Article written by Pitt Street Research. For the full report click here
WBT recently extended its agreement with Leti, its French research collaboration partner, to facilitate the next step in commercializing its ReRAM memory technology. This extension includes adapting the production process of WBT’s ReRAM technology to 300mm wafers at the 28nm node.
Promising results from latest tests
Related to this commercialization process, WBT’s latest ReRAM cells demonstrated stable voltage and endurance levels. WBT and Leti have demonstrated array-level endurance, i.e. across an array of memory cells, above 100,000 cycles, which is on par with expectations in the storage memory market and significantly better than NAND Flash, today’s dominant non-volatile memory (NVM). Combined with earlier announced retention data, we believe the SiOx ReRAM investment case has been substantially derisked, i.e. transferring the technology to 300mm wafers should be relatively straightforward.
Partnerships with Silvaco to accelerate OEM incorporation
WBT also announced a Development Program Partnership with Silvaco (a leading global provider of software, IP and services for design of IC’s and electronic systems for the semiconductor industry) to model the electrical behavior of SiOx ReRAM devices. This collaboration should help accelerate the
incorporation of WBT’s ReRAM modules into OEM customers’ designs.
Using memory cells in logic functions would be groundbreaking
Additionally, in a new collaboration agreement with Technion, one of Israel’s leading Technology research institutes, WBT’s ReRAM technology will be tested in Real ProcessingIn-Memory (Real PIM) environments. Simply put, this means using the memory cells themselves to perform logic functions, i.e. structure them so they actually form logic gates.
Traditional chip architecture separates the logic functions from the storage functions, i.e. on different chips (with the exception of SRAM which is typically very close to the processor on the same chip). This distance takes time and energy to bridge and thus leads to efficiency loss. However, if memory modules could somehow process information instead of just storing it, data could be sent to the memory module for local processing before being sent back to the CPU for further processing. This would save valuable time and money.
Theoretically, due to the structure of a ReRAM cell, WBT’s technology could have an application in this field. The company’s new research collaboration with Technion is all about finding out if theory holds up in practice.
Collaboration with Politecnico di Milano to progress AI applications in hardware
In addition to the earlier announced collaboration with IIT Delhi, WBT has set up a collaboration with Politecnico di Milano, a leading Technical university in Italy, also to further develop ReRAM’s potential in the field of hardware-based Artificial Intelligence (AI). Because of the ReRAM cell structure, which resembles biological synapses, the technology can potentially be tweaked to serve AI applications in a way that will be much faster and cheaper (less energy consumption) than software-based neural networks.
Additionally, hardware-based neural networks are much faster to train on new data sets and can often even learn autonomously. BrainChip (ASX:BRN) is also developing a hardware-based AI chip and we believe this field has great potential given the many benefits compared to software-based neural networks.
There is broad future applicability of hardware-based neuromorphic chips in areas such as autonomous driving, advanced driver assistance (ADAS), edge computing in IoT devices etc.
A wide variety of application areas now being investigated
There are now four identified, potential application areas for WBT’s technology:
• Embedded memory in SoC (System-on-a-Chip) designs;
• Storage Class Memory (SCM);
• Artificial Intelligence chipsets;
• Real Processing-In-Memory (Real PIM).
Embedded memory for SoC’s is WBT’s most direct path to market, in our view, but SCM could be close behind if sufficient traction can be garnered from potential licensees. WBT is just starting out on the path to neuromorphic chips, unlike BRN, so that development process is likely to take several more years at least. Lastly, Real PIM is also a fairly new field with a lot of development work ahead.
Buy rating reiterated and A$ 0.07 price target unchanged
We believe that embedded memory will remain the key value driver for WBT for the foreseeable future. The company is taking the right steps in this commercialization process, in our view, with the move to 300mm wafers now well underway. We reiterate our Buy rating for WBT as well as our price target of A$ 0.07 per share.
Please note, the usual disclaimers apply – click here
Pitt Street Research work is commissioned by the listed companies it covers, and Pitt Street Research has received or will receive payment for the preparation of such work. Please refer to the bottom of the research notes as published on Pitt Street Research’s web site for risks related to the companies being covered, as well our General Advice Warning, disclaimer and full disclosures. Also, please be aware that the investment opinion in this report is current as at the date of publication but that the circumstances of the company may change over time, which may in turn affect our investment opinion.
Categories: Australian Stocks