4DS Memory (ASX: 4DS): A Second Life at 60nm in India

We’ve been following 4DS Memory for a very long time, and for most of that journey it has been a story of heartbreak. The company’s dream, bringing its proprietary Interface Switching ReRAM technology to sub-20nm advanced process nodes, ran aground not once, but twice. After years of platform lots, etching problems, manufacturing setbacks at imec and, ultimately, the formal disengagement from its development collaborations with imec and Infineon in September 2025, we wrote what felt like an obituary. We said then that we hoped 4DS Memory could find a future in some form, even if it was not the one shareholders had anticipated. On 12 May 2026, the company announced something that suggests it may have done exactly that and the answer lies not in Belgium or Silicon Valley, but in India.

What went wrong for 4DS Memory at 20nm … twice

To understand why today’s announcement a little bit better, you need to understand what killed 4DS’s original ambitions. The company’s technology, based on a PCMO (Pr0.7Ca0.3MnO3) material, was developed specifically for large-scale Storage Class Memory applications. The vision was compelling: Interface Switching ReRAM that used the full surface area of each memory cell, offering superior endurance, data retention and, crucially, the ability to scale to the advanced nodes where Flash memory runs into physical and economic limits. Flash, after all, cannot reliably scale below 40nm. ReRAM, can go much further.

The problem was getting there. 4DS Memory needed to scale its technology down from 60nm, the node at which it had already demonstrated successful PCMO-based ReRAM operation with its development partner imec, to somewhere below 20nm to be commercially competitive in storage-class applications. That scaling work proved brutally difficult. The etch process repeatedly caused residue contamination of memory cells. Yields were poor. Lots came back from Belgium not performing as hoped. By the time the sixth platform lot reported in June 2025, the results were, to put it gently, not what 4DS Memory needed. The strategic review followed. The imec and Infineon collaborations ended. The shares lost two-thirds of their value in a single day.

The 20nm dream was over. But the 60nm reality had always been there, i.e. proven, de-risked and, as it turns out, commercially interesting after all.

India rewrites the calculus

India has committed more than AUD$25 billion to building a sovereign semiconductor manufacturing industry. The critical detail for 4DS is that India’s fabrication capability is currently concentrated in mature-node technologies, exactly the 60nm to 65nm range at which 4DS’s PCMO-based ReRAM platform has already been demonstrated to work. This is not a coincidence and today’s announcement is a strategic realignment that flips the script entirely.

At 60nm, Flash memory is indeed the dominant embedded non-volatile memory technology, and no one is pretending otherwise. But ReRAM at this node offers real, differentiated advantages that Flash cannot always match. ReRAM writes data by toggling resistance states rather than through the high-voltage charge-trapping mechanism that Flash relies on. The result is meaningfully lower power consumption per bit operation, a critical advantage in battery-powered or power-constrained environments, as well as significantly faster write speeds and much higher endurance in terms of program/erase cycles. Flash, particularly NOR Flash at embedded applications, can struggle with endurance beyond one hundred thousand cycles. ReRAM can reach into the millions. For applications where data is written frequently and reliably, that is not a trivial distinction.

Where 60nm ReRAM actually wins

So, which end products benefit from these characteristics? Quite a few, as it turns out, and 4DS Memory has named several of the most compelling verticals in today’s announcement.

Automotive electronics are a natural fit. Modern vehicles carry dozens of microcontrollers managing everything from engine management units and transmission controllers to ADAS sensor fusion modules. These systems must write configuration data and fault logs repeatedly over a vehicle’s lifetime, often in high-temperature environments where Flash endurance degrades faster. ReRAM’s superior endurance and lower operating voltage make it an attractive embedded memory option here.

Industrial control systems tell a similar story. Programmable logic controllers, motor drives and industrial IoT nodes are frequently updated in the field, placed in electrically noisy environments and expected to operate for decades. The reliability and write-cycle endurance of ReRAM over embedded Flash is a genuine selling point, not a marketing slide.

Defence and aerospace applications push these requirements even further. Avionics computers, guidance systems and secure communications hardware demand memory that is radiation-tolerant, highly reliable under thermal stress and capable of withstanding extended read-disturb conditions. ReRAM’s resistance-switching mechanism has demonstrated better radiation hardness characteristics than floating-gate Flash — a property that has attracted serious interest from defence electronics designers globally.

Edge AI inference hardware represents perhaps the most forward-looking opportunity. As AI processing moves to the network edge, i.e. into smart cameras, industrial vision systems and autonomous agricultural equipment, the need for fast, low-power, non-volatile weight storage becomes acute. At 60nm, a ReRAM-based embedded memory block can store neural network weights with fewer read latency penalties and lower active power than NOR Flash alternatives, enabling more capable inference at the same battery budget.

A second life — if not quite the original dream

None of this is to say that 4DS Memory has suddenly become the company it set out to be a decade ago. The Storage Class Memory vision that is competitive with NAND Flash at advanced nodes, disrupting data centres and mobile storage is gone. That chapter closed with the imec collaborations. What 4DS is offering today is something more modest, but arguably more achievable: a proven 60nm PCMO ReRAM platform deployed into India’s emerging semiconductor ecosystem through licensing, technology transfer, embedded memory partnerships and OEM integration pathways.

Long-suffering shareholders deserve a realistic assessment. The 60nm opportunity is real, but it is not the jackpot that was once imagined. Licensing revenues and technology transfer fees are the likely revenue model in the medium term, not the royalties from billions of advanced-node chips, even though royalties on commercially produced chips will be what 4DS Memory is aiming for longer term.

But for a company that appeared to be heading for the exit in September 2025, finding a commercially credible second life for technology that works, at a node that India is actively building out, is a far better outcome than the alternative, a share price that goes to zero. Sometimes the road not taken turns out, unexpectedly, to be the one worth travelling.