Innosilicon LPDDR6 Memory IP Adopted by Major Korean DRAM Maker as AI Bandwidth Race Intensifies
Innosilicon has confirmed a key milestone for its next-generation memory technology: a major Korean DRAM manufacturer is now using the company’s LPDDR6 memory IP. The announcement was made during a semiconductor conference in Shanghai, where Innosilicon presented its expanding portfolio of memory and storage controller IP designed for the AI, high-performance computing, and advanced data center markets.
The company had previously revealed that its LPDDR6 and LPDDR5X memory controller IP had been delivered to its first customers. Now, Innosilicon has clarified that one of those early adopters is a leading Korean DRAM producer, signaling growing international interest in its high-speed memory interface technology.
At the event, Innosilicon showcased a wide range of advanced semiconductor IP solutions, including HBM4, HBM3E, GDDR7, LPDDR6, LPDDR5X, DDR5 MRDIMM, PCIe 6.0, and 224G SerDes. These technologies are aimed at solving one of the biggest challenges in modern computing: the memory wall.
As artificial intelligence models become larger and more complex, processors need faster access to memory to maintain performance. AI inference, large-language-model training, high-frequency KV cache operations, and long-context workloads all place enormous pressure on memory bandwidth. Without faster memory interfaces, even powerful GPUs and accelerators can be held back by data transfer bottlenecks.
Innosilicon says its LPDDR6 memory IP is designed to reach speeds of up to 14.4 Gbps. That level of performance could help improve memory read and write bandwidth, reduce latency, and increase stability in AI computing systems. The company also emphasized that it has used end-to-end timing optimization and repeated signal integrity simulations to reduce transmission jitter and signal loss.
The LPDDR6 IP includes several advanced features, such as support for I/O speeds up to 14.4 Gbps, Link ECC, auto and manual ECS, System Meta Function Mode, multiple refresh modes, Read-Modify-Write with Mask Write, dynamic write NT-ODT, DVFSL mode, x24 and x48 burst modes, and efficiency modes designed for both normal and dynamic operating conditions.
For AI workloads, these improvements are especially important. Faster and more stable memory can directly impact token inference speed, single-node AI performance, and the ability to handle long-context large models. Innosilicon is positioning its LPDDR6 and broader memory IP lineup as a foundation for next-generation AI hardware.
Another major point highlighted by the company is its reliance on China’s domestic semiconductor supply chain. Innosilicon claims its 14.4 Gbps LPDDR6 IP was developed using a domestically available toolchain, showing an effort to reduce dependence on foreign technologies for critical chip design capabilities.
The company also stated that its IP and services can scale across process nodes from 28nm down to 3nm. This is notable because China’s access to the most advanced chip manufacturing equipment remains restricted, particularly in areas involving cutting-edge EUV lithography. While domestic lithography development continues, the global semiconductor industry is increasingly relying on EUV for advanced chip production.
Innosilicon’s announcement arrives as the broader memory industry prepares for LPDDR6 mass production. Leading memory companies are expected to push LPDDR6 into high-volume manufacturing around the second half of 2026. Demand is expected to come from AI devices, premium smartphones, edge computing systems, automotive platforms, and high-performance mobile processors.
The adoption of Innosilicon’s LPDDR6 IP by a major Korean DRAM manufacturer is an important signal for the company’s ambitions. It suggests that its memory interface technology is gaining traction beyond China and could become part of the next wave of high-bandwidth memory solutions.
At the same time, the development shows how competitive the global DRAM and memory IP market has become. As AI systems continue to demand faster data movement, companies are racing to deliver LPDDR6, GDDR7, HBM4, and next-generation interconnect technologies. Innosilicon’s latest announcement places it firmly in that race, with a focus on high-speed memory IP, domestic development, and AI-ready bandwidth solutions.
