It’s that time of year again! However, this year a little differently…. Why? Well on revisiting last year’s edition of this post, I was struck how the only successful ‘predictions’ were of the negative variety (suggestion that something that wouldn’t happen) whereas most (all?) of those that were ‘positive’ (suggested something that would happen) were mostly wide of the mark. I didn’t even suggest that 3D-NAND chips would actually appear in the market! And upon reflection, my initial thoughts were a number of predictions of what I don’t see happening in 2015….
Looking back of 2014, most of the ReRAM news from smaller companies has been dominated by two: Adesto and Crossbar. What is interesting is that the two companies appear to be taking opposite approaches to technology introduction and the future of their company (and technology). Adesto seem determined to make it on their own. They purchased part of Atmel which gave them access to a apparently competing technology but with a customer base and presumably cash flow. The customer base (small density flash with a focus on low power) is exactly where one would see their CBRAM fitting in. They have introduced some new niche products such as demonstrating that their technology is sterilization tolerant (maintains the stored data whilst undergoing irradiation to sterilize the part) allowing use in a number of medical applications without the additional complexity of reprogramming after sterilization. Note that this is different from being ‘rad hard’, i.e. being able to operate in a hostile radiation environment such as space or near a reactor. For that you have to rad hard circuitry as well as rad hard memory elements.
Adesto now have a VP of Sales and Chairman of the Board of Directors as per a recent announcement . Interestingly, these both appear to be new positions, i.e. the Board has apparently not had a Chairman until now. Not announced was that the CTO has moved on hopefully to pastures new. As a result, Adesto’s new-look leadership (with large, more relaxed photos!) does seem to have shifted towards the importance of revenue.
Crossbar, publically at least, are focused on (presumably) stand alone memory with their 1Tb on a postage stamp. What’s more they claim to have solved all the major technology challenges with just the transition to volume manufacturing before products will appear. Of course this is quite a hurdle to overcome for numerous reasons as the trials and tribulations of the SK hynix/HP partnership to productize resistive memory clearly demonstrate. It’s clear (to me at least), Crossbar cannot possibly perform the (hopefully, small)R & (large)D needed to develop volume manufacturing on their own. For a start they need a Fab, and a state of the art one at that. Those that exist either belong to the Foundries (TSMC, SMIC, GF, Samsung etc) or the remaining IDMs (Intel, Micron, Toshiba, Samsung etc). The Foundries are great at what they do (provide volume manufacturing capability for the fables community) but not so great at developing radically different technology with new processes and new materials simply because it takes away resources (wafers, people, equipment, etc) from their main purpose. The IDMs are much better in these terms but almost without exception have competitive technology of their own.
So Crossbar need to form a partnership with an entity that can provide them with the development capability (wafer runs, and lots of them) they will need to demonstrate volume manufacturing. I’m sure they are exploring every avenue to do so but even if they just require a BEOL process with new materials, it still has to be in a Fab with a BEOL process at leading edge design rules. Given the problems that these types of partnerships inevitably run into, especially when they are partnerships between unequal entities, frankly, I can’t see that happening either! (But I hope I am wrong….)
Rather, I suspect the endgame for Crossbar is that they will be purchased by one of the remaining volume manufacturers, perhaps following a nominal partnership phase where wafer runs will be promised to assess the state of the technology.
Developing an embedded technology (integration of new memory into an existing CMOS process) is being pursued by both companies. For example, Adesto with their partner Altis offer a number of options for integrating CBRAM with CMOS for such applications. Crossbar has made several announcements about developing a similar capability but I haven’t seen the same level of detail to date. Clearly critical for both companies, it would appear to fit in rather better with Adesto’s approach to productization (low density parts at mature technology nodes). Hopefully, 2015 will see an announcement of an embedded next generation memory product to follow Panasonic’s ReRAM containing MCU.
I wish you all a Happy and Prosperous New Year and thank you for your support of the ReRAM Forum
Christie Marrian, Moderator