Morgan Stanley's "Memory Grim Reaper" Shawn Kim Issues Another Warning! The Real Bomb Is the "2028 Profit Cliff," Comprehensive Downgrade of Memory Stock Target Prices

Morgan Stanley analyst Shawn Kim once again issued a warning to the memory industry, this time targeting the sector’s “2028 profit cliff.”

According to The Whirlwind Trading Desk, in his latest research report released on November 9, Shawn Kim slashed the target prices for Samsung Electronics and SK hynix, and upgraded Samsung Electronics to his new top pick, replacing SK hynix. He wrote in the report: “Concerns about memory are less about delayed data center construction or AI demand following the US mid-term elections, and more about the risk of the ‘2028 profit cliff’ narrative—the market doubts the sustainability of memory growth momentum as supply begins to catch up with demand.”

He believes the market debate has shifted from the near-term fundamentals to the “2028 profit cliff”—with new capacity gradually coming online by the end of 2027, delays in data center construction, a narrowing HBM premium, and other pressures casting doubt on the sustainability of the memory super cycle. Based on this, Shawn Kim sharply compressed his long-term profit margin assumptions for SK hynix and Samsung to their historical averages, directly lowering the output values of both companies’ valuation models.

This isn’t Shawn Kim’s first warning during periods of high optimism in the memory industry. As cited in WallstreetCN, in August 2021 he published “Storage: Winter Is Coming,” which precisely predicted the ensuing two-year downturn in the semiconductor cycle; in September 2024, his report on HBM oversupply was seen as one of the triggers for that year’s sharp drop in Samsung and SK hynix stock prices; on July 21 this year, he issued another report warning that the AI-driven semiconductor memory boom was nearing an inflection point, predicting memory contract prices would peak in Q4—after which shares of SK hynix and Samsung Electronics again tumbled.

These events have earned him the nickname “Korea’s Semiconductor Grim Reaper” in the Korean market. This report serves as yet another warning amid the ongoing deflation of the memory AI trade.

Valuation Reset: Significant Target Price Reductions, Samsung Rises to Top Pick

Shawn Kim’s latest adjustments are substantial. The target price for SK hynix was slashed by 38%, from 2.6 million won to 1.6 million won, corresponding to a projected 2027 price-to-book ratio of about 1.8x. Compared to the current reference price of 1.132 million won, there’s still about 41% upside, maintaining an “Overweight” rating.

The target price for Samsung Electronics common stock was cut 19%, from 381,000 won to 309,000 won, while the preferred stock target dropped from 304,800 won to 247,200 won. Both maintain overweight ratings, and Samsung Electronics common stock is elevated as the new top pick.

The core driver of these target price cuts is Shawn Kim’s marked downward revision of the long-term sustainable profit margin assumptions for both companies. Kim previously assumed a sustainable DRAM profit margin of around 60% until 2035, but now lowers SK hynix’s 2029-2035 DRAM operating margin average to about 32% (close to the historical average of 31% from 2010-2025) and Samsung’s to about 39%. Meanwhile, the HBM premium over standard DRAM is assumed to narrow from 20 percentage points in 2029 to zero by 2033.

The report also raises near-term profit forecasts for 2026-2028 for both companies, mainly due to the reduction in share count from buyback programs. SK hynix’s EPS forecasts for 2026-2028 change by -1%, +22%, and +29% respectively; Samsung’s are raised by 3%, 16%, and 26% for the same period.

“2028 Profit Cliff”: Four Overlapping Pressures

Shawn Kim points out in the report that the “2028 profit cliff” is the core debate in memory investment. He lists four key factors pressuring long-term profit margin assumptions.

First, supply is responding faster than expected. New capacity will gradually come online around the end of 2027, and recent higher Q4 shipment guidance and faster process migration mean more bit supply than previously anticipated. Chinese suppliers entering the consumer market will also exert additional pressure on bulk memory pricing.

Second, delays in data center construction in 2028. The Morgan Stanley Global Thematic Research Team delayed around 5 GW of US data center projects from 2028 to 2029. As a result, global data center capacity growth for 2028 is cut from about 21% year-on-year to 9%, before rebounding to 20% in 2029. This could reduce the tightness of memory supply and weaken pricing power.

Third, BOM cost constraints limit content growth. Memory already accounts for about 60% of the BOM cost of AI servers, forcing buyers to downgrade high-density 96GB/128GB configurations to 64GB RDIMM. NAND spot prices have fallen about 20% over the past two weeks, with 512Gb TLC wafer prices dropping to about $15, down 27-29% from the March high and below the lows of July-August, creating additional pricing risks for the contract market.

Fourth, the HBM premium is narrowing. Morgan Stanley has cut SK hynix’s HBM operating margin forecast from 78-80% to about 70%, and Samsung’s from about 80% to 65-72%. As competition intensifies and standard DRAM prices normalize, the likelihood of HBM maintaining a high premium diminishes.

Cycle Positioning: Valuation Has Adjusted but the Earnings Cut Has Yet to Arrive

Despite the warnings, Shawn Kim does not take a wholly bearish stance. He offers a nuanced assessment of the current market position in his report.

The report notes that since the June peak, the Philadelphia Semiconductor Index (SOX) 12-month forward P/E has been cut in half from 26.3x to 13.4x—one of the largest valuation compressions since the dot-com bubble burst—while consensus EPS forecasts have actually risen 24% in the same period. The current forward P/E of roughly 13x is in line with the median trough valuations (13.4x) of past down cycles and is at its lowest since the October 2022 Federal Reserve rate-hike dip.

Valuation has done its job. The variable now is the earnings line. Kim reviews the 11 SOX down cycles since 2000, finding that 9 out of 11 were chiefly driven by valuation compression rather than earnings cuts. Only in 2000-2002 and 2007-2008, when demand really collapsed, did earnings get cut more than 50%.

A second wave of declines would require 2028 earnings forecasts to be halved as in 2000-2002 and 2007-2008—not our base case, but this is why the earnings cut phase, not valuations, deserves the most attention now.

Kim also notes that the breadth of EPS revisions in the AI supply chain has narrowed since the July 2026 peak—memory’s 3-month average fell from 69% to 53%, AI compute from 55% to 45%, and networking from 54% to 40%—but remains positive, i.e., earnings forecasts are still being upgraded, just at a slower pace.

Samsung vs. SK hynix: End-of-Cycle, Quality Comes First

For stock selection, Kim elevates Samsung Electronics to his top pick, citing a preference for “the largest, most well-capitalized, and strongest tech leaders” at the end of a cycle.

The report says Samsung’s HBM market share is expected to overtake SK hynix in 2027—TrendForce forecasts Samsung’s 2027 HBM shipments to rise about 82% to 21 billion Gb (about 39% market share), surpassing SK hynix’s 20 billion Gb (about 37%). Meanwhile, Samsung’s more diversified business (including the potential recovery of non-memory segments) offers extra support. At the current reference price of 188,700 won, Morgan Stanley’s target of 309,000 won implies roughly 64% upside.

SK hynix remains an Overweight pick. Kim believes its HBM leadership, strong 2027 earnings, and long-term agreements (LTAs) offering protection into 2028, still provide support. However, the report makes clear that the target price cut “reflects a shorter period of excess profitability after 2028.”

The report also describes the worst-case scenario: if AI financing costs rise or AI monetization progress falls short, buyers may suspend purchases, and already-committed capacity could face a demand shortfall risk.

~~~~~~~~~~~~~~~~~~~~~~~~