April 27, 2023
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
- · Demand recovery in the second quarter of 2023 is unlikely because of limited investment by customers, especially hyperscalers, and continued inventory consumption.
- · Mobile and PC segments should reach healthy levels in the second quarter. If that is true, demand from those customers should increase in the third quarter of 2023 and beyond. If prices start to rise, I expect their pace of buying to increase as they seek to replenish inventory with lower-priced components. This is one feedback loop that fuels memory upturns.
- · Demand is expected to gradually recover in the second half of 2023 as customer inventories reach healthy levels.
Earnings Call
These are the highlight quotes regarding DRAM and NAND:
… Mobile and PC applications, although customer inventories of finished goods in distribution channel are relatively improved. The demand situation was sluggish with purchasing delays and set few deductions as consumer sentiment has not recovered yet. While inventories are reaching healthy levels for the segments that first entered the downturn, customers in those areas remain wary of increasing their rate of purchasing. This is keeping demand weak.
Next, I will talk about NAND market. The slowdown in demand for server and storage applications has been more clearly observed compared to the other applications. And as I previously mentioned for DRAM, our analysis revealed that the demand sluggishness is linked with reduced enterprise IT spending. Actually, in addition to this, server SSD demand showed a further contraction due to customer inventory adjustment. Server customers went into the downturn so they are further from the inflection point up. Also, cost concerns with many of those customers is lowering demand as their budgets are under scrutiny.
Moreover, as we announced in our preliminary first quarter result, we have already started rolling down our productions. Therefore, our inventory levels are expected to start to decline from the second quarter, and such reduction is expected to expand in the second half. This is the announcement that got so much attention a few weeks ago; that Samsung is finally giving in to the weak market and lowering their production. Those production cuts will be visible before the mid-point of the year and will fully flow through as a larger supply reduction in the second half of 2023.
Now, let me tell you about the outlook for the second half of the year. Demand is expected to gradually recover in the second half of this year, as customer inventory levels will have declined due to inventory adjustment that have been occurring since last year. This demand recovery is expected to begin with consumer products such as mobile and PC, for which inventory or the just month by customer began earlier than they need for commercial applications such as software and studies. First of all, for mobile and PC, we finished good inventory level decreasing in completion rate of first half, safety is expected to improve with the launch of new smartphone and PC promotion in the second half of the year. The memory downturn started in PC and mobile so those segments should see recovery first. New consumer products will lead to higher demand in these segments in the second half of 2023.
In addition, the trends toward high density based on overall price electricity is expected to continue. Thus, we are expecting to see our demand pattern over weak in the first half and strong in the second half. For server, the timing of demand recovery may be later than for server and mobile since customers’ inventory orders started relatively late. Low memory prices lead to higher demand as customers respond by buying more bits. Demand for memory in PC and mobile in the second half of 2023 will be strong. Bold statement. The server market recovery will be later, probably in 2024.
For DRAM, in the first quarter, our bit growth decreased by a percentage just into the double digits, and ASP decline by a percentage in the low to mid-teens. For the second quarter of this year, we expect market bit growth to increase by a low double-digit percentage, and our EBIT growth should be similar to the market. Samsung’s DRAM bit supply fell 11% sequentially in Q1 of 2023. DRAM ASPs dropped 12-13%. In Q2, they expect to match the market with 11-12% bit growth. Production cuts underway should lead to lower bit growth in the third quarter of the year and beyond.
For NAND, in the first quarter, our bit growth increased by a low single-digit percentage, while ASP fell by a percentage in the high teens. For the second quarter, we expect market bit growth to be in the mid-single-digit range and our bit growth should be similar. The NAND market is even weaker than DRAM. Samsung is not seeking to gain share, at least in the second quarter.
Price changes are an item that has a direct impact on earnings and so during the first quarter, the additional price decline had a direct impact on our profitability as well. Another factor in our first quarter results is the inventory valuation loss. Actually, in Q4, the inventory valuation loss started to kick in from the NAND product, but with the additional price decline recently in DRAM, there was a larger inventory valuation loss recognized this quarter, and that had an additional impact on our performance. The inventory write-down they took in Q4 was on NAND. In Q1, the pricing environment deteriorated to the point where a write-down was taken on DRAM as well.
On the production side, we do see big growth restrictions given the fact that there is a die-sized penalty as part of the switch towards the new interface products. Also, there is growing difficulty with the advanced nodes in terms of technology and also increased production lead-times. DDR5 DRAM has a larger die size, all else equal, so the conversion from DDR4 will lower output. Newer technology nodes are harder and have longer production travelers.
That said, we were looking into our production by product lineup and found – determined that in certain products, we have already secured sufficient volume to supply and meet customer demand changes and therefore, decided to lower production in these products. This reads to me as a soft way to couch production cuts. It makes it sound like “we lowered output of these product because we have enough already,” which is true, but really the market is oversupplied so we are lowering supply. It’s just a weird way to say it.
Therefore, the lowering of production that has been announced, what we mean is being carried out mainly around the legacy products that we have already secured sufficient volume to meet mid to long-term demand and this comes on top-of-the-line optimization that already started in Q1, and so we expect the size of the reduction to be far more meaningful. They already started lowering production in Q1. The reductions underway now are bigger. Most of these reductions are “legacy” products, which may be any older node. Those are the ones to reduce as they are the most expensive.
Summary
Samsung finally gave in and said they will cut DRAM and NAND wafer production. Unlike Hynix and Micron, they did not say by how much. They indicated wafer output cuts had already begun in the first quarter of 2023, they will just be increasing them in Q2 and beyond. The company said they will not reduce their capital expenditures. Samsung has been vague and cagey about their production and investments levels. They could be reducing wafer starts a lot or a little, though their commentary on matching the market supply growth in Q2 suggest their cuts are at least in the neighborhood of the 20-25% that Micron and Hynix announced. Samsung’s superior size and cost structure can put crushing pressure on Hynix and Micron. They can hold prices down and take more market share. However, Samsung can’t get much bigger in DRAM than the 45-45% share they already hold. Customers won’t let them, to the degree they have a choice. Their profits have been reduced by more than 80% because of the weak memory market. I think they have simply had enough. Why continue to extend the downturn and your own pain for a few percent market share? It doesn’t add up. They can’t kill off any more DRAM companies as they did back in the 1990s and 2000s with Taiwan. NAND needs another consolidation move but that will probably take more pain that Samsung is willing to suffer, and the effect on overall supply health might not be worth it. WD and Toshiba already act almost as a single entity. If they merge the effect on supply rationality will be negligible. Hynix already bought out Intel. Micron is the odd company out with a sub-scale NAND business. The only consolidation left is for Micron to somehow join with WD and Toshiba. That would get NAND down to three players, the same number that are left in DRAM.
-S. Hughes (cyclical long MU)