Monday, May 18th 2026

TetraMem Announces 22nm Multi-Level RRAM Analog In-Memory Computing SoC Milestone

TetraMem Inc., a Silicon Valley-based semiconductor company developing analog in-memory computing (IMC) solutions, today announced the successful tape-out, manufacturing, and initial silicon validation of its MLX200 platform, a 22 nm multi-level RRAM-based analog IMC system-on-chip (SoC).

The achievement marks a significant step toward the commercialization of analog computing architectures based on emerging non-volatile memory technologies, addressing the growing challenges of data movement, power consumption, and thermal constraints in modern AI systems.
As AI workloads continue to scale, system performance is increasingly constrained by the cost of moving data between memory and compute units. Analog in-memory computing offers a fundamentally different approach by performing computation directly within memory arrays, significantly reducing data movement and improving system-level efficiency. TetraMem's MLX200 platform integrates multi-level RRAM arrays with mixed-signal compute engines to enable high-throughput vector-matrix operations within memory, while maintaining compatibility with advanced CMOS processes.

The multi-level RRAM technology demonstrated at the TSMC 22 nm process provides key attributes required for practical deployment, including CMOS compatibility with minimal additional process complexity, low-voltage and low-current operation, strong retention and endurance characteristics, and high multi-level capability that supports improved memory and compute density. Early silicon results indicate consistent functionality across arrays, supporting the viability of this approach for both embedded non-volatile memory and compute-in-memory applications.

This milestone builds on TetraMem's earlier work on the MX100 platform, fabricated on the TSMC 65 nm CMOS process, where the company demonstrated multi-level RRAM devices with thousands of conductance levels ("Thousands of conductance levels in memristors integrated on CMOS," Nature, March 2023), as well as high-precision analog computing capabilities ("Programming memristor arrays with arbitrarily high precision for analog computing," Science, February 2024). These prior results established a strong scientific and engineering foundation for scaling the technology to more advanced nodes.

Since 2019, TetraMem has worked closely with the world leading semiconductor foundry to advance RRAM technology from early-stage research into manufacturable silicon. The progress achieved at 22 nm reflects continued development in process integration, device uniformity, and system-level co-design.

The MLX200 and MLX201 platforms are designed to support power- and latency-sensitive edge AI applications, including voice and audio processing, wearable devices, IoT systems, and always-on sensing. Evaluated sampling is expected to begin in the second half of 2026, and multi-level RRAM memory IP is available for evaluation and potential licensing.

Dr. Glenn Ge, Co-founder and CEO of TetraMem, commented, "This milestone reflects years of close collaboration with our foundry partner TSMC and demonstrates the feasibility of bringing multi-level RRAM and analog in-memory computing from computing architecture breakthrough into advanced-node commercial silicon. We believe this approach provides a practical path to improving energy efficiency and scalability for next-generation AI systems."

The successful realization of the MLX200 platform highlights the viability of multi-level RRAM-based analog computing on advanced semiconductor processes. TetraMem will continue to advance this technology to support emerging AI workloads with improved energy efficiency and system scalability.
Source: TetraMem
Add your own comment

6 Comments on TetraMem Announces 22nm Multi-Level RRAM Analog In-Memory Computing SoC Milestone

#1
lexluthermiester
If they can make it viable, this tech could prove a solid advance.
Posted on Reply
#2
son_of_wasps
lexluthermiesterIf they can make it viable, this tech could prove a solid advance.
Agreed, it looks promising imo. Especially exciting that this is only 22nm still, they could make improvements very rapidly by switching to newer nodes.

Also, I'm curious what the storage size of this chip is... Judging by the wafer size I fear it might be less than a gigabyte. Fine for single-use low-power edge deployment but if it's too too small even that could be limited.
Posted on Reply
#3
Athena
if they can do this at scale, and the price is right, I can see this being used in a ton of things
Posted on Reply
#4
tommo1982
I want to see a detailed description on how a system will work with this RRAM. I bet it's going to be really interesting what role CPU plays in it.
Posted on Reply
#6
son_of_wasps
tommo1982I want to see a detailed description on how a system will work with this RRAM. I bet it's going to be really interesting what role CPU plays in it.
If I understand correctly this kind of chip doesn't require CPU, at least not for compute. Vector multiplication (the compute) is done in-memory by simple electrical properties.

Ofc many systems using this will still have a microprocessor to control the general operation, but theorthetically I think it might be possible to use it standalone, with just ADCs or whatever. Feed the chip some data, get some data back, boom, that's your AI inference.
Posted on Reply
Nov 17th, 2026 13:14 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts