top of page

The Future of Memory

Resistive Random-Access-Memory (ReRAM) is a very promising technology for embedded Non-Volatile Memory (NVM).

The IT Press Tour had the opportunity to meet with executives from Weebit Nano. With memory at the center of the digital world, Weebit Nano is enabling innovation with drones, autonomous vehicles, IoT, artificial intelligence (AI), robotics, security, and 5G.

Every year the world uses more memory. New applications are driving an explosion in data usage leading to an increasing need for more memory.

Memory is a critical market segment in modern data-centric societies and is driven by important megatrends, such as mobility, cloud computing, AI, and IoT. The long term demand . . . will result in memory continuing to increase its share of the overall semiconductor market. – Yole’ Development, February 2021

Weebit is targeting both segments of the non-volatile memory (NVM) market with embedded memory modules and discrete memory chips. Weebit is focusing on non-volatile ReRAM in both. There is a strong need for new memory technology to keep pace with technology developments.

Weebit’s ReRAM is made of simple, fab-friendly materials. Two-mask added with very few added steps. Easily integrated into any fabrication using various deposition techniques to allow for manufacturing flexibility. Rapid development by using CMOS compatible process materials. The result is highly scalable, low cost and effort, low power consumption, and high endurance and reliability.

The module design is integrating Weebit ReRAM array in a complete module in 130nm technology to demonstrate IP module capabilities. The design has begun with a tape-out planned for 2Q21. The first silicon is planned for the end of 2021. The ReRAM module is further integrated into a complete subsystem based on the RISC-V processor, peripherals, and I/Fs. Customers can use module-as-platform to accelerate the development of applications such as low-energy IoT devices, security, and sensors.

Weebit sees a significant embedded memory opportunity replacing current technology with embedded flash. The embedded emerging market revenue in the U.S. was about $26 million in 2020 and is projected to be $2.3 billion in 2026 – 110% CAGR. Weebit overcomes the limitations of existing embedded NVM with improved power consumption, switching speed, and scalability.

IoT is a natural fit for ReRAM. Billions of battery-operated edge devices require:

  • Low power consumption -- It is not feasible to change batteries frequently, or at all. Flash consumes too much power for coin-cell batteries and energy-harvesting IoT applications.

  • Minimized cost -- Weebit ReRAM is more than 4X lower cost compare to other emerging NVM technologies.

  • High-level of integration – NVM can be easily integrated into SoC, there are no discrete components.

  • Retention and endurance even in high temperatures for years of uninterrupted field operation.

According to IDC, by 2025 there will be 55.7 billion connected devices worldwide and 73.1 ZB of data generated from connected IoT devices.

ReRAM is also ideal for analog integrated circuits (IC) which have a diverse set of technologies including power, mixed-signal, RF, MEMS, and more. They increasingly require embedding low-density, high-endurance NVM. Support many components, most built-in front-end-of-line (FEOL), involving complicated tradeoffs. Support many process nodes and derivatives.

The challenge is that e-flash is integrated with FEOL, forcing compromise with PMIC/analog components, leading to degraded performance, larger size, and higher cost. ReRAM Is integrated with Back-end-of-line (BEOL), allowing full optimization of PMIC/analog components so there is no impact on design rules; optimized performance, and size. As a low-cost, high-performance, highly scalable BEOL technology, Weebit’s ReRAM enables key requirements for embedded NVM in analog IC’s.

Weebit’s vision for the future of ReRAM technology includes mimicking the brain as accurately as possible. ReRAM resembles a biological synapse. Physical similarities lead to functional similarities. They are highly energy efficient which makes them an enabler for brain-inspired AI systems using ReRAM.

Weebit collaborates with research in both academia and industry. First demonstration of ReRAM-based spiking neural networks in 2019 with CEA-Leti. Research continues with Politecnico di Milano in Italy (Polimi), Technion in Israel, IITD in India, and others.


bottom of page