Emerging Memories After Optane

Prism made of memory hierarchy splitting a beam of light into a color spectrum. Each color has the name of a new memory technology.Now that Intel is exiting the Optane market what will happen to the market for new memory technologies?  This is an interesting question that The Memory Guy has focused considerable attention over the past few years.  In a nutshell, the market will continue to develop, but at a slower pace, with the bulk of revenue growth going to memories embedded into SoCs.

Even so, the market will grow significantly, with revenues reaching a solid $44 billion by 2031.  This is explained in the update to the Objective Analysis and Coughlin Associates acclaimed annual report on emerging memory technologies: Emerging Memories Enter Next Phase.

This growth will be fueled as new memory technologies displace NOR flash for embedded program storage in microcontrollers, ASICs, and other SoCs, since NOR cannot be produced on a FinFET logic process.  Since logic migrated to FinFETs after the 28nm node, the opportunity is already open for MRAM, ReRAM, or another technology to replace its functionality.  This is indeed beginning to happen on a number of IoT chips as well as chips for wearable devices and other applications.

After establishing itself on SoCs, emerging memories will move on to replace not only NOR flash, but also SRAM, and DRAM on SoCs, and then will win designs on all advanced-process chips, even compute processors.  This will all come about as increasing production volume drives the costs out of these technologies, and they become cheaper alternatives to current memory types, even as discrete memory chips.

The report focuses on the technologies, the processes used to produce them, and the equipment requirements that this change will drive, with forecasts not only of emerging memory revenues and costs, but also of the incremental tool purchases that will be required to support this change.

Particular care has been devoted to explaining the challenges faced by discrete memory chips, which do not have the same pressing need to migrate away from today’s highest-running technologies.  A large section of the report details the Intel Optane story and why the technology was unable to reap profits for Intel despite growing into a $400 million annual business.

These losses are estimated in the chart below, one of the report’s 159 figures:

Column chart showing estimated quarterly losses ranging from $100 million to over $600 million from 1Q16 to 4Q20 The history of NAND flash is used to illustrate Intel’s issue.  The report explains the very significant role of the economies of scale, showing how Intel was unable to reach sufficient wafer volume to achieve production costs that were low enough to make Optane a profitable product.

This 241-page report is extremely thorough, with 159 figures, and 36 tables.  It examines not only PCM, ReRAM, FRAM and MRAM, but also a variety of less mainstream technologies, explaining each one’s competitive strengths and weaknesses.  Its 119 company profiles include chip makers, technology licensors, foundries, production tool makers, and research consortia, touching all aspects of the semiconductor business.

It’s the place to look for anyone who wants to understand the technologies, outlook, and business dynamics of emerging memory technologies.