Development of SRAM computing-in-memory AI inference chips, HBF storage and AI inference servers
深圳市九天睿芯科技有限公司 (REEXEN)
Shenzhen (Futian, Shenzhen New Generation Industrial Park, Building 3, 1501-1502) · CNNew on Amplifa
About REEXEN
Shenzhen Jiutian Ruixin Technology Co., Ltd. (REEXEN) is a technology company for AI inference architecture founded in 2019 in Shenzhen. At its core is the MPM methodology (Memory–Processor–Model), in which memory, processor and model are co-developed. The company develops SRAM computing-in-memory (CIM) chips, realizes stacking of NAND or DRAM memory on top of the compute chips via 3D vertical integration with TSV and hybrid bonding, and delivers with HBF (High Bandwidth Flash) a solid-state memory architecture for model weights, KV cache and checkpoints. The ADA product family comprises the inference boards ADA300N (NAND stack), ADA300D (DRAM stack) and the ADA400-DN board with over 2 TB capacity and over 4 TB/s bandwidth for trillion+ parameter models, complemented by AI inference servers with system software for LLM, multimodal and agent applications. Fields of application include cloud inference platforms, edge nodes, robotics, end-cloud collaboration and AI memory systems. The company is located in the Shenzhen New Generation Industrial Park in the Futian district and has been recognized as a National High-Tech Enterprise as well as a Specialized and Innovative SME.
- Site
- 0 Shenzhen, Futian (Meilin)
- Processes
- SRAM 存算一体芯片设计 (SRAM Computing in Memory), 3D 垂直集成/堆叠 (TSV, Hybrid Bonding, NAND/DRAM 垂直堆叠), HBF 高带宽固态存储 (HBF SSD ADA400), AI 推理板卡与服务器 (ADA400-DN-BOARD, 2 TB+ 容量, 4 TB/s+ 带宽), AI inference chip and board design, HBF high-bandwidth solid-state storage, AI inference servers and system software, Computing-in-memory (SRAM-CIM) technology
- Materials
- SRAM, NAND (闪存), DRAM, SRAM Computing in Memory, High Bandwidth Flash (NAND), DRAM (vertically stacked)
- Certificates
- National High-Tech Enterprise, 专精特新中小企业 (Specialized/Innovative SME)
- Founded
- 2019
- Industries
- Generative AI / LLM inference, Edge computing, Robotics and edge nodes, End-cloud collaboration, Cloud AI inference platforms, AI memory
- 01
MPM system definition and chip architecture
- 02
SRAM CIM chip design (custom circuits, data flow, software mapping)
- 03
3D vertical integration (TSV, hybrid bonding, NAND/DRAM stacking)
- 04
HBF 固态存储 development (ADA400)
- 05
Board design and server integration
- 06
System software and model mapping
Technologies & Memory Technology
SRAM Computing in Memory
Memory and computation merge at the transistor level; multiply-accumulate operations are performed directly in the SRAM array with reduced data transfer and energy consumption.
High Bandwidth Flash (NAND)
High-bandwidth flash organizes model weights, KV cache and checkpoints closer to the computation with high capacity and sustained bandwidth.
DRAM (vertically stacked)
Stacked DRAM (ADA300D) provides high-bandwidth, low-latency access for attention computations and hot KV cache.
Locations
Shenzhen (Futian)
CNHeadquarters, chip and system development
- Materials
- SRAM, NAND, DRAM
- SRAM CIM chip design
- 3D integration (TSV/hybrid bonding)
- HBF SSD development (ADA400)
- Board and server integration (ADA400-DN board)
Industries
Processes & technical limits
| Processes | Tolerance | Max. part | Weight | Lot size | Materials |
|---|---|---|---|---|---|
| SRAM 存算一体芯片设计 (SRAM Computing in Memory)Engineering | – | – | – | – | SRAM |
| 3D 垂直集成/堆叠 (TSV, Hybrid Bonding, NAND/DRAM 垂直堆叠)Engineering | – | – | – | – | NAND, DRAM |
| HBF 高带宽固态存储 (HBF SSD ADA400)Engineering | – | – | – | – | NAND |
| AI 推理板卡与服务器 (ADA400-DN-BOARD, 2 TB+ 容量, 4 TB/s+ 带宽)Assembly | – | – | – | – | – |
Additional services
Quality & Awards
The company has been recognized as a National High-Tech Enterprise (2020, recertification 2024), hosts a postdoctoral innovation practice and is a member of the Supercomputing Internet working group of the national standardization body for intelligent computing.
| Evidence | Shenzhen (Futian) |
|---|---|
| National High-Tech Enterprise | |
| 专精特新中小企业 (Specialized/Innovative SME) |
History
- 2019
Company founded in Shenzhen; first award wins, including 1st place in the Shenzhen Innovation & Entrepreneurship Competition (Electronics/IT) and the FPGA competition of the China Smart Expo.
- 2020
Recognition as a National High-Tech Enterprise.
- 2021
Alibaba DAMO "Pingtouge Track Star" (Futian, Shenzhen); Technology Innovation Award of the Shenzhen High-Tech Promotion Association.
- 2022
Shenzhen "Peacock Team" for industry-academia-research technology development; NICT Innovation Product Award for ADA100; inclusion in the China Potential Unicorn List.
- 2023
Recognition as a Specialized and Innovative SME (专精特新); Top 30 China AI Chip Pioneer at the Global AI Chip Summit.
- 2024
Recertification as a National High-Tech Enterprise; Shenzhen Major Science & Technology Project; member of the Supercomputing Internet standards working group.
- 2025
Series B financing round of over 100 million CNY; establishment of a postdoctoral innovation practice; Top 100 Shenzhen Industry Leaders.
- 2026
Series B+ financing round; presentation of a solution for trillion-parameter model inference on a single card (HBF/SSD as an HBM alternative).