Weebit Nano ReRAM to Power Korea's Bid for AI Chip Supremacy

HOD HASHARON, Israel & SEOUL, South Korea – March 05, 2026 – In a significant move that validates a new frontier in artificial intelligence hardware, Israeli technology developer Weebit Nano has been selected to provide a foundational technology for a major Republic of Korea government-funded program. The company’s Resistive RAM (ReRAM) will be a core component in a national initiative to develop ultra-low-power analog compute-in-memory (ACiM) technology, aiming to position South Korea at the forefront of the next generation of AI semiconductors.

This selection marks a critical milestone, transitioning Weebit Nano’s innovative memory technology from a promising R&D component to a cornerstone of a strategic national project backed by a global semiconductor powerhouse. The program brings together industry, academia, and government to tackle one of the most significant challenges in modern computing: the immense energy consumption of AI.

The Quest to Overcome AI's Energy Crisis

Modern artificial intelligence, particularly deep learning, has an insatiable appetite for data and power. The conventional computer architecture, known as the Von Neumann architecture, separates processing units (CPUs/GPUs) from memory units (RAM). This design forces a constant and massive shuttle of data between the two, creating a bottleneck that wastes both time and, more critically, energy. This “memory wall” is a primary reason why AI data centers consume vast amounts of electricity and why running sophisticated AI on small, battery-powered devices remains a formidable challenge.

The South Korean initiative aims to shatter this wall using analog compute-in-memory, or ACiM. This radical paradigm shift performs computations directly within the memory itself. Instead of moving data, the AI model’s parameters (or weights) are stored within a dense array of memory cells. The computation, specifically the vector-matrix multiplication that forms the backbone of neural network processing, occurs in place through the physical properties of the memory array. This can dramatically reduce data movement, leading to orders-of-magnitude improvements in energy efficiency and performance.

The consortium's ambitious goal is to develop silicon-verified ACiM blocks targeting an energy efficiency of approximately 200 TOPS/W (Tera Operations Per Second per Watt). This figure represents a monumental leap over many current AI accelerators, where practical efficiency is often limited by data transfer overhead. Achieving such efficiency would enable more powerful AI in everything from edge devices and autonomous vehicles to massive data centers, all while significantly reducing their energy footprint.

South Korea's Strategic AI Gambit

This program is more than just a technological endeavor; it is a key piece of the Republic of Korea’s broader “AI Transformation Initiative.” The government is strategically investing in foundational technologies to secure a leadership position in the global AI hardware race, fostering a self-sustaining domestic ecosystem that spans research, design, and high-volume manufacturing. By funding this consortium, South Korea aims to build domestic capability and reduce reliance on foreign-designed AI chips.

The project assembles a powerhouse of Korean expertise. Weebit Nano has extended its agreement with DB HiTek, a leading Korean foundry that will manufacture the advanced silicon. They are joined by top-tier academic institutions—Daegu Gyeongbuk Institute of Science and Technology, Seoul National University, and Chungbuk National University—which will provide deep research into device physics and circuit design. The government’s own Electronics and Telecommunications Research Institute (ETRI) will help bridge the gap between academic research and industrial application, while a dedicated company, AnalogAI, has been tasked with commercializing the resulting ACiM products. This integrated structure provides a clear pathway from the lab to the market.

Fred Kim, General Manager of Sales at DB HiTek, highlighted the national significance, stating, “This project is part of the Republic of Korea’s broader AI Transformation Initiative, which supports technologies critical to future AI semiconductor leadership. By combining emerging memory devices with proven CMOS manufacturing, the consortium aims to significantly improve AI energy efficiency while building domestic capability and a sustainable ecosystem spanning academia and industry. Weebit ReRAM is the ideal memory device to use as a foundation for this work.”

ReRAM: A Synapse for the Silicon Brain

At the heart of this initiative is Weebit Nano’s ReRAM technology. This non-volatile memory is uniquely suited for ACiM applications due to its fundamental characteristics. Unlike conventional digital memory that stores a '0' or a '1', ReRAM cells can be programmed to hold a wide range of resistance values, exhibiting analog behavior that closely mimics the synaptic connections in the human brain. This allows for the storage of multi-level AI weights and the execution of analog computations with exceptional efficiency.

Furthermore, Weebit Nano's technology is based on Silicon Oxide (SiOx), a material that is standard in virtually every semiconductor fabrication plant worldwide. This “fab-friendly” nature means it can be integrated into existing CMOS manufacturing flows at foundries like DB HiTek without requiring exotic materials or expensive new equipment. This manufacturability is a crucial advantage over other emerging memory technologies that may offer theoretical benefits but pose significant production challenges, and it is a key reason for its selection.

Coby Hanoch, CEO of Weebit Nano, commented on the program's practical focus: “AI system designers are increasingly looking to bring memory closer to compute to reduce power and latency. In memory compute is a practical path toward that goal, but it requires validation at realistic scales. This initiative combines device innovation, circuit and architecture co-design, and manufacturable silicon, which is exactly what’s needed to move ACiM from research to deployable technology.”

From Prototype to Production: Charting the Commercial Path

The program's key objective is to move beyond small-scale academic test chips and create large, device-array-based silicon implementations. The consortium is focused on establishing a complete and repeatable development flow for ACiM, a crucial step for commercial adoption. By co-optimizing the device physics, circuit design, and system architecture, the project aims to create not just a proof-of-concept, but a robust and verifiable technology ready for real-world AI applications.

The involvement of DB HiTek ensures that the technology is developed with high-volume manufacturing in mind from day one, while AnalogAI’s participation guarantees a commercial focus on creating market-ready products. This strategic alignment significantly de-risks the technology and accelerates its path to potential revenue through licensing and royalties for Weebit Nano.

This collaboration between an Israeli innovator and a South Korean national consortium underscores the global nature of the race for next-generation computing. As AI models become more complex and their energy demands continue to soar, the technologies being developed in this program could define the future of efficient and sustainable artificial intelligence, with Weebit Nano’s ReRAM playing a pivotal role in that transformation.