Newer nonvolatile memory technologies are poised to take over from flash in embedded applications on advanced process nodes, with MRAM and RRAM emerging as the frontrunners while phase-change RAM falls behind.
Magnetic RAM (MRAM) offers higher speed and better endurance for demanding use cases such as persistent working memory, while resistive RAM (RRAM) targets a broader range of systems based on its favorable cost profile. Both technologies require fewer additional manufacturing steps than embedded flash on finFET nodes, where flash has become increasingly difficult to integrate below 28nm.
“MRAM is frequently positioned for higher-speed, higher-endurance embedded use cases,” said Suhail Zain, vice president of regional marketing at UMC. RRAM, by contrast, targets firmware storage in IoT microcontrollers and configuration memory in power management ICs.
Phase-change RAM (PCRAM), once a promising contender, has stalled. No commercial offerings have appeared on finFET nodes, and no major foundry or IDM is actively porting it. Synopsys confirmed it would build PCRAM compilers if customer demand materialized, but “there’s not been enough demand,” said Daryl Seitzer, principal product manager at Synopsys.
The shift is partly driven by flash supply constraints. With HBM, DRAM, and NAND all consuming massive wafer capacity for AI demand, traditional flash suppliers are reducing availability for embedded markets.
Ferroelectric RAM (FeRAM) is also seeing renewed interest, and at least one company is developing a completely new NVM bit cell that could eventually challenge DRAM. Neither technology is expected to threaten standalone NAND, which maintains an insurmountable cost advantage for bulk storage.
