This 40nm geometry, demonstrated by 4DS, is smaller than the latest generation of 3D Flash - the most dominant non-volatile memory technology used in billions of mobile devices, cloud servers and data centers.
In 2016, 4DS has
- Demonstrated scalability, consistency and behaviour of memory cells with high yield at 40nm
- 40nm is a breakthrough development at a scale smaller that existing 3D Flash, the most dominant non-volatile memory technology
- JDA with HGST renewed in July 2016 taking the collaboration into its third year
4DS Memories Ltd. (West Perth, Western Australia) claims to have achieved 40-nanometer resistive random-access memories (ReRAMs) that are denser than flash and rival the recently reported Crossbar Inc.'s (San Francisco)
4DS claims its 40-nanometer ReRAM is a first, but many other labs besides 4DS and Crossbar are known for serious ReRAM efforts using memristors including Adesto Technologies, Elpida, Fujitsu, Global Foundries, Hewlett Packard, Hynix, IBM, Macronix, Nanya, NEC, Panasonic, Rambus, SanDisk, Samsung, Sharp, Sony, ST Microelectronics, Winbond, and several research-only labs like Imec collaborating with foundry partners like TSMC.
Flash is reaching the end of its ability to scale linearly, prompting the move to 3D, such as Samsung's, Toshiba's and Western Digital's recent demonstrations of 64-layer stacked-die flash memories.
4DS also claims to have invested only $12 million to research and develop its recent demonstration chips. The demo chips, 4DS claims, prove its ReRAM memory cells are faster, cheaper and lower power than 3-D flash, giving the company hope at carving out a segment of the $40 billion global market for flash.
Beyond 2016: Commercial deal or strategic action most likely
From the current state of technological development, we essentially see three scenarios for 4DS going forward; firstly, a scenario in which the company succeeds in finalizing the minimum required development of its technology to the point where prospective licensees step up to the plate. In this scenario 4DS will need to demonstrate scalability of the Interface Switching ReRAM memory cell down to sub-45nm resolutions with consistent cell behavior using a scalable manufacturing process.
Still several years of development required by licensees
Any licensee will need to further develop the technology in the next several years to the point where high density, Interface Switching ReRAM memory chips can be manufactured in existing fabs, which we would expect around 2019-2020. This development process will require tens of millions of dollars, possibly up to US$100M, in our view, which is why 4DS will likely not embark on this journey, at least not by itself.
Any license agreement that is non-exclusive would see 4DS receiving multiple up-front, oneoff license payments, which could amount to several millions of dollars each. Additionally, 4DS would receive royalties per chip sold once the technology goes into commercial production. Memory chip royalties typically amount to a single digit percentage of the sales price of the chip.
An exclusive license to use 4DS’ technology would likely require the licensee to pay a substantially larger up-front license fee, potentially several tens of millions of dollars, in 4DS' view, in addition to royalties once the chip goes into commercial production.
Potential strategic action ahead of any license deal
The second scenario would be an acquisition by an established manufacturer in the data storage space. In this scenario as well, we believe 4DS would first need to complete the minimum required development of its technology in order to prove commercial viability of Interface Switching ReRAM. Any acquirer will then need to further develop the technology to commercial insertion into the market, similar to the first scenario.
Given 4DS’ development agreement with HGST and the acquisitive nature of its parent company, we believe Western Digital would be a very likely acquirer in this scenario.
Additionally, companies like SK Hynix, Micron and Samsung might have a keen interest in securing Interface Switching Memory technology to gradually take over from 3D NAND Flash in a few years’ time.
The third and final scenario would see 4DS’ technology not being rolled out commercially in due time, either due to insurmountable and/or overly expensive technological issues with Interface Switching ReRAM, or the emergence/dominance of another non-volatile memory technology, e.g. filamentary ReRAM, MRAM etc. While the likelihood of this scenario seems relatively small at this stage given the rapid technological progress being made together with HGST, it can never be excluded in the dynamic semiconductor industry.
There are two Approaches to ReRAM : Interface switching ReRAM and Filamentary ReRAM
Interface Switching ReRAM - high density memory for mobile and cloud
The development of Interface Switching ReRAM, a unique type of Non-Filamentary ReRAM, represents a breakthrough in ReRAM technology and is unique to 4DS.
Developing memory storage that is not reliant upon a filament allows cell currents to scale down in line with cell size enabling the smaller geometries necessary to put more storage on a memory chip creating high density memory.
A filament-less switching mechanism can operate with low switching currents, due to much more stable currents, essential for high density gigabyte range memories and the retention of data.
4DS has developed a way of controlling the overall resistance of the memory cells using the diffusion of oxygen atoms across the interface and this mechanism is used to reliably control gigabyte memory intended for large-scale storage.
Importantly, Interface Switching ReRAM does not rely on a destruction mechanism thereby increasing endurance, reliability and functional behaviour.
Filamentary ReRAM - low density memory for IoT and connected devices</>
The formation of filaments is the most common approach in ReRAM cell research and development today.
Filamentary mechanisms may work well at relatively large cell geometries but pose significant current density, retention, endurance, access and control problems when trying to achieve gigabyte range memories.
Filamentary ReRAM has inherent scaling limitations because cell currents are high and are independent of cell size. High switching currents are needed for long data retention and the large current fluctuations typically observed in filament-based ReRAM.
The potential for scalability to smaller geometries is limited by wire current densities.
Furthermore, the create and destruct switching mechanism in filamentary ReRAM results in eventual cell breakdown and poses a number of significant limitations for GB silicon storage.