Summary
Resistive random-access memory (ReRAM or RRAM) is a type of non-volatile (NV) random-access (RAM) computer memory that works by changing the resistance across a dielectric solid-state material, often referred to as a memristor. This is in contrast to traditional forms of memory like flash, DRAM and SRAM which store data as an electric charge which also means they cannot store information without power. By instead storing information as a change in resistance, ReRam offers the potential for non-volatile fast memory. There are two main types of ReRam: oxygen displacement memory (OxRAM) and conductive bridging memory (CBRAM). ReRam is one of many next-generation memory technologies under development which also include phase-change memory (PRAM), ferroelectric RAM (FRAM), magnetic RAM (MRAM), and spin-transfer torque magnetic random-access memory (STT-RAM or STT-MRAM).
Viability (4)
RRAM has been around since the early 2000s and touted as the next step in non-volatile memory (NVM) to replace NAND flash memory, DRAM and SRAM. It gained notoriety in 2008, when HP hyped one type of ReRAM, the memristor, as part of a futuristic system called “The Machine”, which has since been discontinued. However adoption has been slower than expected mainly because of the difficulties in manufacturing ReRam were underestimated just as the scalability of NAND flash exceeded expectations. As NAND flash improved, the relative cost and performance benefits of ReRAM have not been enough for companies to proceed with the replacement. Non-volatile memory market was $57 billion in 2021 growing at 10% to $130 billion in 2030. It’s hard to find a catalyst that will push RRAM into the 2020-2025 timeline. The market RRAM is selling into is plenty large enough, but improvements in flash and relatively high costs continue to limit adoption.
Drivers (4)
The push for RRAM and next-gen memory technologies is by the fact incumbent memory technologies - DRAM, Flash and SRAM - are becoming more difficult to scale. In much the same way that CMOS-based technologies are getting harder and more expensive to scale, traditional memory is reaching a scaling limit after which new memory technologies are needed to increase performance. Specifically for RRAM, interest comes from applications that want to hold data in RAM rather than in storage and is an enabler of in-memory computing, Computational Storage and Neuromorphic Computing.
Novelty (1)
The amount of acronyms is a decent proxy for the competitive nature of the next generation memory space. RRAM competes with incumbent NVMs like flash, DRAM and SRAM as well as a host of emerging NVMs like PRAM, FRAM, MRAM and STT-RAM. All of these technologies compete on density, speed, power and most importantly cost. When compared against other next-gen memory technologies, all new technologies have greater densities, faster read and write times, lower power consumption and better scalability. The key thing will be how quickly any of these technologies can make it down the cost curve and begin to replace flash.
Diffusion (3)
Every major memory maker has some research into RRAM ready to replace DRAM and flash when it runs out of road. There hasn’t been any catalytic event that materially changes the fact RRAM has been around for two decades and has yet to see much adoption. Vendors continue to talk up the potential and reposition RRAM for low-power application areas. IoT and embedded systems look to be the most relevant markets but cost relative to flash and DRAM will continue to be a barrier. There will be a tipping point where power consumption costs become a barrier to new products at which point RRAM might be more expensive but makes a product viable.
Impact (3) Medium certainty
Flash, DRAM and SRAM will be replaced in time with higher-performance memory of which RRAM will be one of a few memory technologies. Next-gen memory technologies will differentiate across one dimension and carve out a niche, with MRAM and STT-RAM for low-power IoT type applications and RRAM for in-memory and Computational Storage type applications. The high impact scenario is one in which RRAM is an enabler and catalyst of new computing architectures that have previously been impossible with flash and DRAM like in-memory computing, analog computing and Neuromorphic Computing. This makes the impact much large than capturing a chunk of the $130 billion NVM market in 2030.
Relevant companies