DorsaVi (ASX:DVL) is Combining the Best of 3 Worlds for Tomorrow’s Edge AI Devices

Silicon With a Brain

DorsaVi’s (ASX:DVL) Resistive Random Access Memory (ReRAM) development is progressing toward integration at the 22nm (nanometer) technology node. Now, for many investors, this is highly technical language, but the key point is simple and we will break down what all this means for the company and how it ties into its newly acquired IP (Intellectual Property).

Last month, DorsaVi purchased neuromorphic processing-in-memory technology from Technion, led by Professor Shahar Kvatinsky, a globally recognised pioneer in neuromorphic hardware. This was a significant move for the business. It brings world-class brain-inspired computing technology into DorsaVi’s hands at exactly the time the industry is moving towards this kind of technology.

This acquisition positions DorsaVi at the forefront of one of the fastest-growing segments of the semiconductor market. The global neuromorphic computing sector is projected to rise from US$5.3 billion in 2023 to more than US$20 billion by 2030, according to IBM, a shift that highlights just how transformative this space is becoming.

What are the Best ASX Semiconductor Stocks to invest in right now?

Check our buy/sell tips

The Bottleneck Problem, Why Today’s Computers Waste So Much Energy

A quick refresher for investors is that there’s a fundamental problem with the energy-intensive nature of today’s common computing processes. In modern systems, we still follow the classic Von Neumann architecture, where graphics processing units and central processing units are separated from the memory module. That means these components constantly send massive amounts of data back and forth to perform calculations before storing the computed information in memory.

On top of that, many software applications then need to send this data repeatedly to external servers in a remote data center, adding another layer of back-and-forth communication. All of this movement takes up a huge amount of energy. It also creates latency, slows performance and limits efficiency because the system is spending more time shuffling data around than actually processing it where it lives.

DorsaVi’s Brain-Inspired Chip Breaks the Old Model

Neuromorphic processing, in DorsaVi’s case, reshapes how computing works. Instead of the memory chip acting purely as a passive storage unit, the memory itself can also process data at the same time. It’s similar to how the human brain operates, using neurons to process information and synapses to store data simultaneously. That parallel behaviour is what makes the biological brain so incredibly energy efficient.

In DorsaVi’s neuromorphic memory-processing chip, there is a neuron-like and synapse-like circuit architecture. The chip only responds when data comes in, meaning it doesn’t need to be constantly powered or running. On top of that, it activates only when the incoming data is deemed “relevant” and adjusts itself based on the “importance” or relevance of the incoming data. This behaviour mirrors a spiking neural network, the same biological principle that underpins the efficiency of the human brain.

To put this into a real-world example for DorsaVi’s sensor business: imagine a customer wearing one of their devices. When the wearable detects muscle movement, it can store, interpret and even act on that data within the device itself, rather than sending everything out to an external data centre for processing and then waiting for the response to come back. The result is much faster decision-making, lower energy use and far more efficient, on-device intelligence.

DorsaVi and Artemis Push RERAM to 22nm

Now turning to DorsaVi’s announcement on 5 December 2025, the company is progressing its Resistive Random Access Memory technology at Artemis Labs to fine-tune the circuit design so it can operate at a 22-nanometre scale. This is a semiconductor manufacturing node that is not leading edge, i.e. not below 10nm, but it is a node that is still used in massive amounts of modern chips. The “22nm node” simply refers to how small the electronic features are on a chip. The smaller the features, the more memory and performance you can pack into each device.

However, shrinking chips comes with challenges. As you scale down, issues like data leakage and inefficiencies can occur. So, DorsaVi needs to ensure the technology can operate effectively at these smaller dimensions without creating inefficiencies that would hurt the yield and reliability of the chips.

DorsaVi is able to do this by optimising the material stack, essentially identifying the best combination of materials and the right resistance levels within the memory layers to make the RERAM function properly at 22nm. So far, the team has successfully demonstrated clear on and off states, meaning the chip can reliably distinguish between a “1” and a “0”, the most fundamental requirement for storing data without constant power.

They have also confirmed that the memory retains its low-power switching behaviour, which is ideal for wearables and robotics, and Edge AI devices in general, where energy efficiency is everything. Lower energy intensity translates directly to longer battery life, more responsive devices and the ability to process data closer to where it is captured.

Promising Progress, Complex Path

It’s important to note that in the announcement, DorsaVi didn’t specifically state that this progress has been achieved with the newly acquired PIM IP yet. However, we can reasonably assume there are clear links between the fine-tuning work and the broader goal of ensuring the PIM technology can operate effectively at smaller scales.

For investors in DorsaVi, this is a very positive signal. The company’s technology continues to hit meaningful milestones, but investors should also recognise that this level of development requires exceptional engineering capabilities and involves highly complex work.

Even if the technology performs well at a lab level, moving from proof-of-concept to large-scale semiconductor production takes time. The company will still need to secure foundry partnerships, access manufacturing capacity and build commercial customer relationships to bring this technology to market. But it is definitely on the right track!

DorsaVi is a research client of Pitt Street Research whose directors own shares in the stock.