There’s a “soap opera” unfolding in the memory industry, according to Ed Doller, Micron’s former Chief Strategist of the NAND Solutions Group. A year and a half ago, memory manufacturer Kioxia (a spin-off company from Toshiba) was expected to go public. After rumors of more private equity money invested in the company and a possible acquisition by Western Digital or Micron, Kioxia is expected to IPO in September, according to Reuters.
To get his thoughts on Kioxia and the state of the rest of the memory industry, Blair Kenney of GLG’s TMT team spoke with Doller. Below are a few select excerpts from our broader discussion.
If Western Digital acquires Kioxia’s assets, how would that help it?
Strategically speaking, in the NAND industry, like the DRAM industry, further consolidation would likely be a good thing. WD’s NAND market segment share in Q4 of 2020 was around 15%. Micron was at 11% and SK Hynix at 12%. The leaders are Kioxia and Samsung. SK Hynix is acquiring Intel’s NAND business, whose market share is around 9%, so it puts the company at more than 20% market segment share, a nice place to be. If WD were to acquire Kioxia, it’d be a solid number two player in the NAND industry.
Furthermore, the position that Kioxia has in the enterprise SSD market, while certainly nowhere near as strong as Samsung or Intel, is definitely better than Micron and WD. Micron’s been struggling to further its penetration into that market, so having access to that market would help it. It echoes the SK Hynix acquisition of Intel’s NAND memory unit. The thing that surprises me most is that Micron didn’t acquire the Intel business, given the relationship those two companies had over the past decade and a half.
We’ve seen a lot of NAND and DRAM capital expenses build-outs from SK Hynix and Samsung. How will this impact the industry’s cost curve and pricing?
Anyone who’s paid attention to this industry knows that as it grapples with Moore’s Law and the inability to keep up with it, what that fundamentally means is that keeping technology scaling will require more and more capital. Everybody has their own way of doing it. There’s no wrong and right way — it’s just a matter of where do you want the pain to occur? As much as you would like to think that capital will ultimately pass through to pricing, that doesn’t occur in this industry. This is a commodity business, so to speak. You do your best, hope for the best, try to manage your supply demand such that you’re not creating too much, which I don’t think will occur. We’ll just have to watch and see how this all plays out.
Micron is focusing the strength of its long-term demand on its technology transition. Does Micron’s current architecture and technology upgrades give it enough runway relative to competitors, specifically in the enterprise SSD and NAND space?
First, it’s important for people to understand a couple of things about the solid-state drive market. There’s the client SSD market, which for all intents and purposes is a commodity solution, and there’s the enterprise SSD market, which is a great market to service. The bit demand is high and, generally speaking, the margin dollars are high. This market is all about reliability and performance. If firms can build an SSD that is more reliable and has better performance, the customers will want it. They’ll likely pay a premium. But within SSD, there is a controller that manages the NAND, and building a controller is very difficult.
That’s why companies such as Samsung and Intel, who were first movers to go after the enterprise space, have done so well and continue to garner the lion’s share of that market. Micron frankly has been struggling to penetrate deeply into the enterprise market. The margin dollars are on the high-performance, high-reliability PCIe side. Micron is only doing that now with an acquisition it made several years ago. Intel and Samsung will do everything they can to keep Micron from penetrating far into that market.
All the while, Intel has had only a small amount of NAND capacity. Once the acquisition of SK Hynix closes, it’ll have access to a whole lot more NAND, which means Intel’s ability to penetrate deeper into the market is greater. It’ll further keep Micron at bay.
Could you give a range of possible NAND demand outcomes, given SSD and controller shortages through 2021?
What’s happening is there’s a constraint across the board on logic components. That primarily has to do with the consolidation of logic foundries. Companies have not been investing properly. Now, it’s becoming problematic for companies to get logic, specifically solid-state drive controllers, both on the client side and on the enterprise side.
What do you do about that? If a company is a NAND manufacturer and sells to the enterprise, it’ll try to get customers to buy higher-density, solid-state drives. For enterprise players such as Intel, they might have excess NAND, but they won’t lower average selling price (ASP). They’ll remain high, but the volume may come down, which leads to excess NAND. The NAND market is different from the DRAM market from an elasticity standpoint. You can always move NAND by lowering the ASP, since you can never have enough storage, but more DRAM is unlikely to change the performance of a laptop.
There certainly is a scenario where, over the next year, we start to see NAND’s ASP coming down as a result of the inability to ship a solution. My expectation is that it’ll be a tight balance in the industry as we watch logic providers expand their capacity, but this is a multiyear problem. Hopefully, it’ll work itself out, but it wouldn’t shock me if this has a foundational impact on the NAND market over the next 12 months.
What do you see as the key areas of growth over the rest of 2021 for the memory market?
The key areas for growth are these transitions that will occur from a technology standpoint. Keep in mind that when companies come out and say they’re introducing the 177 layer or 144 layer, these transitions take time. For example, an enterprise solid-state drive firm not only has to get the fundamental NAND technology healthy, but integrate that healthy technology into a new SSC, and get its cloud service provider to qualify that SSC. It always takes longer than you’d think. We’ll continue to see these newer multi-hundred-tier-level NAND devices be integrated into SSEs and ramped into a high volume, especially in the second half of this year. On the DRAM side, as we see companies such as Micron and its competitors with their 1-Alpha, they’ll continue to drive lower costs because that’s what the memory industry does.
Is there a risk of overcapacity going into the second half of 2022?
Always. In this industry, it’s a struggle. But I remain bullish on the memory industry. It makes me sleep well at night knowing that demand as well as the challenges to continue scaling this technology will be extremely strong. There’s probably never been a better time to be in the semiconductor memory business. All you have to do is look at the historical gross-margin dollars of DRAM and NAND. Even when people say it was a horrible quarter, they still got 30% gross margin. If we can continue to have the worst quarters that we’ve recently had, it’ll be a phenomenal industry to participate in.
About Ed Doller
Ed Doller is a storage industry expert, veteran, and independent consultant with more than 30 years of experience in the storage space. He was most recently Vice President and Chief Strategist at Micron Technology, which offers the industry’s broadest portfolio of silicon-to-semiconductor solutions. Prior to this, he served with Micron as Vice President and General Manager Enterprise Storage and served as Vice President and Chief Memory Systems Architect with the company. He joined Micron in May 2010. Prior to Micron, Dollar served as Chief Technology Officer at Numonyx and at the Flash memory group at Intel.
This technology industry article is adapted from a GLG teleconference. If you would like access to events like this or would like to speak with technology industry experts like Ed Doller or any of our more than 900,000 industry experts, contact us.
Sample Questions from Teleconference:
- As a potential deal with Kioxia heats up, what do you think is the likely outcome?
- If Western Digital acquires KIOXIA assets, how would that help them achieve a subscale in the NAND market?
- With cap-ex build-outs from SK Hynix and Samsung, how is this likely to impact the industry’s cost curve and then pricing in the next three to five years?
- Would you quantify what you think Micron could lose in market segment share?
- Micron is focusing the strength of its long-term demand on its technology transition. So, does Micron’s current architecture and technology upgrades give it enough runway relative to competitors?
- Is Micron’s architecture still enough to keep them competitively profitable?
- So, is the margin uplift from 2D to 3D memory condition largely behind the industry?
- How will supply-and-demand trends vary by DRAM types, and that would be DDR3, DDR4, and DDR5?
- How will the impact of the new Intel server chips Ice Lake and others impact the DRAM consumption and DRAM requirements per server?
- Could you give a range of possible NAND demand outcomes, given SSD and controller shortages through 2021?
- What is your outlook for the next-gen memory space after backgrounds quickly pivot from 3D XPoint to CXL?
- Does it change your opinion on next-gen memory in general and its near-term adoption rates if we see a strong pivot to CXL?
- To the 2022 to 2025 time frame, who do you think is most likely to win, and can you choose the company, or can you break this out individually, by wafer cost, bit cost, and density?
- So what do you see as the key areas of growth over the rest of 2021 for the memory market?
- Any risk of overcapacity going into the second half of 2022?