Sumco Corp
TSE:3436
Earnings Call Analysis
Q3-2023 Analysis
Sumco Corp
Chairman Hashimoto highlighted that the company's third-quarter profits exceeded expectations, coming in at approximately JPY 4 billion higher than projected. This was attributed to a combination of delayed depreciation and cost reduction initiatives, with each contributing equally to the lower-than-expected costs.
Looking ahead to the fourth quarter, the company expects to maintain sales levels comparable to Q3. However, they anticipate a significant decline in operating profit (OP), primarily due to a substantial reduction in production volumes, which will have an estimated JPY 6 billion negative impact quarter-over-quarter. This strategic production cut aims to avoid an excessive buildup of inventory in response to a projected drop in volumes for 2024.
In terms of shareholder returns, the company is guiding for a year-end dividend of JPY 10 per share, resulting in a full-year dividend total of JPY 52. This represents a stable dividend payout ratio of 30.5%, reflecting the company's commitment to delivering shareholder value even in less favorable market conditions.
The market for silicon wafers remains challenging due to factors such as the U.S.-China trade war, geopolitical tensions, and inflation impacting consumer sentiment. Despite finding a probable bottom for 200-millimeter wafers, the current market conditions remain uninspiring, with the market down approximately 25% from its peak.
Chairman Hashimoto expressed optimism for the long-term outlook, citing technological advancements that are likely to propel new wafer demand. This optimism is driven by trends in logic chip functionality, chip stacking techniques, and developments in 3D DRAM stacking, as well as the growing need for silicon interposers. These factors suggest an eventual upswing in wafer demand, despite the existing market woes.
Current customer wafer inventory trends suggest that while there might seem to be a decline in inventories, the reality is that high levels persist. The inventory levels across both logic and memory segments are notably high. The demand for logic use at epi wafers is declining gradually without a drop in prices, contrasting with the sharp fall in demand for memory use. This paints a picture of an industry undergoing correction while bracing for improved conditions anticipated in 2025.
You don't have any saved screeners yet
Thank you for your participation today. This is the results briefing for the third quarter of the fiscal year ending December 2023. Before starting the presentation, allow me to confirm today's materials, which consists of 3 items: the brief statement on consolidated financial results for Q3 fiscal 2023, the announcement regarding revision to dividend forecast and the presentation deck entitled Results for Third Quarter of Fiscal Year 2023.
This will be a 60-minute briefing, which will end at 5:00 p.m. Next, a disclaimer. The estimates, expectations, forecasts and other future information discussed here shown in today's materials were prepared based on the information available to the company as of today and on certain assumptions and qualifications, including our subjective judgment. Actual financial performance or results may differ substantially from the future information contained in this material due to risk factors, including domestic and global economic conditions, trends in the semiconductor market and foreign exchange rates.
We will have presentations today from Representative Director, Chairman and CEO, Mayuki Hashimoto; and Vice President, CFO, Michiharu Takii. Hiroshi Ito, Executive Officer and General Manager of Accounting is also on hand. Chairman and CEO, Hashimoto, will discuss our forecast and operating environment to be followed by an explanation of the financial results by CFO, Takii. We have set aside time for a Q&A session as well. I will now hand over to Chairman Hashimoto.
I am Chairman Hashimoto. I will start with the overview on Slide 5 of the presentation. This is a summary of the Q3 results. Profits were well ahead of our forecast, but the main factors were the weekend and better-than-expected cost split roughly evenly. The lower-than-expected costs were due to delays in incurring depreciation and cost reduction initiatives also split roughly evenly. Profits came in roughly JPY 4 billion higher than projected. Sales was largely in line with forecast, down only slightly.
Turning to the Q4 earnings forecast, while we expect sales to be unchanged Q-on-Q, we project significant sequential declines in OP. This begs the question of why OP is expected to fall so much despite sales being largely flat sequentially. The decline is a function of an expected drop in production volumes. Although sales are forecast to be unchanged Q-on-Q, Sumco has lowered production volumes substantially. In addition, we will be conducting regular maintenance at our flagship plant in November. As a result, we project a JPY 6 billion impact from lower production volumes on a Q-on-Q basis.
There are a number of other factors as well, such as an expected increase in depreciation, which will contribute to the JPY 8.6 billion Q-on-Q drop in OP. The decision to reduce overall production volume is because we know that volumes going forward into 2024 will decline. We have chosen to act preemptively in adjusting production to ensure our inventory does not rise to excessive levels. Please turn to Page 6. This slide shows shareholder returns. We are guiding for a fiscal year-end dividend per share of JPY 10 for a full year total of JPY 52. This represents a dividend payout ratio of 30.5%. Next page, please.
So what does the market environment for silicon wafers look like? Very simply, the market environment in both Q3 and the outlook for Q4 continues to be weak. The weakness is in both logic and memory. Consumer and other applications for 200-millimeter wafers are also weak. The reason for this weakness is clearly the U.S.-China trade war which has significantly depressed the Chinese economy. Other geopolitical risks are also contributing to the lackluster markets. The situation in Ukraine has led to inflation, which is depressing consumer sentiment. This is also a significant factor. The drying up of elevated pandemic demand is also a factor, inventory turnover is sluggish.
On our outlook going forward, if we look at 2024, the only area of end demand that is firm is EV-related demand. All other areas are seeing continued weakness. Although PCs appear to have bottomed out, demand has only reverted to previous low levels, highlighting just how depressed recent conditions have been. We expect demand to remain at low levels. On the cadence of a recovery, there are differences between logic and memory. Obviously, with Logic, the major players are the foundries, which produce to order. In contrast, the IDMs are the major players in memory, so production is based on anticipated demand.
By nature, therefore, memory has a tendency to overshoot. This time is no exception. This time around memory players are carrying elevated levels of product inventory. As a result, even if the markets recover, it may take time before wafer inventories are worked down and wait for demand picks up. In the case of memory, I suspect there is likely to be a sizable time lag before wafer demand recovers. This makes it very difficult to forecast, but I believe that logic may gradually recover from around the middle of 2024. For memory, it may well take until the end of the year before there is a recovery in terms of wafer purchases in my view. Page 8, please.
This slide shows the wafer trend for 200-millimeter by quarter. 200-millimeter wafers have been impacted by weakness in consumer applications. The lowest level since we began collecting data had been in 2016. However, market trends are lackluster with the wafer trend dropping below 2016 levels. However, I think the bottom for 200-millimeter wafers may be around this level. Maybe what we see is some bottom calling for a while before things start to look up. In any case, current conditions are uninspiring. Relative to the peak, the market is down by around 25%. Page 9, pleased for the trend for 300-millimeter. As you can see, there was a drop off in Q3. I expect to see this trend persist. As mentioned earlier, I expect we may see a recovery around mid-2024 or later, so the weaker trend is likely to continue until then.
Please jump forward to Page 11, please. Up to this point, most of what I have talked about has been depressing. However, if the question is, will the market trend for wafers remain weak going forward? I would say, obviously not. By nature, the semiconductor market is a market that is very lumpy. However, if we look beyond the cyclical trends for the market, there are technological developments that can change the slope of market growth by generating new demand. I would like to cover this in more detail. For instance, if we look at logic chips, there are continued improvements in functionality, which are driving growth in the transistor count. The vertical access on the left is transistor count with the blue dots representing transistors.
As you can see, transistors have grown linearly over time. Transitions are essential for computing operations, so as computing performance rises, it also drives growth in chip area. Chip area is represented by the orange dots with the scale shown on the vertical access to the right. We have plotted companies A, B, C and D on the chart. As we have identified company D as producing GPUs, you can probably guess which company this refers to. As you can see, chip area has been growing significantly. As such, not only will there be an increase in the number of leading-edge wafers, but the surface area of leading-edge wafers is also likely to grow over time. Slide 12, please.
Increasingly, there has been a trend toward combining chips, mounting separate chips onto an interposer. Currently, this is limited to only leading-edge chips. However, the individual chips are not simply placed on a base but must be interconnected. Given that the substrate must incorporate connecting circuitry, the substrate must be relatively high grade silicon. This is a new and emerging application that did not exist previously. Slide 13, please.
If we look at CMOS image sensors, demand for ever better image resolution is driving a trend towards chip stacking. This is particularly true for automotive applications. If we think about the requirements for autonomous driving at night, there will be a need to expand the size of the photodiode. However, as the [ photodiol ] gets larger, it will not be possible to accommodate the peripheral circuitry that had previously been incorporated in the same chip. There will also be an increase in complexity as well. This could drive a shift to 2 chips for now. This trend will drive an increase in the number of wafers. Slide 14, please.
This slide examines improvements in 3D DRAM stacking, achieving greater bit density and faster transfer speeds ultimately will require a 3D structure. The stacked structure will need to have a silicon interposer at the bottom. The emergence of interposers or carriers is a departure from the trend to date. We are now starting to see demand for interposers emerge. We have already received many requests from customers and have been shipping samples. This is the trend for leading edge, which suggests demand for more wafers. Slide 15, please.
Finally, looking at memory. Recently, we have seen an emerging trend toward bonding memory to logic. YMTC in China is doing this. To date, logic circuitry would have been incorporated in the periphery of memory, but the structure has recently become more complex. What we are now seeing is a shift to specifically fabricating logic circuitry, which is then bonded to memory. Directionally, the trend is toward increasing layer numbers, but given the challenge of fabricating 300 or 400 layers, it's possible we may see a shift to bonding to 300 layer NAND back-to-back to get to 600 layers. This would imply a doubling of wafer consumption.
So technologically, I believe there is a trend towards using more silicon wafers. Unfortunately, however, the industry is prone to strong volatility as a result of changes in the macro backdrop. This is because the rapid pace of growth makes it difficult to project the future, in my view. This is being compounded by emerging geopolitical risk. The increased complexity of conditions surrounding the market are also making it difficult to forecast. Slide 17, please.
This shows customer wafer inventory trends. Optically, it appears as if September inventory levels declined, but the drop is a reflection of the fact that there were customers that suspended orders in September. The actual trend remains unchanged with inventories at persistently high levels. Slide 18, please. This shows the breakout of customer inventories between logic and memory. You can see that wafer inventory levels are similarly high. Slide 19, please. This chart shows trends for logic use at epi wafers. The orange line is inventory. Unfortunately, inventory is at record high levels. However, the key feature of epi to highlight is that the blue line, which is demand is not dropping off sharply, but instead is declining gradually. As well, prices have not fallen. This is in contrast to Slide 20.
This is the same chart for memory use PW. As you can see, there continues to be a steep increase in wafer inventory. But in the case of memory, you can see that demand is falling too. This is actually quite rare. Although there have been instances in the past where there have been slight declines, it is unusual to see such a sharp decline in demand, although it reflects measures to reduce inventory. Effectively, this is a double whammy with both a decline in demand and an inventory correction coming at the same time. This is the current situation. Please see Slide 21 for the overall forecast.
As you can see, 2023 has fallen sharply, down 10-plus percent, but if second half 2023 conditions remain unchanged, then we would expect to see a slight fall in 2024 as well. I note that the market was relatively solid in first half 2023. However, if you look at 2025, we expect a relatively strong recovery in 2025. The blue dotted line is the PPP, GDP trend, which we have updated with recent data points. Although we have seen significant moves recently, after all is said and done, our view would be that the trend will revert towards the PPP, GDP line. For a while, in 2021 and 2022, I have thought that we might see a divergence from this trend, but it doesn't appear to be the case.
It appears that purchase levels in 2020, 2021 and 2022 were too high relative to PPP, GDP, and we are paying the price for that overshoot now. In other words, what we are seeing now is a correction for excessively high purchase levels. This is an industry that tends to overreact in both directions. If we back out the numbers from the PPP, GDP line, it implies that the magnitude of the excess purchases is probably around 10 million wafers. On that basis, you could argue that the correction could be worked through over the course of a year in 2023, but it's 2024.
Another reason for the unprecedented change in the trend curve is the emergence of geopolitical risk, particularly even in China is a major factor, in my view. If you look back to 2009, even when we had the global financial crisis, we did not see such a sharp decline in silicon wafers. Therefore, the current wafer and semiconductor recession is the result of a complex combination of factors. Given this, I think it will take time for the market to recover. However, the overall trend for semiconductors, as you know, is that they are essential building blocks, so I do expect to see a recovery over time. This completes my section of the presentation. I will hand over to CFO, Takii, to talk about details of our earnings.
I, Takii, will present the earnings in more detail. Please turn to Slide 23. This slide shows the results for the first 9 months of 2023. Nine-month sales were not down that much on a year-on-year basis, but OP dropped JPY 18.1 billion year-on-year. The big decline came in Q3 with stand-alone OP for Q3 at JPY 15.1 billion, down significantly year-on-year. Ordinary profit also fell. However, profit attributable to owners of the parent was up JPY 7.2 billion year-on-year on the extraordinary gains of JPY 20.1 billion generated in Q1 related to the acquisition of the polysilicon business.
Nine-month CapEx was up substantially year-on-year at JPY 225.3 billion on continued greenfield investments. Similarly, depreciation also rose. EBITDA was down on a year-on-year basis with most of the declines coming in Q2 and Q3. The EBITDA margin was also down. Please turn to Slide 24. This is the analysis of changes in operating income. First, the sequential analysis. Q3 operating profit dropped JPY 5.7 billion Q-on-Q. This was much better than expected. If we break down the JPY 5.7 billion sequential decline, costs were largely flat Q-on-Q. We have been forecasting an increase in costs.
In actual fact, costs did increase but were offset by favorable yields and expense discipline across a wide range of expenses. Depreciation was lower than we had expected by around JPY 1 billion, rising JPY 2.7 billion Q-on-Q on the back of slight changes in the ramp-up of new equipment. Sales-related variance was in line with expectations as a result of sharply lower wafer volumes falling JPY 5.7 billion. ForEx made a positive contribution of JPY 2.6 billion, owing to the weak yen. The yen was weaker than we had expected. On the right, we show the 9-month OP waterfall chart. There was a significant year-on-year increase in cost of JPY 18.3 billion.
This breaks down into an JPY 8 billion increase in materials costs, another JPY 8 billion in higher electricity costs, and a JPY 2.1 billion year-on-year rise in labor costs. Costs were slightly better than we had expected on better-than-expected yields and other factors as touched upon in the discussion of Q3 sequential change. Depreciation increased JPY 6.7 billion year-on-year. This was mentioned in the discussion of forecast, but for sales-related variance, there was a negative impact of around JPY 6 billion, with volume declines having a negative impact of JPY 24 billion, which was offset by a positive of JPY 18 billion on the year-on-year increases in price.
ForEx had a positive impact of JPY 13 billion on the back of a JPY 10 weakening of the yen year-on-year. Slide 25, please. Slide 25 shows the balance sheet and cash flow statement. Cash and time deposits on the balance sheet fell JPY 76.4 billion. I will explain this in more detail in talking about the cash flow statement. Finished products and work in progress increased. Raw materials and supplies rose JPY 23 billion. The impact of the acquisition of the polysilicon business pushed up polysilicon inventory by JPY 6 billion. As a result of the rapid reduction in production levels, materials consumption fell slightly leading to the increase.
Supplies also increased on the back of a buildup of inventory related to greenfield investments. Tangible assets increased significantly on the back of higher greenfield CapEx. Interest-bearing debt increased roughly JPY 50 billion. This breaks down to new borrowings at FST of JPY 20 billion and at Sumco in Japan of JPY 30 billion. If we look at the capital account, retained earnings increased, reflecting the impact of the roughly JPY 60 billion in net profits, offset by JPY 30 billion in dividend payments for a net increase of JPY 28 billion.
On the cash flow statement to the right, the subtotal of pretax profits, extraordinary income and depreciation was JPY 113.6 billion. The increase in inventory was a negative JPY 16.4 billion. Under others, a major factor was tax payments on the back of strong profit levels in the previous fiscal year. Operating cash flow was JPY 75 billion. Capital expenditures were JPY 220 billion. With regard to the negative free cash flow after dividends paid, we raised around JPY 50 billion in new borrowings and tapped into cash and time deposits, hence the decline in cash on hand. Slide 27, please.
On Page 27, we show our forecast for Q4. We expect profits to be down significantly in Q4. I will go into more detail later. Chairman Hashimoto, commented earlier on the lower capacity utilization. The focus will be on shipping from inventory. On a full year basis, we project OP to fall JPY 41.3 billion year-on-year. A major contributor to the drop is the year-on-year drop in Q3 given that Q3 2022 OP was around JPY 30 billion compared to the JPY 15.1 billion in Q3 this year. For Q4, last year, we generated close to JPY 30 billion as well but are projecting JPY 6.5 billion for Q4 this year. The year-on-year decline in second half OP accounts for the majority of the JPY 41.3 billion year-on-year drop in OP.
We generated extraordinary income in Q1, but after deducting income tax payments, we project net profit to fall JPY 8.8 billion year-on-year to JPY 61.4 billion. Full year depreciation is projected to be JPY 71.8 billion. The OPM and EBITDA margins will decline as well with the drop acute in second half and particularly Q4. That said, we expect to maintain our ROE over the 10% level. Slide 28, please.
Slide 28 shows the analysis of change in operating profit. First, the Q4 sequential change on the left. Sales are largely unchanged Q-on-Q. We haven't significantly changed our ForEx assumption. There isn't much of a change in terms of sales volume, but in October, we conducted regular maintenance at all 3 of our Crystal ingot plants. In November, the Imari plant, which is the mainstay 300-millimeter plant will conduct regular maintenance. In addition, we will be lowering capacity utilization at the 200-millimeter plant. As a result, there will be a significant decline in capacity utilization for both 300 and 200 millimeter. This will reduce work in progress and finished product inventory, although we will incur increases in costs.
Sales-related variance is expected to drop a substantial JPY 7.1 billion Q-on-Q. Effectively, we are reducing production significantly relative to sales. Looking at the analysis of year-on-year change for full year OP on the right, we expect OP to fall JPY 41.3 billion. We expect costs to increase JPY 18.7 billion year-on-year. This is not significantly higher than the JPY 18 billion year-on-year increase in cost for the 9-month OP, we are not expecting to see a significant year-on-year increase in costs in Q4. That said, the expected breakdown is a JPY 10 billion year-on-year rise increase in materials costs, a JPY 7 billion to JPY 8 billion year-on-year increase in electricity costs, and a JPY 2 billion year-on-year rise in labor costs.
Depreciation should also increase. Sales-related variance breaks down into a roughly JPY 22 billion positive from higher prices year-on-year, but a hefty JPY 45 billion negative from lower volumes. On a year-on-year basis, we expect a positive contribution of around JPY 13 billion from the weaker year. Page 30, please. For your reference, you can see that the EBITDA margin is not off that much up to Q3 but in Q4, for the special factors cited previously, including the significant drop in capacity utilization, we expect to see a major decline. This completes my section of the presentation.
Thank you. We will now open the floor to questions. We will start with Mr. Enomoto of BofA Securities.
Please talk about your expectations for the wafer market on a quarterly basis. Where do you see the big bottom for the market? Was it Q3? Or will it be Q4? Also, this time, you have stepped up your production cuts, but when do you expect to be done with production cuts? The reason why I ask is because it appears that DRAM and memory have already bottomed out, which suggests that semiconductors are now starting to recover. Based on what you have said, it seems that a recovery in wafers may take more time, which makes it challenging to project when wafers will bottom. Any color you can provide would be helpful.
It is the same for me as well. Although I can talk about how things might turn out, there is no guarantee that my prediction will be right. I say this because this time around, it is very difficult to make projections. In particular, with regard to geopolitical risk, we had no way of anticipating that the Trump administration would take such aggressive action on China or that the Biden administration would take it even further with sanctions. We did not expect to see such actions.
Additionally, these actions muddy the water significantly as they cannot be predicted within the framework of normal market economics. This makes it extremely difficult to make projections. What I can say at a minimum is that with regard to memory, we have no sense that the market has bottomed in any way. Customers continue to keep production levels low. The fact that prices have recovered very slightly is more a consequence of the dramatic reduction in production, which has led to lower supply volumes. So I don't feel at all that the market has bottomed. In my view, the bottom will be next year. Conditions for logic are similar. And if anything, the correction in logic really only got started in Q3.
Given this, the September quarter was certainly not the bottom. I think the bottom will come next year. On the question of the timing of a recovery, as you will have seen, customer inventories are at record high levels. Obviously, we would like nothing better than to see an early recovery. But if you look at the inconvenient reality, it seems clear that the market is unlikely to recover that quickly. Instead, I think the market bottoms next year, after which we see a recovery. For memory, the trough is actually quite deep. The trough for logic is not as deep in comparison. I have talked about this a number of times, but if you think about memory, when consumers have money, they will buy a smartphone with plenty of memory, but when economic conditions are weak, they trade down.
I think this is the way to think about it. Therefore, when there is a recession, memory is the first to be impacted. The same can't be said of logic. Just because you don't have money, you can't buy a smartphone without a logic chip. It wouldn't be a smartphone without a logic chip to begin with. Given this, a downturn in logic demand is likely to be milder in my view. So to reiterate, I think the end markets for logic are likely to recover sometime midyear next year. However, a recovery in memory is likely to take until the end of next year in terms of wafer demand. In terms of the recovery of the end markets, logic benefits from trends such as AI, so end product demand should recover toward the end of the first half of next year, in my view. But for memory, I think it will take until the second half of next year for a recovery.
This is because not only our end product inventories at very high levels, the memory players are carrying significant inventory. In terms of recovery, if you think about it, the world's population is 8 billion, but China alone accounts for 1.45 billion. On the back of China being driven into a recession, I think the drop in consumption in China is having a meaningful impact. The timing of a recovery in Chinese consumption is significant. And if the U.S. were to put further pressure on China, it would further prolong the recession. These factors make it very difficult to make projections. I would be interested in knowing what you, Mr. Enomoto think about this.
Thank you. The reason why I was asking was because the leading edge customer has bottomed out. So I was hoping that while it may be some way away, there might be some signs of light at the end of the tunnel.
Leading edge has indeed bottomed out, not memory, but leading-edge logic. But legacy product is still weak. What is particularly problematic is that all of the memory players across the board with no exceptions have cut production significantly. If you look at logic, while there were customers that have made major production cuts, there are others that have been able to get by with only a slight dip in margins despite cutting production. Such players have not dropped into the red. There are very significant differences between the individual companies in logic, so it's tough to generalize.
However, leading-edge production at the companies that have managed to remain relatively unscathed, haven't dropped much to begin with and seem to be coping relatively better.
Thank you. Next is Mr. Ikeda of Goldman Sachs.
How should we think about the supply-demand balance related to what you show on Page 21? You have said that volume, which has been falling from second half 2023, will decline significantly into 2024. Given these conditions, do you expect you will be able to continue to raise prices? It seems to me that the best you can expect is that prices do not fall given the sheer scale of customer losses. That said, you have indicated that the prices reflect the appropriate compensation for greenfield investments and that, therefore, you need to see prices continue to rise.
Please talk about the balance between volume and price for 2024. For volume that customers choose to push out beyond 2024, how do you propose to ensure that the LTAs are respected? Any color you can provide would be helpful.
100% of Sumco's 300-millimeter wafers are covered by LTAs out to 2026. Our thinking is that we will not make changes to either the total volume of the LTAs or prices. However, with customers cutting production and with no warehouse space for more wafers, it would be impossible to force the customers to take delivery of wafers. In such cases, we have agreed to defer delivery timing without changing the overall contract volume, although actions may vary from customer to customer. We aim to arrive at a solution where the pain is shared. So we have not made any changes to prices whatsoever.
If we were to lower prices, we would not be able to achieve our target return on investment. Our customers are respecting total contract volumes, but we have allowed them to have some flexibility in terms of delivery timing, agreeing to delivery push-outs where necessary. Forcing customers to take wafers they cannot use or store simply because it is in the contract is not a realistic way to approach business. However, lowering prices would have implications for our return we have not changed pricing.
With regard to pricing for 2024, I imagine that renegotiations are upcoming. But even if the prices don't rise 10% in 2024, directionally, you will be raising prices. Is that correct? If you are unable to raise prices in 2024, directionally, you will be looking for the customer to make up for it in 2025.?
Yes, that's correct. There are customers that are slated for price hikes in 2024. We haven't started negotiations yet, but basically, we expect customers to respect the terms of LTAs.
Understood, I hold high expectations for Sumco. Thank you.
Next is Mr. Nishiyama of Citigroup Securities.
In looking at the chart on Page 18, showing inventory levels for the memory makers the situations in March and September appear to be very similar. In other words, in both of these months, purchase volumes fell, which seems to have led to a dip in customer inventory levels. What I find interesting is that your peer has made similar comments. In announcing March quarter results, they said that wafer shipments would increase in the June quarter.
This time around at the September quarter results briefing, your peer indicated that they expected to see a slight recovery in shipments in the December quarter. Is the background to the dramatic drop in wafer purchase volumes in September, similar to what you saw in March? If that is the case, can we expect to see a similar trend to the June quarter in December quarter shipments?
The drop in March is different than the drop in September, in my view. What happened in March was that the decline in that month was offset by an increase in shipments in April. The customer was looking at the average for these 2 months, and it was simply a timing difference. So the basic underlying trend for March and April was the same. However, the drop in September was a reflection of the absence of orders. There was no suggestion that customers would make up for this by buying more in November, for instance. So the September drop is a reflection of an overall push out in delivery timing and is different than the activity in March and April. In other words, customers are truly reducing production.
If that's the case, and there won't be a rebound in purchase volumes in October, does that mean you expect memory customer inventories will have peaked out?
So we don't expect inventories to keep falling. We expect inventory levels will remain high for a while, but we hope that over time, there will be a gradual decline. However, all of the customers have significantly cut production from September onwards and are reducing purchase volumes, so maybe there will be a flattening of the rising trend in inventories with inventory levels declining gradually over time from next year. I think that is likely, given current capacity utilization. Customers are likely to feel the same way, which is why they are deferring deliveries.
Does that mean it has peaked out?
I am not clear on what you're referring to when you say peak out, but with the customers cutting production dramatically, I don't think you can say that things have peaked out.
Understood. Thank you.
Next is Mr. Miyamoto of SMBC Nikko.
I would like to also ask about the outlook for prices as we head into 2024. First, can we expect the increase in prices for 300-millimeter wafers covered by LTAs to rise by the same magnitude as they rose in 2023? For the non-LTA portion, which is probably only the FST portion, is there a risk that prices may fall slightly? For 200-millimeter wafer, similarly, what is the outlook for prices, particularly given a more challenging supply-demand situation? Please talk about 2024 prices in more detail.
With regard to LTAs, there are some LTAs where the prices rise in a step-wise function every year, and some where the price is fixed for the 5-year period. So I don't know if the magnitude of the increase will be the same as it was for 2023. Some of that is because it is subject to negotiations, but there are also differences related to product mix. However, for those contracts where prices are set to rise, of course, we will be asking for higher prices.
In terms of spot, there is no spot business at the parent, but there is spot business at FST. It is true that spot prices have declined, although the magnitude of the declines is not as bad as we have seen in the past. The reason for this is that the spot market has virtually dried up. All of the customers are locked into the LTAs. The current market is not one where you can drive more volume by lowering prices. If you can't buy, the price point is irrelevant. This is the current situation with customers having cut production significantly.
For 200-millimeter, my sense is that surprisingly, prices have not fallen that much. Spot prices may have fallen, but within 200-millimeter, Sumco Japan is focused on high performance, so prices have not declined that much. However, for simple commodity PW, I think prices have fallen.
Next is Mr. Okazaki of Nomura.
You have said you will reduce production significantly in the December quarter. Please comment on the image of production levels for the March 2024 quarter and beyond. December production will be down on the combination of regular maintenance and demand factors, but can you comment on your current view of expected production levels for both 300-millimeter and 200-millimeter in the March 2024 quarter and beyond?
For the March quarter, I am aiming to raise production levels slightly Q-on-Q. The December quarter production is impacted by regular maintenance, which we are not offsetting in any way. In addition, we are cutting production significantly to lower inventory because I don't want to carry too much inventory. The plan for the March quarter is to align production volume with sales demand, so production levels should rise slightly.
Thank you. So basically, from a production standpoint, the December quarter will be the bottom, correct? And subject to demand, production in the March quarter will increase?
Yes.
Next is Mr. Nakahara of Tokai Tokyo Research Center.
Can you comment on how you are thinking about price and volume as it relates to the waterfall charts for change in OP? Given progress in scaling, there should be a higher volume of high price point wafers. I suspect that the production cuts are being targeted at lower price point wafers. Am I correct to assume that the impact of the implied improvement in mix is included in the price impact?
Basically, the answer is yes. Leading-edge product commands higher prices, and it is also true that leading-edge volume is increasing. It is also true that legacy volumes are falling. That said, total volumes are down, which makes it tough to say. However, it is fair to say that mix should be better.
Basically, is it fair to say that higher-priced wafers are profitable?
Chairman Hashimoto. Yes, of course.
Next is Mr. Yoshida of CLSA Securities.
With regard to CapEx and depreciation, given the challenging operating environment, how are you thinking about next year? If possible, can you talk about how you are thinking about both 2024 and 2025? Last time, you indicated that depreciation would be flattish out to 2025 and then would drop off significantly in 2026. Has there been any change in your long-term view of CapEx and depreciation?
On depreciation, I will have CFO, Takii comment later, but we are slowing down CapEx, particularly in Taiwan, where PW is the main focus. We have significantly moderated the pace of CapEx in Taiwan. On the domestic side, we are monitoring the situation and monitoring the pace where possible. However, we have already placed orders and bought all the equipment. Also, it isn't possible to defer capacity expansion investment for multiple years. So we will do what we need to do. Fortunately, this is a growth industry, so the equipment we have purchased will not go to waste.
So we are monitoring the situation and adjusting ourselves accordingly. That said, as I mentioned earlier, with geopolitical risk, politics does make it difficult to make projections. It is certainly true that we are seriously engaged in discussing how things may play out. Can you comment on depreciation?
We shown 2024 depreciation in the table on Slide 21. For 300 millimeter, we expect a slight decline compared to the average for 2023. Given there isn't a strong need to ramp up facilities in a hurry, depreciation may be pushed out slightly from what we have talked about previously.
I see, so some of the depreciation forecast for 2024 could potentially be deferred? What is your view on depreciation for 2025 and 2026?
Likely that there will be some further pushouts in terms of timing for depreciation.
I see, understood. Thank you.
Next is Mr. Yamada of Mizuho Securities.
This relates to the previous question, but in terms of how to think about cost for Q4, there will be a negative impact on profits from the change in production levels? A negative impact from regular maintenance, a negative impact from the earnings-linked bonuses since this can only be fixed in Q4 as well as depreciation. Given that you apply the declining balance method, deferring the timing of the ramp-up of equipment, I believe, will lead to Q-on-Q declines in at least Q1. Of the components of costs I have listed, are there items where you expect to see Q-on-Q increases in Q1 2024?
Takii will respond. This year, materials costs and electric power unit prices rose substantially. This was the major reason for higher costs in 2023, although we did raise base salary levels as well. I would expect that these 2 major items, materials costs and electric power unit costs should stabilize next year. This is because a key component of the increased cost this year was the weaker yen which should stabilize next year.
Labor costs will be impacted by annual salary increases, but apart from depreciation, there really aren't cost items where we expect strong year-on-year increases. To your point about depreciation, it is true that depreciation levels will reset as we start the new fiscal year and should be down versus Q4. However, as new facilities come online, depreciation will increase over time, so depreciation is likely to be higher than this year.
On that point, can I confirm, Q1 depreciation will definitely increase year-on-year and full year depreciation for 2024 will be higher than 2023. But relative to the Q4 level, Q1 2024 depreciation is likely to fall sequentially, but then depreciation should rise sequentially over the course of the year. Is that correct?
Yes, that's correct.
So that's the correct way to think about it, right?
Yes, that's fair.
And on materials and electric power costs, given that they are already high, it doesn't feel like there will be a significant Q-on-Q increase into Q1. Is that fair?
Yes, that's fair.
And labor costs will increase to reflect the increase in base salaries, but given that you are paying earnings-linked bonuses in Q4, which will push up labor cost. In Q1, the absence of this bonus should mean that there won't be a big Q-on-Q increase in labor costs for Q1, right?
We actually accrue for the bonus, so this would not be a driver for a specific increase in labor costs for just Q4.
I see, understood. So labor costs will increase Q-on-Q?
Yes.
Understood. Thank you.
Thank you, everyone, for joining the Q3 2023 results briefing. We are grateful for your participation today. We will end the meeting here.
Thank you.
Thank you.