Longsys launches HK IPO as H1 profit surges 700x on memory price spike

scanx
Reviewed by
Shraddha JScanX News Team
Key Highlights
  • Longsys seeks up to $801 million via Hong Kong IPO with implied market value of HK$195 billion
  • H1FY26 net profit surged to 10.58 billion yuan, up more than 700 times YoY on rising memory prices
  • Revenue jumped 136.3% YoY to 24.09 billion yuan, with enterprise storage segment tripling
  • International sales account for nearly 70% of revenue, led by Zilia and Lexar brands
  • 78.3% of proceeds allocated to R&D for chip design and advanced memory products
powered bylight_fuzz_icon
49904112

*this image is generated using AI for illustrative purposes only.

Shenzhen Longsys Electronics Co. Ltd. launched its Hong Kong IPO on Monday, seeking to raise up to HK$6.28 billion ($801 million) as it reported a record first-half profit fueled by soaring memory chip prices.

The Shenzhen-listed company is offering 26.1 million shares for up to HK$240.60 each, implying a market value of up to HK$195 billion. Citic Securities and Citigroup are underwriting the deal, with cornerstone investors including Lenovo and Ingenic Semiconductor agreeing to buy 18.89% of the shares on offer.

Financial Performance

Longsys reported a dramatic earnings rebound in the six months through June, driven by higher memory prices from AI-related buying. Revenue jumped 136.3% year-on-year to 24.09 billion yuan ($3.38 billion). Net profit soared to 10.58 billion yuan from just 14 million yuan a year earlier, an increase of more than 700 times.

Metric H1FY26 H1FY25 Change
Revenue 24.09 billion yuan ~10.2 billion yuan* +136.3%
Net Profit 10.58 billion yuan 14 million yuan >700x

*Derived from disclosed growth rate and current revenue.

Segment Growth

Enterprise storage was one of the fastest-growing businesses, with revenue tripling year-on-year to 2.14 billion yuan. Products entered the supply chains of major internet companies and server manufacturers. International operations also fueled expansion: Zilia generated 3.95 billion yuan in external sales (up 184.6% YoY), while Lexar’s global sales rose 84.9% to 3.97 billion yuan. Nearly 70% of revenue comes from outside Mainland China.

What the Numbers Show

The divergence between revenue growth (+136.3%) and net profit expansion (>700x) highlights extreme operating leverage. With prior-year profits near negligible levels (14 million yuan), the current period's 10.58 billion yuan profit indicates that fixed costs were largely covered by baseline volume, allowing almost all incremental revenue from higher memory prices to flow directly to the bottom line. This suggests margins are highly sensitive to commodity price cycles.

Business Overview

Founded in 1999 and listed on Shenzhen’s ChiNext board in 2022, Longsys ranks ninth globally by 2025 storage-product revenue, holding about 1.2% of the global market. It markets industrial products under Foresee, consumer storage under Lexar, and overseas industrial products through Zilia. Production of self-designed controller chips has exceeded 280 million cumulative units.

Use of Proceeds

Longsys plans to use 78.3% of IPO proceeds for R&D in chip design and advanced memory products. The balance will fund capacity expansion in Suzhou and Brazil, global brand promotion, and potential strategic investments or acquisitions. Subscriptions close on Thursday, with final pricing expected on Sept. 4 and trading debut set for Sept. 8.

How might a potential downturn in AI-driven memory chip demand impact Longsys's highly leveraged profit margins in the coming quarters?

What competitive advantages will Longsys gain by allocating 78.3% of IPO proceeds to R&D, particularly regarding self-designed controller chips?

Could the expansion of manufacturing capacity in Brazil and Suzhou expose Longsys to new geopolitical or supply chain risks given its heavy reliance on international revenue?

like15
dislike

Longsys unveils AMD-co-optimized AI storage solutions at FMS 2026

scanx
Reviewed by
Ritika DScanX News Team
Key Highlights

Longsys demonstrated its 'Edge AI Storage Fusion' strategy at FMS 2026, featuring AMD-co-optimized AIDIMM memory for AI hosts, high-speed PCIe Gen5 SSDs for PCs, and HLCache technology for mobile devices. The showcase highlighted significant improvements in inference efficiency and memory utilization for local AI workloads.

powered bylight_fuzz_icon
47497663

*this image is generated using AI for illustrative purposes only.

Longsys (301308.SZ) showcased its end-to-end Edge AI storage solutions at the Future of Memory and Storage (FMS) 2026 exhibition in San Jose, California, on Aug. 5, 2026. Under the theme 'Edge AI Storage Fusion,' the company demonstrated technologies designed to address memory bottlenecks and latency in local large language model (LLM) deployment across three core scenarios: AI Agent Hosts, AI PCs, and AI Mobile embedded devices.

AI Agent Hosts: Co-Optimized with AMD

To tackle KV cache expansion and latency challenges, Longsys partnered with AMD to integrate AIDIMM™ high-bandwidth memory, the iSA™ (Intelligent Storage Agent), and AISSD™. A live demo developed jointly with SixUnion showed a system running 70B-, 80B-, and 122B-parameter models smoothly using just 64GB of AIDIMM™ memory. This configuration delivered significantly lower memory usage and higher inference efficiency compared to traditional setups. Longsys also unveiled an AMD Ryzen™ AI Halo platform integrated with AISSD™ and iSA™, which is now AMD AVL-certified and ready for mass deployment.

AIDIMM™ features a native 256-bit interface, up to 307.2 GB/s peak bandwidth, and up to 128GB capacity per module. Its plug-and-play design supports scaled manufacturing. Meanwhile, iSA™, Longsys' proprietary scheduling engine, utilizes MoE offloading, tiered cache management, and predictive prefetching to dynamically balance memory and storage resources in real time.

AI PCs: Next-Generation SSD Innovations

For the AI PC segment, Longsys introduced a DRAM-less PCIe Gen5 SSD powered by its 5nm SPU™ (Storage Processing Unit). The drive achieves 14.8 GB/s sequential read and 13 GB/s write speeds at ≤6.3W power consumption, which is roughly 10% lower than comparable controllers. Built-in lossless compression expands usable capacity by up to 2:1 without impacting speed.

Product Type Read Speed Write Speed IOPS Power Consumption
PCIe Gen5 SSD 14.8 GB/s 13 GB/s N/A ≤6.3W
PCIe Gen5 mSSD 11 GB/s 10 GB/s 2200K/1800K N/A

The company's PCIe Gen5 mSSD reaches 11 GB/s read and 10 GB/s write speeds, with 2200K/1800K IOPS. Available in flexible form factors including M.2 2230/2242/2280, Longsys supplies partners such as Lenovo and ASUS, with manufacturing capacity exceeding one million units monthly.

AI Mobile: HLCache™ and AILPBGA™

In the mobile sector, Longsys' HLCache™ UFS technology intelligently offloads cold data from DRAM to storage. Testing on a Google Pixel 7a revealed that a 4GB LPDDR5 device saw background app capacity rise 64%, from 17 to 28 apps, while DRAM usage dropped 20%, from 3.1GB to 2.5GB. This performance approaches that of a native 6GB configuration while supporting 13B–20B parameter models.

Complementing this is AILPBGA™, a 256-bit, 307GB/s embedded memory solution in a compact 22×22mm package, purpose-built for space- and power-constrained devices.

Global Operations

Longsys operates a multi-brand strategy, including the consumer brand Lexar and the South America manufacturing and operations center Zilia. These entities support localized production, delivery, and technical support worldwide. The company stated it will continue advancing proprietary technologies and collaborating with global ecosystem partners to accelerate the adoption of intelligent computing.

How might Longsys' AMD AVL certification accelerate the adoption of Edge AI in enterprise environments compared to non-certified competitors?

What impact could the 2:1 capacity expansion via lossless compression in Longsys' DRAM-less SSDs have on the pricing dynamics of the AI PC market?

Will the success of HLCache™ technology prompt major smartphone OEMs to reduce base DRAM specifications in future mid-range devices to cut costs?

like17
dislike

More News on Longsys