Jiang Bolong: Systematic Deployment of Edge AI Storage Opens a New Paradigm of Hardware-Software Collaboration

Invest in stocks with the Golden Kylin Analyst Research Reports—authoritative, professional, timely, and comprehensive—helping you uncover potential thematic opportunities!

（Source: Securities Times）

In the wave of AI accelerating from the cloud to edge devices, storage is upgrading from traditional media to become a system-level critical capability. At the CFMS|MemoryS 2026 Summit held recently, Jiang Bolong rolled out a series of new products, including SPU (Storage Processing Unit), iSA (Storage Intelligent Agent), HLC (High-Level Cache), and the next-generation high-speed storage media PCIe Gen 5 mSSD. With a data-tiering approach for finer granularity and software-hardware co-design solutions, it balances performance and cost in the AI era and builds competitiveness for edge AI storage systems.

In interviews with media outlets such as Securities Times, Jiang Bolong Vice President and General Manager of the Enterprise Storage Business Unit Yan Shuyin, and company Vice President and General Manager of the Embedded Storage Business Unit Huang Qiang, shared insights into the overall trends in the storage industry, the design thinking behind edge AI storage products, and the company’s capability to deploy an end-to-end supply chain.

Edge AI Storage Changes

Recently, Google launched the TurboQuant technology, claiming it can achieve about 6x memory savings and an 8x improvement in inference speed. The related news has also drawn attention from the industry and capital markets to efficiency upgrades in storage technology, and has led to some short-term fluctuations in market sentiment.

Yan Shuyin told a reporter from Securities Times that the industry continues to explore related technologies for optimizing AI storage efficiency. If breakthroughs can be achieved and applied on the ground, it will further promote large-scale adoption of edge AI application scenarios.

Compared with short-term fluctuations, the long-term supply-and-demand trajectory in the storage market remains a key focus. Some storage controller manufacturers expect it to last at least until 2027. However, the latest forecasts from institutions in the flash memory market suggest that storage product prices have already risen for three consecutive quarters. It is expected that starting from the 3rd quarter of 2026, the rate of increase will gradually converge, and specific product line online prices may show some divergence.

In response to the industry’s diverse judgments on cycles and price trends, Jiang Bolong believes that improving AI storage efficiency through technological innovation and expanding application scenarios remains the core main line for industrial development.

According to the company, as AI technology continues to penetrate edge devices and all kinds of smart terminal applications keep maturing, the industry as a whole has put forward higher requirements for storage performance, architecture, and collaboration capabilities. It is precisely the rapid deployment of edge AI that brings brand-new technical challenges and demand changes to traditional storage architectures, and has also become the core entry point for Jiang Bolong’s targeted product and technology layout.

Software-Hardware Collaboration

Unlike cloud AI, which focuses on specialized storage services for GPUs, edge AI is built around three core requirements: high-performance capacity, SiP (system-in-package) system-level integration and encapsulation, and customized services. The requirements for storage fundamentally differ from those in the conventional standard storage ecosystem.

Yan Shuyin said that with the continuous development of AI technology, traditional data-tiering models are changing. In addition to cold and hot data partitioning, warm data scenarios are increasingly emerging. In response to this change, Jiang Bolong launched the SPU storage processing unit, iSA storage intelligent agent, and the HLC advanced cache technology, enabling intelligent scheduling functions, and jointly tuned them with ecosystem partners to deepen software-hardware collaboration and optimization.

Unlike conventional SSD controller chips, SPU is a dedicated processing unit built for an intelligent storage architecture. The chip is manufactured using a 5nm advanced process, with a maximum capacity of 128TB per disk, effectively addressing the challenge of balancing capacity and cost. The SPU’s core features include two major capabilities: in-storage lossless compression and HLC advanced cache technology. This significantly reduces the required SSD capacity and cost, and can also use HLC technology to sink warm/cold data down to SSDs, saving nearly 40% of DRAM capacity requirements. In addition, by adding iSA’s intelligent scheduling capability, it can enable heterogeneous disk deployment at the edge and be compatible with multiple flash architectures, including SLC and QLC.

As introduced, KV cache warm data is typically stored on local SSDs, and balancing capacity and access speed is an important direction for edge AI storage.

“In the past, the storage industry’s hardware-software system did not perform similarly fine-grained data recognition and tiered management. Against the backdrop of high memory costs, by distinguishing and recognizing data and optimizing high-frequency data scheduling, we can effectively improve flash access speed,” Huang Qiang said. Relying on software-hardware collaboration and embedded storage optimization to improve end-to-end storage runtime efficiency has become the core development direction that the industry widely agrees on.

In terms of ecosystem collaboration, Jiang Bolong, together with AMD, jointly optimized an intelligent agent host based on the Ryzen AI Max+395 processor, enabling local deployment of the 397B model and reducing DRAM usage by nearly 40%. In the embedded storage field, Jiang Bolong and UNISOC (Unigroup Zhanrui) jointly validated that with a 4GB DDR paired with HLC, 20 apps can start in just 851ms, achieving an experience close to the 6—8GB DDR level, and is expected to lower terminal costs.

Enhancing Supply Chain Capabilities

As the edge AI trend places higher demands on storage vendors’ system-level capabilities and customized services, Jiang Bolong has launched the customized service Foundry model, breaking through the bottleneck of traditional single upgrades and delivering comprehensive improvements. This model covers key core links across the entire upstream-to-downstream industrial chain, including chip design, hardware design, firmware software, packaging processes, industrial design, automated testing, materials engineering, and production and manufacturing. Its SiP can integrate multiple chip types—such as SoC, eMMC/UFS, LPDDR, WiFi, Bluetooth, and NFC—into a single package.

Huang Qiang said that with its controller chip design capability, software design capability, and high-end packaging and testing capabilities such as SiP (system-level packaging), Jiang Bolong forms an end-to-end storage capability that can meet future edge AI demands for diversification and customization.

“On the one hand, relying on the advantages of Chinese engineers and the supply chain, we will strengthen the landing and deployment of localized products in China; on the other hand, relying on overseas production bases, we can achieve low-cost and rapid replication of core technology results, reduce the difficulty of overseas manufacturing, stay close to global terminal markets, and leverage regional and cost advantages.” Huang Qiang revealed that the company is highly confident about its future overseas development.

When discussing cooperation with upstream storage original manufacturers, Yan Shuyin emphasized to Securities Times that Jiang Bolong’s advantage lies in scale—it can maintain long-term, good cooperation relationships with upstream vendors and credibility.

“Scale is very important for the storage industry. As AI brings changes to the storage supply relationship, this means that during periods when resources are scarce, original vendor resources will inevitably tilt toward more competitive fields and customers. On the basis of leveraging our existing advantages, we will continue to increase product innovation—this is also an important strategic focus for the company,” Yan Shuyin said.

Massive information and precise insights—available on the Sina Finance app