Webb gave a series of presentations at the recent virtual Flash Memory Summit. While there are many flavours of emerging memory technology in development right now Webb's conclusion is that Intel's phase-change memory – known as 3D XPoint memory or Optane will dominate the market for stand-alone emerging non-volatile memory in 2025 and 2030 and this is partly because Intel is selling the memory at a loss.
Webb defines emerging memory as PCM, MRAM, ReRAM and FeRAM plus others. It does not include NAND, DRAM, NOR, SRAM, EEPROM and so on. In these analyses Webb does not consider embedded/SoC emerging memories where MRAM is starting to gain traction.
It is Webb's view that PCM, supplied by Intel, will represent 90 percent of the stand-alone emerging memory market in 2025 with MRAM in a minor second place and that will still be the case in 2030.
Year Optane MRAM All emerging memories
2020 $1.1B $80M $1.25B
2025 $4.3B $250M $4.75B
2020 $9.7B $610M $10.7B
Stand-alone emerging non-volatile memory markets through to 2030. Source: MKW Ventures.
And this is despite the fact that Intel sells its phase-change memory at a loss, according to Webb. In an FMS presentation Webb said that 3D Xpoint operating margins are "very negative" and estimated that selling Xpoint was costing Intel more than $300 million a quarter. Although that loss making may diminish as memory production scales it is set to persist, Webb said.
He added that it doesn't matter. Persistent Memory is great for the data center and differentiates the Intel Architecture from AMD and other competition. Margin on memory is not the goal for Intel, Webb said.
"We predict huge revenue increase for Optane simply because Intel is investing billions in it and is driving new bus connections for it. Without Intel ... divide the numbers by 5 to 10," Webb said in a blog post.
While Webb's analysis may seem bleak for proponents of alternative memories such as ReRAM and FeRAM he provides some justification. He points out that five years after the introduction of PCM by Intel and Micron together, Micron's sales, apart from those to Intel, are worth less than $10 million per year. Webb said it takes a long time to hit high volume after it is agreed that a technology works.
"If ReRAM or FeRAM take off. it will be in four to five years and it will not replace other markets," Webb observed.
Related links and articles: