NextFin News - Shenzhen Longsys Electronics Co., the world's second-largest independent memory vendor, has opened subscription for a Hong Kong initial public offering that could raise about US$800 million, betting that the current memory upcycle will carry it into the city's busiest IPO market in five years. The share sale, priced at up to HK$240.60 per H-share, values the company at a steep discount to its Shenzhen-listed A-shares even as first-half 2026 net profit surged more than 71,000 percent.
Longsys began selling shares to investors on Monday, with the subscription window running through Thursday. The company is offering 26.08 million H-shares, with a maximum price of HK$240.60 each, implying gross proceeds of roughly HK$6.27 billion, or about US$805 million. Net proceeds are expected to reach approximately HK$6.02 billion. If the listing succeeds, Longsys will become the first independent memory manufacturer in China to maintain a dual "A+H" listing, complementing its 2022 debut on Shenzhen's ChiNext board.
The timing is the story. Memory prices have been rising for four straight quarters, and Longsys's profit has exploded as a result: first-half 2026 net profit attributable to shareholders reached 10.58 billion yuan, up 71,528.66 percent from a year earlier, on revenue of 24.09 billion yuan, up 136.26 percent. In the first quarter alone, gross margin hit 55.53 percent, the highest level in the company's public history. The question investors must answer during this four-day subscription window is whether they are buying a memory company at the top of its cycle or a storage-platform company that has found a way to keep the cycle at bay.
The Deal: Terms, Cornerstones, and What the Money Is For
Longsys is selling 26.0778 million shares globally, split 10 percent to Hong Kong retail investors and 90 percent to international institutions, with a 3.9117 million-share overallotment option. Each board lot contains 50 shares, putting the maximum entry ticket at about HK$12,151. Shares are expected to begin trading on the Hong Kong Stock Exchange on September 8, 2026.
The offering has already lined up a deep bench of cornerstone investors. Fifteen cornerstones, including Transsion International, Lenovo Group, CITIC Securities' Hong Kong asset-management arm, Beijing Injoinic Technology, and several supply-chain partners, have agreed to commit about US$151 million at the maximum offer price, which translates to 4.926 million shares, or nearly 19 percent of the base offering. That level of industrial commitment is a signal that Longsys's customers and suppliers see strategic value in the company beyond the quarterly memory-price cycle.
The company plans to allocate roughly 78.3 percent of net proceeds to research and development in chip design and advanced storage products, 11.7 percent to strategic investments and acquisitions, and 10 percent to working capital and general corporate use. The emphasis on R&D is not incidental. Longsys has designed eight controller chips, six of which are already in commercial products, and in March 2026 launched its WM8500 storage processing unit aimed at AI-era workloads. The prospectus frames the raise as a move up the value chain, away from commodity modules and toward customized enterprise and edge-AI storage.
The Numbers Behind the Profit Explosion
Longsys's financial trajectory over the past three years reads like a textbook memory cycle. Revenue grew from 10.13 billion yuan in 2023 to 17.46 billion yuan in 2024, then to 22.77 billion yuan in 2025. Net profit attributable to shareholders swung from a loss of 837 million yuan in 2023 to 499 million yuan in 2024 and 1.42 billion yuan in 2025. Then the upcycle accelerated violently: in the first quarter of 2026 alone, the company earned 3.86 billion yuan, more than its entire 2025 profit, on revenue of 9.91 billion yuan.
The first half of 2026 confirmed the trend. Revenue of 24.09 billion yuan already exceeded the full-year 2025 total, while net profit of 10.58 billion yuan was roughly seven times the prior year's full-year figure. The company's own forecast for the first half pointed to net profit between 9.2 billion and 11.0 billion yuan, and the actual print landed at the top of that range.
But the profit math deserves scrutiny. Longsys sits in the middle of the memory supply chain: it buys NAND and DRAM wafers from integrated device manufacturers such as Samsung, SK Hynix, and Micron, then adds its own controller chips, firmware, packaging, and testing before selling finished modules under the FORESEE, Lexar, and Zilia brands. In an upcycle, module makers enjoy a double benefit: the inventory they built at lower prices appreciates, and they can pass higher prices through to customers. That is exactly what happened in the first quarter of 2026, when gross margin reached 55.53 percent.
The same mechanism works in reverse. Inventory at the end of the first quarter stood at 17.96 billion yuan, up from 11.68 billion yuan at the start of the quarter, an increase of 6.3 billion yuan in three months. Advance payments to suppliers jumped 239 percent, to 4.64 billion yuan, as the company locked in wafer capacity. By the end of June, inventory represented 60.12 percent of total assets. That is a lot of working capital riding on the direction of memory prices. If the cycle turns, the inventory that amplified this year's gains will amplify the losses next year.
Why Hong Kong, and Why Now
Longsys first applied for a Hong Kong listing in March 2025. That application lapsed because the company did not complete its listing hearing within the six-month validity period, a delay that coincided with a weak memory market and a still-cautious Hong Kong IPO window. The company resubmitted on May 29, 2026, cleared its hearing on August 23, and received its overseas-listing filing approval from China's securities regulator, which had been issued in September 2025.
The delay, unintentional as it may have been, worked in the company's favor. Hong Kong's IPO market has revived sharply. New listings in the city raised about US$22.45 billion in the first half of 2026, up nearly 57 percent from a year earlier, the busiest start to a year in five years, according to LSEG data. Semiconductor names have been at the center of the revival: Nexchip Semiconductor raised US$890 million in a Hong Kong listing in July 2026 and jumped as much as 14 percent on its debut. An analysis by the research firm Kharon found that more than 85 percent of Chinese AI-related companies that went public in 2026, 23 of 27, chose Hong Kong over Shanghai or Shenzhen.
For Longsys, a Hong Kong listing solves three problems at once. It opens access to international capital that cannot easily buy ChiNext shares. It creates a currency and jurisdiction that global customers and acquisition targets find familiar. And it establishes a valuation anchor in a market that has shown a clear appetite for semiconductor stories tied to AI. The A-share price has already done much of the heavy lifting: from its October 2022 listing low to July 2026, the stock gained more than 1,500 percent, and market capitalization briefly exceeded 300 billion yuan. As of late May, the A-share market cap stood at about 233 billion yuan.
The Strategic Pitch: Edge AI as a Cycle Dampener
Management's argument is that Longsys is not simply a module assembler riding a commodity wave. The company is repositioning itself around edge AI storage, where demand is tied to device shipments and product differentiation rather than to the spot price of memory chips. In an investor-relations response in August 2026, the company said:
"Our self-developed SPU controller chip combined with iSA storage agent can efficiently identify and offload warm and cold data, reducing DRAM capacity requirements in terminal devices."
The technology, called HLC for high-level cache, uses fast NAND flash to store data that does not need to sit in expensive DRAM. Longsys says the approach, jointly optimized with AMD and Unisoc, can cut DRAM usage in edge AI products by about 40 percent. That sounds counterintuitive for a memory seller, but the logic is that cheaper storage per device expands the total number of devices that can run AI models locally, which expands total storage unit demand even if DRAM content per device falls.
The market numbers behind the pitch are large. Estimates compiled by Omdia, TrendForce, Yole, and China Insights Consultancy put the non-IDM memory market at US$275.4 billion in 2025, growing to US$875.6 billion by 2030, a compound annual growth rate of 26 percent. Within that, AI storage is expected to grow from US$13.8 billion to US$141.5 billion, a 59.4 percent annual pace. The same research suggests that while integrated device makers currently outsource about 20 percent of their memory needs, 75 to 90 percent of end-market storage products ultimately require completion by non-IDM companies.
That is the structural hole Longsys is trying to occupy. The three dominant memory manufacturers, Samsung, SK Hynix, and Micron, have shifted their most advanced capacity and engineering talent toward high-bandwidth memory and enterprise DRAM for cloud AI data centers. That has tightened supply of edge DRAM, including DIMM and LPDDR products, and pushed prices higher. Longsys argues this creates durable room for independent vendors in edge and consumer segments that the IDMs are deprioritizing.
There is evidence the pivot is already underway. Enterprise storage revenue grew 93.30 percent in 2025, to 1.78 billion yuan, the fastest-growing segment. Longsys has become the first storage company in mainland China to commercialize a self-developed controller for UFS 4.1, the high-performance flash standard used in AI smartphones. Its B2B brand, FORESEE, ranked second globally among independent memory brands in 2025, while its consumer brand, Lexar, ranked second globally among independent consumer storage brands with a 3.3 percent share. Overall, the company held 1.2 percent of the global memory market in 2025.
The Counter-Thesis: This Is a Cycle Trade, Not a Platform Trade
The strongest case against Longsys's valuation is the simplest: the company is a memory module maker, and memory is one of the most brutally cyclical businesses in technology. The profit surge is not the result of a new business model; it is the result of rising wafer prices and inventory gains. Every participant in this market has seen this movie before. The 55.53 percent gross margin in the first quarter of 2026 is not a new normal; it is a peak-cycle reading that will compress when supply catches up with demand.
The inventory build is the tell. Adding 6.3 billion yuan of inventory in a single quarter, while advance payments to suppliers rose 239 percent, is a bet that prices keep rising. It is a leveraged bet on the cycle, financed with working capital that now represents 60 percent of the balance sheet. When memory prices peaked in previous cycles, module makers that had stocked up at the top were the first to report write-downs. As one analysis of the sector put it, for memory module makers, excess profit earned during an upcycle can still disappear during a downcycle.
There is also a valuation problem. At the maximum offer price of HK$240.60, the Hong Kong offering implies a market capitalization far below the A-share level of roughly 233 billion yuan. Even at a substantial discount, new investors are paying for a company whose earnings are at a cyclical peak. If 2026 is the earnings high-water mark, the IPO multiple will look expensive in retrospect regardless of how compelling the edge-AI narrative becomes.
The bullish response is that Longsys is not a pure module play. Its controller-chip portfolio, its HLC architecture, and its enterprise-storage traction are real attempts to capture more of the value chain and to make revenue less sensitive to wafer prices. But these initiatives are early. The company's own investor-relations statement noted that robotics and humanoid-intelligence applications:
"Remain in the early stages of development, with a certain degree of uncertainty, and their current contribution to actual performance remains low."
The platform story is a plan, not yet a profit center.
What to Watch: The Signal That Would Break the Thesis
The base case is that Longsys prices at the top of its range, raises close to the full US$800 million, and debuts with a modest gain, supported by the US$151 million cornerstone book and a still-warm Hong Kong IPO window. The company's earnings momentum into the listing is undeniable, and the industrial depth of the cornerstone group suggests the offer will clear.
The upside case requires the edge-AI narrative to be validated faster than expected. If HLC-based products win design wins with major device makers and enterprise storage continues to grow at triple-digit rates through 2027, Longsys could be re-rated from a cyclical module maker into a storage-platform company. That would justify a higher multiple and narrow the discount to the A-share price.
The downside case is the one cycle veterans know well. Memory prices peak in late 2026 as new capacity comes online, gross margin compresses from the 55 percent level back toward the mid-teens, and the inventory on the balance sheet turns from an asset into a write-down risk. In that scenario, the Hong Kong listing becomes a cautionary tale about raising capital at the top of the cycle.
The falsifying signal is specific and observable: if Longsys's gross margin falls below 30 percent for two consecutive quarters while inventory remains above 55 percent of total assets, the structural-pivot thesis is wrong and the cyclical thesis is confirmed. Margin compression alone could be managed; margin compression combined with sticky inventory is the classic signature of a cycle turning.
For investors, the horizon matters. In the short term, momentum and the cornerstone book favor a successful debut. Over the medium term, the question is whether enterprise and edge-AI storage can offset a consumer-memory slowdown. Over the long term, the question is whether an independent vendor can build defensible technology moats against IDMs that control the wafer supply. Longsys is betting the answer is yes; the next two years of quarterly gross-margin and inventory prints will decide who is right.
The memory cycle rewards the impatient and punishes the late. Longsys has arrived in Hong Kong with the wind at its back, but the same wind that fills the sails can capsize the boat when it shifts. This is a company selling a structural story at a cyclical peak, and the market will soon learn which one it actually bought.
Explore more exclusive insights at nextfin.ai.
