January 31, 2022
Samsung is the market leader in both DRAM and NAND and their memory divisions are just part of a massive conglomerate. I look through their earnings presentation and conference call transcripts for insight into the state of the memory market. This is not an analysis of their earnings, prospects, etc.
Earnings Presentation
- As the company projected, overall memory bit growth was above market in Q4, following a below-market Q3. Prices continued to fall (they didn’t distinguish between DRAM and NAND in this comment, so it means both memory types.)
- Plan is to expand sales in leading-edge nodes (to reduce costs) and shift sales to the highest-density products. I read this to mean they are not adjusting production and instead will reduce their costs as fast as possible to minimize losses in this downturn.
- Only comments on when the market may recover was to say “possible recovery on mobile demand in the 2nd half [of 2023] with high-density trend” and that customers will continue lowering inventory in the short-term.
Earnings Call
These are the highlight quotes regarding DRAM and NAND:
… memory price fell further as deepening macro issues eroded customer – consumer incentive. Additionally, with the impact of meaningful loss from the valuation of inventory, our results decreased significantly compared to the previous quarter. This is surprising, that Samsung took an inventory write-down this quarter. Their cost structure is lower than Micron’s so this may be the result of a difference in accounting practices in Korea (higher cost used for inventory).
In DRAM, for server, as you know, is that component supply issues gradually eased. However, demand growth was limited because of our decline in Set build caused by economic uncertainties and customer stance to maintain their inventory adjustments. DRAM components are less oversupplied than they were ninety days ago.
Mobile and PC, it seems that our demand was muted because of continued inventory adjustment at major customers and a reduction in Set builds, amidst shrink consumer sentiment coming from concerns of economic slowdown. End demand for consumer products continues to be weak so PC and mobile phone makers are working down inventory instead of buying more memory.
While decline in demand was steepening across the overall industry, we expand the portion of cutting-edge node by optimizing our product portfolio. In addition, we actively responded to demand for high-density products focusing on major data centers and several OEM customers and our quarter-on-quarter bit growth exceeded market. Samsung is shifting more quickly to their most advanced process nodes to reduce their costs. They plan to outgrow the market. This is a contrary approach to what Micron is doing. Micron is reducing output and slowing node migration, taking a large hit to their cost structure in order to bring supply and demand into balance more quickly. Samsung’s strategy will grow their market share and lower their costs faster than Micron and Hynix. The company is flexing their superior size and leading cost structure to increase their market power over Hynix and Micron.
Next, I will talk about the NAND market. For server SSD, despite continued growth in our contents per box, customers continue to delay purchases because of their inventory adjustments and demand was somewhat stagnant. As for mobile, while consumer sentiment remained weak due to the economic slowdown, demand was sluggish, especially in China because of Zero-COVID policy until early December and production disruption at some customers. End customer demand for NAND is weak so consumers of NAND are using up inventory instead of buying more memory from Samsung.
Now let’s move to the outlook for the first quarter. For DRAM, in server, as economic uncertainties continue, our customers have maintained their stance on inventory adjustment so far. Thus, under concerns over weakening momentum for a demand recovery in the short term, we should monitor if demand will improve in accordance with macro variables such as interest rate policies. And for mobile and PC, under the effect of slow seasonality and weakened consumer sentiment due to concerns of deepening economic slowdown down, there is a possibility that demand will shrink. The pain in DRAM will continue, and possibly get worse if the global economy slows further.
Now let’s move on to the outlook for 2023. While the industry’s overall inventory level has increased due to the lowest ever demand bit growth last year, our customers are likely to maintain their stance on inventory adjustment in the near term. We expect customer purchasing sentiment to recover after inventory of those months are complete. First of all, for server, for the time being, our customers are expected to keep adjusting inventories due to economic uncertainties. However, we expect fundamental demand to remain solid given the investment in core infrastructure such as for AI and machine learning. In particular, demand for DDR5 is projected to increase in [earnings] in the second half of the year, thanks to increasing contents per box and the launch of new CPUs. For mobile, consumer sentiment is likely to keep muted due to various macro issues. However, demand may recover relatively in the second half of the year, thanks to the high-density trends and the spread of new competitors in smartphones. For PC, similar to mobile, the surge in sales induced by COVID-19 is likely to cause the replacement cycle to take some time, and shipment are focused to weaken for the time being. Underlying memory demand for data centers is strong but PC and mobile are weak. All segments have too much customer inventory, thus demand for memory from suppliers is low. 2022 had the lowest ever memory bit demand growth. This is why the memory downturn has been historically severe; memory makers couldn’t possibly have seen such a large decline in bit demand.
For DRAM, in the fourth quarter, our bit growth increased by a percentage in the high single digits and ASP declined by a percentage in the early 30% range. For the first quarter of 2023, we expect market bit growth to decrease by a low single-digit percentage and our bit growth should be similar. DRAM bit demand in the first quarter will shrink, a rare occurrence. Weak markets in memory almost always mean low bit growth, not absolute reduction in bit demand. Samsung will match the market, meaning they are planning, at least for this quarter, to hold inventory instead of selling into such a soft market.
For NAND, in the last quarter, our bit growth increased by a percentage just into the double digits. Our ASP fell by a percentage in the high 20%. In the current quarter, we expect market bit growth to decrease by mid-single digit, and our bit growth should slightly outperform the market. A high 20% ASP drop in one quarter for NAND is severe. Like DRAM, bit demand in NAND will shrink in absolute terms. Samsung will shrink their bit growth also, but not by quite as much as the market. This means they plan to hold inventory in NAND also.
In response to the question about Samsung reducing memory supply to help bring the market into balance, … and our CapEx approach this year is to continue to make the infrastructure investments that are necessary to respond to mid- to long-term demand and to make the – and to secure the essential clean rooms that we would need to do that. So in conclusion, this year’s CapEx plan is expected to be similar to previous year.
Now at the same time, in order to secure best quality and also line operation optimization. We have strengthened our production line maintenance efforts and are going through some equipment layout adjustments, which and also we are efficiently pursuing the migration to future cutting-edge nodes. Also, in order to increase our process technology competitiveness and also to stabilize our process technology early on, we have been increasing the share of engineering runs and as a result of that, within our total CapEx, the R&D-related portion is expected to increase versus previous years.
And so in the process of carrying out these initiatives, we think that impact on our bit at a meaningful scale in the near term would be inevitable. However, if you look at the long term, these are all essential activities that are necessary to enhance our competitiveness and responsiveness to the market. And that’s why we will be carrying these activities out at a brisk pace as part of our preparations for future growth. Capital expenditures in the year will be similar to 2022 with more of the money spent on cleanroom and facilities. They will take actions that indirectly lower bit supply growth, making equipment layout adjustments in fabs and running more non-product lots. These are activities fabs always need to do but they lower output. In this weak environment, they are deciding to do some of these activities because the output hit is positive for market balance.
Summary
This was not a good call for Micron investors in both the short and long term. Samsung plans to use their superior size and cost structure to put downward pressure on memory pricing. They are doing this by shifting their production mix to leading nodes faster. This increases supply of memory and lowers their cost per bit. Micron has already given in to the weak market and lowered production by 20% as well as slashing their capital expenditures. By not following suit, Samsung will increase their market share in both the short term and long term over Micron and also grow their lead in cost per bit. During down cycles in the 2000s and early 2010s, Samsung took similar actions to drive the Taiwan DRAM makers out of business. They have not exercised this market power in the last two down cycles, so this is a change in strategic behavior. Micron overtook Samsung in recent years in DRAM process technology and this action may be in response to Micron becoming more competitive. Being the third-place player, Micron can’t afford to cede ground to Samsung as they have with their decision to cut production. This is a major strategic mistake. Micron’s executive team is highly focused on the price of their stock and near-term results, which is why, I believe, they took the unprecedented step of reducing their production. This cycle will go down as a major strategic misstep by a Micron management team that are short-term thinkers. I continue to hold my large cyclical position in Micron but plan to sell all of it when they market turns up and the price of the stock recovered. I don’t believe this is a buy-and-hold stock. The only investment here is to invest in the cycles.
-S. Hughes (cyclical long MU)