expected to report more than $33 per share of earnings (EPS) for its current financial quarter on Aug. 5. This would represent a staggering 167x growth on a year-over-year basis.
In his report, Hosseini highlighted key structural catalysts for his constructive view, including the ramp-up of BiCS8 and BiCS10 NAND technologies, which are expected to drive about 10% annual front-end cost-per-bit reductions.
Finally, the analyst is positive on SNDK's long-term contractual agreements with hyperscalers as they offer significant visibility into future revenue.
Four leading AI models discuss this article
"The article's core claims rest on non-existent SNDK financials and cannot be taken at face value."
The article frames Susquehanna's PT cut from an implied ~$5,400 to $3,050 as a mere modeling correction while still calling the stock an 80% upside name. That math only holds if SNDK currently trades near $1,700, yet the real SanDisk (acquired by Western Digital in 2016) no longer exists as a standalone ticker. The piece appears to recycle outdated or hallucinated data on 167x EPS growth and BiCS8/10 ramps that belong to WDC or Micron, not SNDK. Real memory names are already pricing in AI-driven NAND demand; any further delay in KV-cache offloading or DRAM substitution would hit margins faster than the article admits.
If this is simply an AI-generated article with fabricated numbers, the real risk is that hyperscalers accelerate SSD offload in 2025 and the actual memory complex (WDC, MU) rerates higher, rendering today's skepticism moot.
"The reliance on a non-existent ticker and the mischaracterization of cyclical recovery as structural AI growth suggests this analyst note is built on shaky fundamental assumptions."
The narrative here is fundamentally flawed. SanDisk was acquired by Western Digital (WDC) in 2016; referring to 'SNDK' as a standalone stock suggests either a hallucinated ticker or a significant reporting error. Assuming the analyst is actually covering Western Digital, the 167x EPS growth figure is likely a distortion of a low-base recovery rather than structural hyper-growth. While the transition to BiCS8/10 NAND is a legitimate cost-reduction lever, the 'unsettled' AI inferencing architecture is a massive risk. If hyperscalers prioritize HBM (High Bandwidth Memory) over NAND for KV caching, the demand for enterprise SSDs will be cannibalized, rendering the 'contractual visibility' argument moot.
If the shift toward edge AI forces a move away from expensive, power-hungry HBM in favor of high-capacity NAND, SanDisk/WDC could become the primary beneficiary of a massive infrastructure pivot.
"A downward PT revision paired with 80% 'upside' is a yellow flag, not green—it signals model uncertainty at precisely the moment when AI memory architecture decisions are being locked in, and SNDK has little control over which way they go."
The headline buries the lede: Susquehanna cut its PT yet still calls it 80% upside. That's not a vote of confidence—it's a recalibration after admitting prior model errors. The real story is KV cache uncertainty. If hyperscalers route AI inference through DRAM/HBM instead of SSDs, SNDK's demand thesis collapses. The 167x YoY EPS growth cited is a single quarter comparison (likely lapping a depressed prior year) and tells us nothing about sustainability. BiCS8/10 cost reductions are real, but commoditized NAND margins are structurally compressed. The hyperscaler contracts sound reassuring until you remember those same customers negotiate aggressively and can pivot suppliers. A 30% pullback in weeks suggests the market already priced in AI upside; this cut may signal consensus is rotating.
If KV cache offloading becomes industry standard (increasingly likely as inference costs dominate TCO), SNDK's SSD attach rates could actually accelerate, making the analyst's caution premature and the 80% target conservative.
"Upside hinges on AI-driven memory demand and BiCS8/10 cost declines, but cyclicality and execution risk threaten the projected 80% upside."
This note frames SNDK as a high-risk, high-uncertainty play: a trimmed target despite an implied 80% upside suggests the bull thesis hinges on AI memory architecture favorable to SanDisk’s NAND and a long runway of hyperscaler contracts. The bullish view rests on BiCS8/10 cost declines and a meaningful enterprise SSD ramp from AI workloads, but the article glosses over NAND price cycles, the pace of AI adoption, and execution risk in ramping new nodes. The reported >$33 EPS for the current quarter seems dubious without one-time factors; without cash-flow and balance-sheet context, the upside is not guaranteed and could vanish if demand or pricing deteriorates.
Bear case: the target cut signals model risk and potential overestimation; even if AI demand materializes, NAND prices tend to cycle lower and BiCS8/10 ramps may disappoint or take longer than expected.
"Post-pullback valuation already discounts most KV-cache substitution risk given oligopolistic NAND supply."
Claude correctly flags the KV-cache risk but underweights second-order supplier concentration. If only two NAND makers remain post-consolidation, hyperscalers' pivot to HBM still leaves WDC/MU with pricing power on legacy enterprise SSD sockets through 2026. The 30% pullback already prices in most architecture uncertainty; further downside requires both HBM substitution and a simultaneous NAND glut.
"Hyperscaler vertical integration poses a greater threat to NAND margins than simple supply-demand cycles."
Grok, you're overestimating pricing power. Even with duopolistic supply, hyperscalers like AWS and Meta are vertically integrating, designing their own silicon and storage controllers to bypass standard NAND vendors. The risk isn't just a NAND glut; it's the structural erosion of the SSD's value proposition as compute-in-memory architectures evolve. If the 'SNDK' entity is actually WDC, they are fighting a losing battle against the commoditization of their own product line by their largest customers.
"Vertical integration risk is real, but the question of whether hyperscalers *want* operational ownership of storage is more decisive than whether they *can* achieve it."
Gemini's vertical integration risk is real, but underestimates WDC's moat in enterprise SSD controllers and firmware—not easily replicated. However, the larger blind spot: nobody's addressed whether hyperscalers actually *want* to own storage stack complexity. If they do, WDC becomes a contract manufacturer with razor margins. If they don't, pricing power returns. The article's silence on this tradeoff is deafening.
"NAND remains core for most workloads; commoditization is not guaranteed despite vertical integration."
Gemini overemphasizes hyperscalers bypassing NAND; in practice, NAND remains core for most storage workloads, and WDC’s controller/firmware moat isn’t trivial to replicate. Commoditization is not guaranteed even with vertical integration. The main risk stays NAND price cycles and capex discipline; if pricing weakens or AI demand softens, the '80% upside' script collapses regardless of who owns the stack. Thus the upside needs disciplined execution and favorable pricing dynamics, not a secular shift away from NAND.
Despite a recent price cut, Susquehanna still sees 80% upside in the stock, but panelists are bearish due to uncertainty around AI memory architecture, NAND price cycles, and execution risks. The panel agrees that the transition to BiCS8/10 NAND and hyperscaler contracts are key opportunities, but these are offset by significant risks such as vertical integration by hyperscalers and potential substitution of NAND with HBM or DRAM.
Cost reductions through BiCS8/10 NAND and potential growth from hyperscaler contracts
Hyperscalers' pivot to HBM or DRAM for AI inference, leading to reduced demand for enterprise SSDs