Why Morgan Stanley lifted target prices for SIMO, Longsys, and Phison
Morgan Stanley’s latest report delivered one of the sharpest re-ratings in the storage space, arguing that the NAND market is no longer being defined primarily by the traditional consumer electronics cycle. The firm raised Silicon Motion (SIMO.O) from $155 to $400, Longsys from RMB 300 to RMB 673, and Phison from TWD 2,248 to TWD 2,588. The key thesis is not just another round of SSD price recovery. Instead, the bank believes AI data centers are shifting NAND demand away from the old smartphone, PC, and client SSD inventory cycle and toward enterprise SSDs, AI boot drives, and long-duration procurement from cloud operators.
Among the three names, SIMO saw by far the most aggressive revision. Morgan Stanley’s new $400 target implies 23x 2027 expected EPS, and the bank expects the company’s 2026 revenue to reach a record high. Longsys and Phison also stand to benefit from firmer memory pricing and AI-server-related storage demand, but the bank still rates both at Equal Weight. That distinction matters. It suggests the current upcycle is not a uniform uplift across all storage module vendors. Instead, valuation expansion is likely to be concentrated in companies with more direct exposure to enterprise SSDs, high-value controllers, and AI-specific storage components.
AI servers are pushing NAND toward a supply shortage through 2027
The central conclusion of the report is that AI-driven NAND demand could remain a meaningful force through 2027. Morgan Stanley estimates that global NAND supply and demand will still show around a 2% surplus in 2025, reflecting the lingering impact of the previous oversupply cycle. But by 2026, the market could swing to a 15% shortage, and by 2027, even with additional supply coming online, the deficit may still be about 9%.
The report provides explicit volume estimates. Total NAND demand for 2025, 2026, and 2027 is projected at 1111 EB, 1250 EB, and 1484 EB, respectively. Supply over the same period is estimated at 1128 EB, 1058 EB, and 1347 EB. That implies the market moves from a modest surplus to a notable deficit within two years. The importance of that shift is not simply its scale, but the source of the demand. Morgan Stanley is not relying on an aggressive recovery in smartphones or PCs. In fact, its assumptions for those end markets are relatively restrained: NAND content per device is broadly flat, and unit shipments still follow a declining hardware-team model.
That means the shortage thesis is built much more on AI servers, cloud SSD procurement, enterprise storage build-outs, and boot-drive demand than on a broad-based recovery in consumer electronics. By 2027, Morgan Stanley expects AI-related NAND demand to grow 60% year over year to 609 EB, accounting for 41% of total NAND demand. Against total demand of 1484 EB and supply of 1347 EB, the market would remain structurally tight.
This is a meaningful change in how NAND should be analyzed. Historically, the market was more tightly linked to smartphone and PC inventory corrections. In the emerging AI cycle, servers need not only GPUs and HBM, but also substantial local storage, enterprise-grade SSDs, and dedicated boot drives. Once cloud service providers shift more of their procurement into long-term agreements, the timing and volatility of NAND pricing may look very different from previous memory cycles.
Price strength is real, but it is concentrated in data-center-related products
Channel checks already show that pricing strength is not evenly distributed across the memory stack. According to Morgan Stanley’s 3Q26 checks, TLC enterprise SSD prices were up about 30% quarter over quarter. Server DRAM prices rose 20% sequentially, while legacy DRAM categories such as DDR3 and DDR4 increased by 30% to 40%. By contrast, consumer NAND saw much smaller gains.
The reason is straightforward. Smartphone and PC customers are operating under tighter margin pressure and have less ability to absorb large price increases. Data-center customers, especially those deploying AI infrastructure, are buying into a different economic model. Their purchasing decisions are tied more closely to capex cycles, platform deployment schedules, and long-term utilization assumptions than to short-term retail demand conditions. As a result, the strongest price moves are emerging in enterprise SSDs and server-adjacent memory products, not across all NAND categories.
For investors, that distinction is critical. Calling this simply an “SSD price upcycle” misses the fact that profitability and pricing power will likely diverge sharply between enterprise-focused and consumer-focused segments. The products most directly tied to AI infrastructure are already demonstrating stronger pricing behavior than general-purpose NAND sold into phones and PCs.
Why SIMO is viewed as a direct beneficiary of AI storage growth
Morgan Stanley’s much higher target for SIMO is rooted in two specific growth vectors: enterprise SSD controllers and AI boot drive modules. The report identifies the company’s MonTitan enterprise SSD business as one of its most important growth engines over the next several years. Morgan Stanley expects MonTitan to contribute 5%, 13%, and 19% of SIMO’s revenue in 2026, 2027, and 2028, respectively.
At the same time, boot drive modules are expected to ramp materially. The bank projects that this segment will contribute roughly 15% of revenue in 2026 and 21% in 2027. In AI servers, the boot drive is not the most visible component, but it is essential for system initialization, management, and operating stability. As AI server shipments rise, demand for the associated controllers and modules should increase alongside them.
This is why Morgan Stanley believes the market may need to reassess SIMO’s identity. Historically, the company has often been viewed more as a consumer-oriented controller supplier. The new valuation framework assumes that enterprise and AI-linked revenue will become a much larger share of the business in a relatively short period. If that transition materializes, earnings quality and multiples could improve meaningfully.
That said, the report does not present this as fully de-risked upside. The $400 target rests on several conditions being met: a successful ramp in enterprise SSDs and boot drives, continued customer adoption, and no significant slowdown in AI server demand. If any one of those variables disappoints, the justification for a 23x 2027 EPS multiple would become more difficult to defend.
Why Longsys and Phison still face constraints despite higher targets
Longsys and Phison also stand to benefit from tighter storage markets and stronger enterprise demand, but Morgan Stanley stopped short of upgrading their ratings beyond Equal Weight. The reason lies in the position module makers occupy in the supply chain. When NAND supply becomes tight, manufacturers are more likely to allocate available output first to large cloud customers and core CSP accounts. That may limit the incremental volume available to module vendors.
In other words, higher prices are helpful for inventory revaluation and average selling prices, and an improved enterprise mix can support gross margins. But if the upstream supply base and top-tier customers lock in most of the volume, the revenue elasticity of module providers may be capped. This is why target prices can move higher without necessarily implying a more bullish rating stance.
The report also highlights long-term agreements, or LTAs, as an important structural factor. Suppliers can obtain some downside price protection through LTAs, and Kioxia’s 2027 LTA coverage is expected to exceed 50%. However, LTAs are not a one-way positive. Micron has also pointed out that such contracts usually include both price floors and price ceilings. That means they can reduce the risk of a sharp price collapse while also limiting upside in a severe shortage environment.
Module makers, meanwhile, are trying to push more inventory burden onto customers through structures such as TCM. Their goal is to stabilize long-term gross margins in a 25% to 35% range. But success still depends on customer acceptance, the tightness of supply, and whether their product mix is differentiated enough to command better economics.
2028 remains the key boundary for the bullish NAND thesis
Despite the constructive medium-term view, Morgan Stanley is clear that the biggest uncertainty sits in 2028. Under the bank’s base-case scenario, even by that year the market could still show roughly a 5% shortage if AI NAND demand continues to grow 60% year over year and YMTC capacity remains around 310kwpm. But if YMTC expands to 470kwpm and AI growth slows at the same time, the NAND market could shift from shortage to balance, or even back into oversupply.
The report’s scenario matrix effectively frames 2028 as a function of two variables: YMTC capacity expansion and AI SSD demand growth. Across combinations of 310-470kwpm of YMTC capacity and 30%-60% AI growth, the outcome can move from clear shortage to near balance or surplus. That is exactly why memory cycles remain so hard to time. Strong near-term prices and low inventories often reinforce bullish expectations, but once supply discipline weakens, oversupply can return quickly.
There are already signs that the consumer side is less able to absorb higher prices. Some orders have been reduced, and smartphone and PC customers appear constrained in what they can pay. That suggests the pricing ceiling for consumer NAND may arrive earlier than for enterprise products.
Ultimately, the most important question raised by the report is not simply whether SSD prices will rise. The deeper issue is whether AI demand will remain strong enough to absorb new supply over the next two years. For companies such as SIMO and other firms exposed to controllers and AI storage infrastructure, 2026 may represent the beginning of a meaningful enterprise and AI revenue ramp. For the broader NAND cycle, however, the deciding variables are likely to be 2028 capacity additions at players such as YMTC, the intensity of CSP capital expenditure, and whether supply discipline across the industry holds.