Exclusive Interview With Jim Handy, Objective Analysis
"Every 2 years number of flash bit cells on single chip doubles."
By Jean Jacques Maleval | November 13, 2020 at 2:11 pmAnalyst Jim Handy came to Objective Analysis with over 30 years in the electronics industry including 14 years as an industry analyst for Dataquest (now Gartner) and Semico Research. His background includes marketing and design positions at suppliers including Intel, National Semiconductor and Infineon. He writes two blogs: The SSD Guy and The Memory Guy. He has a Bachelor’s degree in Electrical Engineering from Georgia Tech, and is a patent holder in the field of cache memory design. He is the author of The Cache Memory Book (Elsevier, 1993). He also holds an MBA degree from the University of Phoenix. He is on the advisory board of the FMS, has served as a member of the Embedded System Conference’s Embedded Leader Community, a member of the Mass Storage Technical Working Group of the International Electronics Manufacturing Initiative (iNEMI), the past chair of the iNEMI photovoltaics Product Emulator Group, a member of the SNIA Solid State Storage Initiative, and is a leader in the Gerson Lehrman Group Councils of Advisors, as well as a member of other expert networks, including AlphaSights, Coleman Research, DE Shaw, DeMatteo Monness, Elliott Advisors, Guidepoint Global, Ridgetop Research, and Tegus Advisors. He also serves as an expert witness and has been designated an Elite Expert by IMS ExpertServices.
When will HDDs definitely die under the pressure from SSDs?
A couple of years ago, WDC, a company that wins whether or not SSDs take over the HDD market, showed a timeline in which they projected that NAND bytes would still cost 10x as much as HDD bytes through 2028. As long as HDD stays cheaper than SSD then there will continue to be an HDD market. Although NAND flash is still undergoing aggressive cost reductions, HDD is still keeping pace, and it looks like that will continue for a long time. I don’t expect to see HDDs die off for over a decade. Of course, I said this same thing in 2006, and it’s now 14 years later and I am saying it again. HDDs may be around for a very long time.
How do you see the evolution of SSD and NAND flash prices?
Most SSDs are sold at cost-based prices, and 80%+ of the cost is the NAND flash, so SSD prices tend to follow flash prices. NAND flash continues to follow Moore’s Law even though other technologies have slowed. Every 2 years the number of flash bit cells on a single chip doubles. This causes manufacturing cost per bit, and therefore price per bit to drop by 50% every 2 years. Shifts from MLC to QLC, and later to PLC, accelerate the cost reduction beyond that. It seems that there’s no limit to the number of layers that can be added to 3D NAND, so today it seems like the cost reductions will continue forever. While that’s not practical, I expect to see it continue for well over 5 years.
We counted around 200 makers of SSDs in the world. How do you see consolidation? Will the flash chip makers finally be the winners?
There’s something to be said for vertical integration, since it allows one company to keep the profits that would normally go to two. But if that were the only consideration then there wouldn’t still be dozens of DRAM DIMM makers after all this time. Memory modules first appeared in the late 1980s. Chip makers are best at satisfying huge orders. Smaller companies do better than the chip makers at satisfying lower-volume needs from end users. They will coexist as long as there is a market.
There will always be numerous SSD makers. The barriers to entry for this market are very low – anyone can get in. To stay in the market a company simply needs to manage its inventory turns very well and to constantly find new customers. Any company that does this well will be able to stay in the SSD business.
Areal density of flash chip seems to never decrease. What’s the limit and when?
Areal density is a measure of HDD capacity, but it doesn’t make sense to use it for memory chips. The best measure for NAND is cost per bit, and that continues to decrease very quickly. As I mentioned above, there is no clear limit to the number of 3D NAND layers that can be made. 3D NAND may be with us for a very very long time.
Who are the main flash chip makers and their market share?
In 2019 Samsung led the market with 35%, followed by Kioxia with 19%. WDC and Micron each had 14%. The bottom 2 major NAND makers were SK hynix with 10% and + Intel at 8%. Both were a little less than that, so their combined share would have been 17%. Some people believe that SK hynix combined with Intel will be the new #2 player after Samsung for 2020.
How will PC NAND undermine DRAM?
We wrote a report at Objective Analysis almost 10 years ago that showed how NAND flash could improve PC performance more economically than DRAM – if you add $50s’ worth of NAND it would improve performance more than adding $50 worth of DRAM. This was a secret that server users knew but that PC makers and users didn’t, so we decided to run an experiment and commissioned someone to run nearly 300 benchmarks. Had the market moved in this direction then PC DRAM sizes would have stagnated and SSD disk caches would have become very popular. DRAM makers would have suddenly seen the market’s growth permanently stall. In the end nobody ventured in that direction.
How much have 3D NAND pitches shrunk since the technology’s 2013 introduction?
3D NAND pitch has shrunk very little since the technology was first introduced. The biggest change came from staggering the columns instead of having them all in a line. This made a small but important improvement in die size. 3D NAND pitch doesn’t shrink because the holes are being made as small as possible, since they need to be coated on the inside with numerous layers of materials. This should not change much over the long term.
What’s China’s memory ambition?
We wrote a report about this. China’s government has a lot of money. The country is centrally managed, and operates to 5-year plans. Naturally, the government looks at all of this money, and they look at the problems that need to be solved and then they put together a 5-year plan to use the money to solve the problems. One problem today is that semiconductors are China’s biggest import, even bigger than oil! The country would like to become more independent and to source more of its chips internally. To accomplish that the government’s investment group Tsinghua assembled something called the “Big Fund” with over $100 billion to establish an important position in semiconductors.
What product should they focus on?
Memories are a commodity: users can replace one company’s chip with someone else’s and the system still works, so the buyers can rapidly shift from one supplier to another depending on who gives them the best price. China will use memory chips to rapidly establish an important presence in the semiconductor market. It’s the same thing that Japan, Korea, and Taiwan all did when they wanted to become important suppliers to the semiconductor market.
How do you explain that Seagate is a poor actor in SSDs?
You need to ask Seagate about the company’s focus, but there is clearly a much stronger focus on HDDs than on SSDs.