The Supply Gap Spilling Beyond HBM: Reading Three Hana Securities Reports as One Story

Korea Invest Insights

Context This post is a follow-up to the HBM-concentration and legacy-DRAM-risk framework laid out in SK Hynix Q2 Earnings Cut and Target Price Held and CXMT IPO and Memory Price Risk. It pairs well with FADU 2Q26 Earnings Preview for the FADU eSSD axis and AI PCB System Bottleneck Thesis for the substrate axis. The related hubs are the AI HBM Hub and the Exclusive Analysis Hub.

TL;DR

Combine the three reports Hana Securities has published over the past two weeks into a single frame, and they read as three scenes in the same story. The supply gap created as the big three shift toward HBM and DDR5 is spilling into the adjacent value chain.
Scene 1, package substrates: The move to larger-area, higher-layer-count DDR5/LPDDR5 substrates is eating into effective substrate capacity. The shortage signal to watch is LTA (long-term agreement) signings (Daeduck Electronics, Simmtech, Haesung DS).
Scene 2, eSSD controllers: AI data-center demand is driving an enterprise SSD surge, and SK Hynix has picked domestic design house ASICLAND as its Gen6 controller partner. Mass production of the FADU-bound Gen6 chip is expected in the second half of 2027.
Scene 3, legacy memory: China’s GigaDevice is filling the DDR4 and SLC NAND space the leaders vacated. Over the past year, SLC NAND ASP has risen roughly 5x and legacy DRAM ASP roughly 10x.
The memory boom has broadened beyond HBM. That said, all three axes still have verification points to check: for substrates, LTA disclosures, pricing, and utilization rates; for eSSD, FADU’s official order wins and revenue recognition; for legacy memory, the speed at which China fills the gap. [Analysis scope: report summary]

Key Framing

The three reports each cover different stocks, but they converge on a single sentence: the supply gap created by the big three's concentration on HBM and DDR5 is spilling into package substrates, eSSD controllers, and legacy memory, broadening the memory boom beyond HBM.

The Throughline: One Gap, Three Branches of Spillover

Read the three reports separately and they look like unrelated stock stories. Overlay them, though, and they are three outcomes branching from a single cause.

Cause: The big three (Samsung Electronics, SK Hynix, Micron) are concentrating their limited wafer and back-end capacity on HBM and DDR5. The result is a supply gap in adjacent components and legacy products.

Spillover: That gap spreads in three directions.

Scene	Where the supply gap sits	Beneficiary axis	Shortage/growth signal
1	Memory package substrates	Daeduck Electronics, Simmtech, Haesung DS	LTA signings, price hikes
2	eSSD controllers	ASICLAND, FADU	SK Hynix Gen6 adoption, mass-production contract
3	Legacy DRAM, SLC NAND	GigaDevice (China)	ASP spike, earnings surprise

Each scene is laid out below together with the figures from the original reports.

Scene 1: Package Substrates, Capacity Erodes Quietly

The first report is Hana Securities analyst Kim Min-kyung’s Buy Recommendation on the Memory Package Substrate Sector (July 3, 2026). [Fact: Hana Securities report]

The Argument: Fears of Price Cuts Are Overblown, the Real Story Is a Shortage

Daeduck Electronics, Simmtech, and Haesung DS all took a sharp share-price hit from AI-bubble concerns triggered by news that Meta would lease out its computing resources externally, compounded by media reports of second-half substrate price cuts. The report’s judgment is that this correction is actually a buying opportunity. Memory IDMs currently posting high operating margins have little incentive to cut prices.

The core logic is erosion of effective production capacity.

Memory package substrate capacity has been flat since the large-scale expansions of 2020-2022.
Back then, the dominant products were DDR4 and LPDDR4; today the dominant products are DDR5 and LPDDR5/5X.
Higher-performance memory and larger die sizes are pushing substrates toward larger area and higher layer counts, which shrinks the number of substrates the same capacity can produce.
Top-tier substrate makers such as Unimicron and Samsung Electro-Mechanics are focused on expanding capacity for FCBGA, their higher-margin product, which widens concerns over the global BT substrate supply-demand balance.

Substrate prices were already raised in April 2026. That increase reflected higher raw-material costs, gold and copper among them, but the report estimates it also partly priced in second-half supply concerns. Substrate demand is set to rise further in the second half as general memory supply to NVIDIA Rubin ramps up and Chinese memory makers expand DDR5 shipments. [Inference: report estimate]

Verification Point: The Shortage Signal Is the LTA

The most useful observation tool the report offers is LTA (long-term agreement) signings. In March 2020, Simmtech signed a single-buyer supply agreement with Micron; in a substrate sector where lead times are short, a contract of KRW 100bn or more was unusual. That disclosure marked the point where the actual shortage intensified and the industry entered a full-fledged price-hike cycle. [Fact: historical precedent]

So the practical implication of this axis is clear: rather than buying the beneficiary stocks first, check supply-contract disclosures, pricing, and utilization rates first. If an LTA disclosure appears, it can be read as a signal of cycle entry similar to 2020.

Scene 2: eSSD Controllers, SK Hynix Has Picked Its Partner

The second report is Hana Securities analyst Kwon Tae-woo’s ASICLAND: The eSSD Partner SK Hynix Chose (July 8, 2026). [Fact: Hana Securities report]

The Company and the Contract

ASICLAND (445090) is a design house founded in 2016 and is the only company in Korea to hold both TSMC VCA (Value Chain Alliance) and Arm Total Design partner status at the same time. Its business splits between non-memory ASIC (AI, HPC, and the like) and memory (eSSD, eMMC), and more than 80% of its employees are engineers.

The key event is its contract with SK Hynix.

On June 29, it disclosed a design and tape-out support contract with SK Hynix for a next-generation eSSD controller.
The contract is worth KRW 31.9bn (43.7% of 2025 revenue), running from May 27, 2025 to December 31, 2027.
Tape-out, on the PCIe Gen6 spec and 5nm process, was completed as of July 1.
The base-case scenario is mass production starting in 2028, but the verification timeline is conservative, leaving room for that start date to move earlier.
The key point is that this is not confined to development. It is a contract structure that carries through to mass production.

SK Hynix’s choice of eSSD controller design house has moved from a domestic Company A for Gen4, to Taiwan’s GUC for Gen5, to ASICLAND for Gen6. [Fact: report]

Demand and Earnings

eSSD demand driven by AI data centers is a confirmed growth axis. In 1Q26, combined revenue at the top five global enterprise SSD makers reached US$18.46bn, up +86.1% quarter-on-quarter and an all-time high (TrendForce). 2026 is expected to be the year enterprise SSD becomes NAND’s largest application. [Fact: TrendForce]

ASICLAND’s mass-production track is also coming online in sequence.

The eMMC card controller is already generating mass-production revenue and is expected to scale to KRW 40-50bn or more per year
The FADU-bound Gen6 controller, on a 6nm process, is expected to enter mass production in the second half of 2027
A contract for the next Gen7 development project could be signed within the year

1Q26 revenue came to KRW 54bn (+242.4% YoY), and the operating loss narrowed to KRW 3bn. With this contract now factored in, 2026 revenue guidance was raised from KRW 160bn to KRW 180bn, and the company is expected to turn profitable for the full year. [Fact: report]

Verification Point: An ASICLAND Contract Is Not the Same as FADU Revenue

There is a distinction to draw here. The report mentions second-half-2027 mass production of the FADU-bound Gen6 controller, which connects directly to current FADU positioning. But there is no guarantee that ASICLAND’s contract translates into customer demand and revenue for FADU. ASICLAND is the design house that designs the controller; FADU is the company that sells the SSD containing that controller. The two companies sit in different places in the value chain. [Inference: value-chain distinction]

So when looking at FADU, what should be checked separately is not the ASICLAND contract disclosure but FADU’s official order wins and revenue-recognition schedule. The direction is set, but revenue is a separate event.

Scene 3: Legacy Memory, China Fills the Space the Leaders Vacated

The third report is Hana Securities analyst Baek Seung-hye’s GigaDevice: A Legacy Memory Company Growing on the Global Supply Gap (July 14, 2026). [Fact: Hana Securities report]

The Company: Same Roots as CXMT

GigaDevice (603986) is China’s number one company by market share in legacy DRAM, SLC NAND Flash, NOR Flash, and MCUs. It is a separate memory company founded in 2005 by Zhu Yiming, the founder of CXMT, who currently also serves as GigaDevice’s board chairman. GigaDevice holds a minority 1.8% stake in CXMT, and CXMT foundries GigaDevice’s DRAM on a contract basis, a collaborative relationship between the two. [Fact: report]

The division of labor is clear. CXMT focuses on DDR5/LPDDR5-and-above commodity DRAM and HBM, while GigaDevice handles specialty memory such as NOR Flash and SLC NAND, legacy DRAM at DDR4/LPDDR4 and below, and MCUs. In other words, this is a structure in which CXMT climbs upward chasing the leaders while GigaDevice absorbs the legacy gap below it.

Earnings: A Surprise Built by the Gap

Preliminary 2Q26 results put numbers to this structure.

Preliminary revenue of RMB 7.3bn, up +226% year-on-year and 46% above consensus
Net profit attributable to shareholders of RMB 5.4bn, up +1,496% year-on-year and 77% above consensus

The strength comes from rising ASPs on core products and expanding share. As the world’s top memory makers shift more of their production mix toward HBM, DDR5, and 3D NAND, the supply-demand balance for legacy DRAM and NAND has tightened, and over the past year, GigaDevice’s SLC NAND ASP has risen roughly 5x and its legacy DRAM ASP roughly 10x. [Fact: report]

Second-Half Catalysts and Verification Points

There are two sources of upside in the second half. First, as Kioxia and Micron scale back legacy NAND production, the rise in SLC NAND prices continues. Through its partnership with CXMT, GigaDevice is one of the few companies able to expand DDR4 capacity over the next two to three years. Its target is to lift global NOR Flash share from around 20% currently to 25% within three to five years. Second, mass shipments of customized 3D DRAM: AI-related customized DRAM for automotive smart cockpits, AI PCs, and robotics is expected to start contributing to revenue from the second half of 2026. [Inference: report outlook]

The dual meaning of this report is itself the verification point. GigaDevice’s rapid growth is both evidence that the memory boom has broadened and a data point for tracking how fast Chinese makers are filling the gap. How quickly China fills the space the leaders vacate also affects the price ceiling on legacy products for Korean memory makers. That is the same logic as the client-DRAM price-ceiling argument covered in CXMT IPO and Memory Price Risk.

Synthesis: A Broader Boom, and Three Buttons to Check

Read the three reports together and the conclusion is singular: the memory boom is no longer just a story about HBM. The more the big three concentrate resources on HBM and DDR5, the more supply gaps open up beside and beneath them, and each gap creates its own separate beneficiary. Package substrates, eSSD controllers, and legacy memory are each riding their own cycle.

That said, all three axes have a check button between narrative and revenue.

Axis	Narrative	Check button to press
Package substrates	Larger area and higher layer counts erode effective capacity	LTA disclosures, pricing, utilization
eSSD controllers	SK Hynix Gen6 adoption, transition to mass production	FADU’s official order wins and revenue-recognition schedule
Legacy memory	China absorbs the gap the leaders left, ASP spikes	The speed at which China fills the gap

In other words, these three reports are not a signal to buy the beneficiary stocks first. They are a map of what to check to determine whether a cycle has actually begun. Does an LTA disclosure appear? Is FADU’s revenue actually recognized? Does China’s legacy capacity expansion push prices back down? These three buttons mark the point where narrative turns into fact.

This post synthesizes the detailed content of three publicly available Hana Securities reports (memory package substrates on July 3, 2026, ASICLAND on July 8, and GigaDevice on July 14, 2026) into a single frame. The targets and outlooks cited are the views of that brokerage. The stocks mentioned are examples used to explain industry structure, not a recommendation to buy or sell any specific stock. The figures and outlooks in the reports are as of their publication date and may change afterward. Investment decisions and their outcomes are the sole responsibility of the investor.