Sumco Corp

TSE:3436

Earnings Call Analysis

Earnings Call Transcript 2024-Q2

[Takayuki Komori, executive]

Thank you for your participation today. This is the results briefing for the second quarter of the fiscal year ending December 2024.

Before starting the presentation, allow me to confirm today's materials, which consists of 3 items: the brief statement on consolidated financial results for Q2 fiscal 2024, the announcement regarding interim dividend and the presentation deck entitled, Results for Q2 Fiscal 2024. 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 and 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, Shinichi Kubozoe. Chairman and CEO, Hashimoto, will discuss our cast and operating environment, to be followed by an explanation of the financial results by CFO, Kubozoe. We have set aside time for a Q&A session as well.

I will now hand over to Chairman Hashimoto.

[Mayuki Hashimoto, executive]

I am Chairman Hashimoto. I will start with the overview on Slide 5.

This is a summary of the results for the second quarter of fiscal year 2024. The results for second quarter significantly overshot our forecast. With regard to sales, the recovery after bottoming out was better than we had initially expected. I will say that our forecasts were conservative to begin with, so maybe the results were about right. There was an overshoot of around JPY 3 billion, with JPY 2 billion from stronger sales and a JPY 1 billion impact from the weaker yen as well as other factors such as cost reduction.

Turning to the earnings forecast for the third quarter of 2024. Optically, our forecasts are for a sequential decline of JPY 4.7 billion in sales to JPY 100 billion, but almost all of the negative impact can be explained by the ForEx assumption at JPY 150 to the dollar compared to the JPY 155 yen level of Q2. We are not expecting a decline in sales.

Operating income is projected to fall by JPY 5.2 billion sequentially, but almost all of the decline is the result of an increase in depreciation which we project will rise by more than JPY 3 billion. The change in the FX assumption is also a factor at JPY 150 to the dollar, for an impact of around JPY 1.3 billion to JPY 1.4 billion. So these 2 factors largely explain our forecast for a Q-on-Q drop in OP.

Next slide, please. This slide shows our dividend guidance. As we have already announced, we will pay an interim dividend of JPY 15.

Next page, please. Looking at market conditions during Q2, 300-millimeter wafers did indeed bottom out in Q1 and gradually recovered in Q2. Although 200-millimeter and smaller diameters also bottomed, rather than recovering, this segment of the market has continued to crawl the bottom, with shipments remaining low. Fortunately, customers continue to respect LTA prices.

On the outlook for Q3, customer inventory levels for leading-edge 300-millimeter wafers, which are mainly used for generative AI, are very low. Customers are running at full capacity utilization and inventory levels continue to decline. In fact, customers are carrying almost no inventory.

However, for wafers used for legacy applications, customer inventories remained high. Demand here is very weak. Customers' utilization for legacy products is running below full capacity, so I think resolving this situation will take time.

As wafers used in AI account for less than 20% of the overall volume, the recovery in overall shipment volumes will likely be gradual. With regard to 200-millimeter wafers, we expect production adjustments that customers will continue. So although the market has bottomed, shipment volumes have continued to bottom crawl.

In terms of prices, LTA prices continue to be respected by customers, but the picture for spot prices is mixed. In certain segments, spot prices have declined. That said, SUMCO has virtually no exposure to the spot market in 300-millimeter. So trends in the spot market don't have much of a direct impact.

On the outlook for semiconductor demand, corrections to industrial and automotive applications are likely to continue. Sales of end product cars are not particularly weak, but channel inventory remains sluggish. Strong demand related to AI applications continues to be the key driver with the recovery in data center applications. PC and smartphone applications have bottomed, so we should start to see a pickup in replacement demand for end products. As such, there should be a gradual recovery here.

Car sales have been firm. So once excess channel inventory has been cleared, we should see conditions return to normal. Having said this, overall demand for silicon wafers is likely to continue its gradual recovery up to the end of calendar 2024.

Next page, please. This is the regular slide, which shows the wafer trend for 200-millimeter by quarter. 2024 is the red line. As you can see, recently, the market has been very weak, dropping off suddenly from last year. 200-millimeter has historically been used primarily for industrial and consumer application. As such, weak macro conditions in China, which is dampening sales of goods to China, is having a major impact. In addition, on the back of rising tensions with the U.S., the Chinese government policy of Buy China is impacting sales of wafers.

Many of the semiconductor companies in China can be said to have a high degree of government involvement. There are suggestions that they are being told not to buy foreign-produced products. Therefore, such players are not buying many non-China produced wafers, instead choosing to procure from local Chinese players. It is very difficult to get a clear sense for the overall picture, but as far as we can tell, this appears to be the situation.

Next page, please. This is the trend for 300-millimeter. I believe we have a clear grasp of the overall market. We are seeing a recovery, as shown here. The magnitude of the pace of recovery is not entirely clear, but as I noted earlier, we don't have a sense that there will be a dramatic recovery.

Next slide, please. This slide looks at the broadening of applications on the back of advances in AI. A good example of future applications would be, for instance, autonomous driving, given that AI will enable high-speed data processing and more effective control of errors. This is the future we envision.

Next slide, please. So what is an AI server? The small diagram on the upper left depicts the conventional server, which is a combination of CPU and DRAM. In contrast, an AI server consists of a combination of GPUs and HBMs with DRAM stacked in 8 layers, all on top of a silicon interposer or substrate. It is considered advantageous to shorten the distance between chips, enabling faster speeds and also reducing power consumption. This is achieved by putting all of the devices on a silicon interposer in a single package.

Fortunately, for us, an AI server consumes roughly 4x more silicon wafers than a conventional server in surface area terms. So strong growth in AI servers is favorable for us. I note that a single AI server consumes 2 300-millimeter wafers.

Next slide, please. This slide depicts the trend of continued surface area growth for AI servers. Over time, AI servers are expected to incorporate 2 GPUs and 8 HBMs in a single package.

Next slide, please. AI aggregates huge data volumes for computation using a trial and error process to work through algorithms, requiring a large number of DRAM chips. There is a need to increase transistor density, but there is a mismatch between the pace of narrowing design rules and the need for higher densities.

The consequence of this is the growth in surface area as touched upon earlier. This is already happening. Chip sizes are getting bigger, especially logic. The assumption that as we move to 5-nanometer or 3-nanometer, chip sizes would get smaller, but because requirements have continued to increase, in fact, chip sizes are getting bigger. This is what is happening.

Next slide, please. In the past, interposers were not used, but silicon wafers are now being used as a substrate. This is to allow for surface connections underneath. As shown in the diagram for logic on the left, on top of a silicon board, we have a computing element on the left, a graphic element on the right and IO in the middle. All of these elements are combined on the same substrate.

Previously, silicon interposers did not exist, but although they do not represent huge volume yet, they are already above the order of several 10,000 worth of wafers. Interposers are expected to grow significantly going forward.

The same is true for HBMs within DRAM. HBMs will be staffed on a silicon interposer with circuitry connections to CPUs or GPUs. Rather than the single layer structure used to date, the trend is towards a stacked structure. Not only that, the base will be silicon. There will be a shift towards 3D structures.

Next slide, please. On this slide, we look at the outlook for growth in leading-edge silicon wafers. This is an area of strength for SUMCO. So I would like to discuss this. for 7-nanometer and finer design rules, wafer demand is projected to increase by 400,000 wafers over a 4-year period. SUMCO's share is greater than 50% in this part of the market. Almost all of the new capacity we are currently building is for leading-edge logic use wafers. We expect utilization of this capacity will be filled by the growth in demand for leading edge. This shows the projected growth in demand for leading-edge wafers.

Next slide, please. This looks at leading-edge wafer trends for HBM. We expect to see continued growth. HBM-used wafer demand is increasing. There has been a recovery in memory. HBM has been the main driver. That said, HBM cannibalizes production capacity, so capacity for conventional DRAM has declined.

As a result, DRAM makers have not been able to keep pace with demand for conventional DRAM. So both DRAM and HBM have been showing improved trends. It is notable that companies that have strength in HBM are struggling, but companies that have lagged with HBM have seen improved conditions. Companies producing HBM have had to sacrifice their capacity for conventional DRAM, which appears to be benefiting other players.

Next slide, please. This slide shows customer wafer inventory trends. We are finally seeing inventory levels start to show relatively solid decline.

Next slide, please. When we look at memory, I actually had felt that it would take much longer for customer inventories in memory to decline. But as a result of the strong trend in AI-used DRAM, there has been a broadening of applications for NAND. Thankfully, this has led to a decline in inventory. However, the improved prices in NAND are partly the result of production adjustments, so NAND production is not necessarily running at full capacity. DRAM, on the other hand, is running at full capacity with trends strong.

Next slide, please. Slide 20 is my final slide. We have seen a number of events, such as, for instance, the pandemic and more recently, equity market volatility and a sharp appreciation of the yen, which makes forecasting extremely difficult. If we attempt to forecast despite this backdrop, when we look at 2023, 2024, there has been a break from the previous prevailing trend as a result of the impact of the pandemic, intensifying trade tensions between the U.S. and China as well as the confluence of a number of geopolitical factors.

The question is, how we get back to the previous normalized trend underpinned by PPP-GDP? We don't necessarily have all the answers, but we assume that there will be a reversion to trend over time.

As such, 2025 is likely to be tough, but in 2026, we should be able to raise utilization for greenfield investments coming online. Next year, however, is likely to be tough in my view.

This completes my section of the presentation. I will hand over to CFO, Kubozoe, to talk about details of our Q2 earnings.

[Shinichi Kubozoe, executive]

I, Kubozoe, will present the earnings for this quarter. Please turn to Slide 22.

This slide shows the results for this fiscal year up to Q2. As discussed earlier, Q2 sales were JPY 104.7 billion, OP JPY 12.2 billion; ordinary profit, JPY 11.3 billion; and profit attributable to owners of the parent was JPY 7.6 billion. First half or 6-month sales were JPY 198.2 billion, OP JPY 20.8 billion; ordinary profit, JPY 20.4 billion; and net profit JPY 12.6 billion. First half sales fell JPY 22.4 billion year-on-year and OP was down JPY 25.9 billion.

Profit attributable to owners of the parent fell a significant JPY 37 billion in the absence of the JPY 20.1 billion in extraordinary gains from the negative goodwill related to the acquisition of Mitsubishi Materials' polysilicon business in the previous fiscal year. First half CapEx on an acceptance basis was JPY 124.7 billion.

We continue to incur greenfield investment. Second quarter depreciation was largely unchanged Q-on-Q for a first half total of JPY 35.9 billion. Further down the table, the OPM EBITDA margin and ROE for both Q2 and first half are as shown on the page.

Please turn to Slide 23. This is the analysis of Q-on-Q and year-on-year changes in operating income. First, the sequential analysis for Q2 on the left. The first half year-on-year comparison is on the right. Sequentially, Q2 sales rose JPY 11.2 billion and operating income grew JPY 2.6 billion. The yen weakened by JPY 8.2 versus Q1, which had an impact on sales. As well, there was a modest recovery in volume in 300-millimeter, which contributed to JPY 11.2 billion Q-on-Q improvement in sales. The analysis of the JPY 3.6 billion Q-on-Q improvement in OP is shown below.

With regard to costs, electric power costs were impacted by the termination of relief measures applied to the renewable energy promotion surcharge pushing up the unit price. Labor costs increased as the impact of annual salary increases kicked in from April. These 2 factors were the key drivers behind higher Q-on-Q costs.

The increase in sales was positive for sales-related variance. There was a positive JPY 2.5 billion impact from ForEx as well. The key driver of the Q-on-Q improvement was the positive sales-related variance.

Looking at the year-on-year change for first half, sales fell JPY 22.4 billion and OP dropped JPY 25.9 billion, the yen weakened by JPY 16.8 to the dollar.

Looking at the bar chart below, costs improved year-on-year by JPY 3.7 billion on a year-on-year decline in electric power unit prices as well as lower labor costs. On the right, there was a JPY 10.8 billion year-on-year positive from ForEx.

However, with the completion of new facilities, depreciation increased. The large negative from sales-related variance reflects the fact that production was running at relatively high levels up to first half 2023, leading to a year-on-year decline in sales volume. The combination of these factors contributed to the JPY 25.9 billion year-on-year decline in OP.

Slide 24 shows the balance sheet and cash flow statement. On the balance sheet, total assets increased JPY 115.7 billion to JPY 1,188.7 trillion as shown in the middle of the table. This is mainly due to the hefty JPY 100 billion increase in tangible and intangible assets to JPY 662.2 billion, on the back of progress on the acceptance of the greenfield CapEx. Total liabilities were JPY 530.4 billion, with interest-bearing debt increasing JPY 114.8 billion to JPY 339.2 billion from the end of the previous fiscal year.

We completed our borrowing for Japan in the March quarter were procured around JPY 17 billion in short-term funds for investments in Taiwan in April, pushing up the balance of interest-bearing debt, up slightly from the end of March. Net assets were JPY 658.3 billion, for an equity ratio of 49.9%. The D/E ratio on a gross basis was 0.57x and 0.31x on a net basis.

On the cash flow statement to the right, relative to first half operating cash flow of JPY 36.5 billion, investment cash flow, composed of capital expenditures on an acceptance basis and others on a net basis, was an outflow of JPY 150.7 billion, resulting in a sizable negative free cash flow of JPY 114.2 billion. As touched upon earlier, the increase in interest-bearing debt was JPY 110.3 billion. Q2 cash and time deposits were largely flat, falling by JPY 1.1 billion versus the end of the previous fiscal year.

On Page 26, we show our forecast for Q3. As covered earlier by Chairman Hashimoto, we project Q3 sales of JPY 100 billion, operating income of JPY 7 billion, ordinary income of JPY 5 billion and net profit attributable to owners of the parent of JPY 3.5 billion. For nonoperating income and expenses, we have factored in a foreign exchange loss of JPY 2 billion.

Depreciation is expected to increase sequentially. We project JPY 21.3 billion for Q3. Our foreign exchange assumption for Q3 is JPY 150 to the dollar. The actual rate for July and August are already largely fixed, so it was a question of what to project for the month of September. We have assumed a level of around JPY 145 to the dollar for September, resulting in an average for the quarter of JPY 150. This is the forecast for Q3.

Slide 27 shows the analysis of change in operating profit. First, the Q3 sequential change on the left. Sales is projected to fall JPY 4.7 billion Q-on-Q. One factor behind the decline is the ForEx assumption of a JPY 5 strengthening of the yen versus the dollar.

In addition, there was a slight pull forward of volume in 200-millimeter and smaller wafers into Q2 as a result of timing of deliveries to customers. The average shipment volume for Q2 and Q3 is not significantly different, but because of the above-mentioned timing, Q3 shipment volumes are expected to decline slightly. 300-millimeter wafers are expected to continue to recover gradually in Q3.

OP is projected to drop by JPY 5.2 billion. This is the result of expected increases in depreciation on progress in greenfield investments, the negative impact of sales variance and the negative JPY 1.3 billion impact from foreign exchange, reflecting assumptions for a stronger year.

Looking at the analysis of year-on-year change for the 9-month forecast. The trend is largely unchanged from the explanation of first half year-on-year changes. While there are positives from ForEx and cost, this is expected to be more than offset by the significant negatives in the form of depreciation and sales-related variance on a year-on-year basis. As such, we expect the 9-month OP to decline JPY 34 billion year-on-year.

The reference materials show trends for a number of metrics. Please refer to this at your leisure. This completes my section of the presentation.

[Takayuki Komori, executive]

Thank you. We will now open the floor to questions.

[Takayuki Komori, executive]

We will start with Mr. Enomoto of BofA Securities.

[Takashi Enomoto, analyst]

Please discuss your view of 300-millimeter demand in second half. If we look at Page 20, which is the global forecast for wafer capacity and demand, the flat year-on-year trend for 2024 implies you are still expecting a significant recovery in second half, but the outlook for volumes in Q3, as implied by the sales variance in the waterfall charts, frankly, appeared to suggest a weaker trend.

I believe you expect the overall market to recover, but your recovery appears to be weak. Are there assumptions you have made in terms of changes in market share? Or are there risk factors that you have baked into your forecast?

[Mayuki Hashimoto, executive]

Frankly, we struggled on this point. It may well be that 2024 does not match 2023 levels. The recovery in the second half of the year for the overall market does not appear to be that strong. Ours is an industry where we don't know how things will turn out until we get there, but the market is surprisingly weak. At this point, I don't think we will see a sharp recovery in second half. So unfortunately, it seems more likely that the recovery will be gradual.

If we look at history, typically, recoveries have tended to be quite sharp, but this time around, although AI-related demand is very strong, for other applications using legacy nodes like 28-nanometer, customers' capacity utilization is still running at 70% to 80%. Unless utilization for these areas get back to full capacity, I think it will be difficult to achieve a sharp recovery.

The major factors are industrial and consumer electronic applications where the pace of recovery is very weak. In particular, our Japanese customer sales to Chinese players on a global basis have apparently fallen off significantly. I think that the weakness in China is indirectly having a substantial impact.

[Takashi Enomoto, analyst]

Can I just confirm that you are not expecting to lose share?

[Mayuki Hashimoto, executive]

We don't expect our share to decline.

[Takayuki Komori, executive]

Next is Mr. Ikeda of Goldman Sachs.

[Atsushi Ikeda, analyst]

This relates to the question from Mr. Enomoto. With regard to the medium-term demand outlook for 300-millimeter you show on Page 20, you talked about the expected growth in silicon surface area for AI servers, but for AI-enabled edge devices such as smartphones and PCs and the potential for increases in silicon content, can you talk about the expected trends here? If possible, can you split this out into logic and DRAM?

In particular, with edge devices, I believe NAND will show strong growth. For 3D NAND, I believe your customers have been gradually raising capacity utilization recently. Please talk about your view of the outlook. Also related to the expected increase in layers for 3D NAND, we are likely to see stacking with COA begin in my view. What impact do you think this will have on the wafer market? If possible, can you comment quantitatively?

[Mayuki Hashimoto, executive]

I can't comment quantitatively as it is still early days and the technology is only just emerging. However, with regard to the magnitude of demand that might be generated by AI-enabled smartphones and PCs, I don't think there will be a big impact.

This is because if you compare the information that you might be able to get by slightly increasing local computing power versus what you can get by communicating with the cloud, the former would be quite limited in scale. Obviously, it would be different if you could incorporate an AI chip, but it is impossible to embed an AI chip in a smartphone handset because of the large form factor. Even with the PC, it would be quite difficult. Also, from a practical standpoint, even if you wanted to, you wouldn't be able to get a hold of AI chips. So my view is the potential from this is limited.

The only area where there could be an impact is if you had smartphones or PCs that had AI-like features, which might become a catalyst for replacement demand, given that we haven't seen much replacement demand recently. I do have some expectations for this kind of demand.

I do expect that the adoption of AI chips will have an impact on servers, and we are already seeing dramatic change. However, I don't see AI-enabled smartphones or PCs, driving major changes in the market. I don't have quantitative data.

With regard to NAND, I believe you were talking about a stacked NAND structure, which incorporates electrodes underneath. There is already a company that has started fabrication. I do think that this will drive some growth in NAND volumes, but my sense is the NAND market appears to be slightly running out of steam. Overall sales are currently up on production cuts that were implemented to boost prices. At this stage, there doesn't appear to be a strong driver for NAV. Where we are seeing strength is really only in leading-edge and logic.

[Atsushi Ikeda, analyst]

Understood, I believe the capacity utilization at your customers may be weak. Is your expectation that a clear recovery in this area will only come in 2025?

[Mayuki Hashimoto, executive]

Conditions at our customers are very widely, so there are some customers where utilization is very low, but others where utilization is very high. But even with those customers where utilization is high, demand is very strong for leading edge, but utilization for legacy nodes has yet to recover significantly.

If we look at the customers that are struggling, they appear to be losing share because they don't have products that are in high demand. There is a clear separation between those that fabricate leading-edge or HBM versus those that do not.

[Takayuki Komori, executive]

Next is Mr. Watabe of Morgan Stanley.

[Takato Watabe, analyst]

Have there been any changes in price trends or other revisions to the 300-millimeter LTA?

[Mayuki Hashimoto, executive]

There haven't been changes to LTA prices. But as discussed last time, we have allowed customers to extend deliveries. We have also allowed customers to change out legacy wafers for leading-edge wafers, However, when customers say that they have no room to store wafers, we have no choice but to accommodate request to delay delivery.

As a consequence of the delays, for contracts where there is a step-up feature, the impact of the higher price is also getting pushed out until we make delivery of those wafers. So for example, in an instance where the price increase was to have been 10%, the push-up comes later.

So the overall blended selling price may only rise several percentage points, effectively not rising as much as we initially expected. That said, we have made no changes to the stated LTA prices. I suspect our peers are doing the same thing.

With regard to volume, there isn't anything we can do. We cannot force our customers to take delivery. We are, therefore, accommodating our customers' wishes. No change from what we talked about last quarter.

[Takato Watabe, analyst]

With regard to the price increases that have been pushed out on the back of delayed deliveries, you expect to see those price increases come through in 2025 or '26?

[Mayuki Hashimoto, executive]

Yes.

[Takato Watabe, analyst]

Understood. With regard to your guidance, you have missed your forecast each quarter. I hope that you can ensure that your guidance is more accurate in the future.

[Mayuki Hashimoto, executive]

Unfortunately, forecasting is not easy in this industry. The only thing we can do is to respond quickly as circumstances change. It is the same as equity prices or ForEx rates. Because this is a very volatile industry, things do not always go to plan.

[Takato Watabe, analyst]

I would urge you to follow the consensus figures more closely.

[Mayuki Hashimoto, executive]

On that point, I don't believe the consensus figures are right.

[Takayuki Komori, executive]

Next is Mr. Nishiyama of Citigroup Securities.

[Yuta Nishiyama, analyst]

On Page 27, you indicate that Q3 sales-related variance will have a negative impact on OP on a sequential basis. Given that demand is recovering, I believe volume should be improving, even if only gradually. Also, SUMCO has a high share in high value-added leading-edge wafers and relatively speaking, the recovery in leading-edge is stronger.

With regard to prices, 3 months ago, you said you expected 300-millimeter LTA prices should start to rise from the September quarter. Also, with regard to 200-millimeter spot prices, you also indicated that you didn't expect prices to fall further. Given all of the above, why are you expecting sales variance to drop Q-on-Q in Q3? Can you elaborate more on this, please?

[Mayuki Hashimoto, executive]

You are asking about sales?

[Yuta Nishiyama, analyst]

No, I'm asking about the waterfall chart for Q-on-Q changes in operating profit, specifically the negative JPY 1.2 billion for sales-related variance.

[Mayuki Hashimoto, executive]

I will ask CFO, Kubozoe, to respond.

[Shinichi Kubozoe, executive]

As I mentioned earlier, there was a pull forward in 200-millimeter in Q2 related to the timing of customer deliveries, which pushed up sales volume in Q2. Because of the front-loading, we expect Q3 200-millimeter to be slightly lower. This accounts for the majority of this impact. So the numbers here do not reflect a sequential change in 300-millimeter.

[Yuta Nishiyama, analyst]

I see. In the presentation, I understood that you didn't expect to see much of a change in 200-millimeter from Q2 into Q3, so I hadn't expected that it would have much of an impact. With 300-millimeter volumes up and LTA prices trending up, if only slightly, are you saying that despite this, if you net everything out, you expect a negative impact from sales variance?

[Shinichi Kubozoe, executive]

Your understanding of the overall trend is correct, but as mentioned earlier, timing of deliveries and shipments does have an impact. Q2 was pushed off slightly by the timing of 200-millimeter shipments, but the effect of this pull forward is what is behind the expected negative impact in Q3.

In terms of the operating environment, the market for 200-millimeter is unchanged, and 300-millimeter will see a gradual recovery into Q3. So there isn't a negative impact related to 300-millimeter.

[Mayuki Hashimoto, executive]

Explaining this is very difficult, but there is inventory that is in transit by ship. There are times when the sea freight arrives early and acceptance happens earlier than expected, but other times when it arrives late. This actually does represent a relatively solid level of volume.

In Q2, the seafreight arrived early and was accepted early, resulting in a pull forward from Q3. Optically, although Q3 sales is down as a result, it is simply an issue of timing, so it isn't the case that we lost revenue. The decline in sales is not the result of a drop in 300-millimeter. 200-millimeter tends to be lumpy, which makes it a little difficult to follow, but the variance is typically within the margin of error.

[Takayuki Komori, executive]

Next is Mr. Yoshida of CLSA.

[Yu Yoshida, analyst]

My question is related to the previous question. If I look at the sales-related variance, as shown in the OP waterfall chart, comparing the Q2 Q-on-Q figure and the Q3 Q-on-Q guidance, previously you shared your image of the breakdown between volume and unit price. If possible, could you provide this breakdown?

Also, with regard to the LTIs, where you have agreed to extend delivery timing, I believe you initially expect to see some positive price impact from second half onwards. Can you comment on what has happened with this? If, for instance, the deliveries have been pushed out even further, when do you expect to see the impact of higher prices come through?

[Mayuki Hashimoto, executive]

So you are correct that there are deliveries that have been pushed out, which will materialize in second half, although the scale of volume is small. So there is a portion which will have an impact. If everything proceeds in line with plan, then there should be a meaningful improvement in the next fiscal year. But as I discussed earlier, it has become very difficult to forecast.

Similar to the fact that no one predicted a pandemic, we are seeing many choppy trends. So all we can do is to prepare multiple backup plans and to respond swiftly to what is happening on the ground. That is the secret to surviving in this industry.

I will share our forecast with you, but as you know, this is an unpredictable industry, and it is just our industry that is unpredictable either. This week saw a dramatic market move. I don't believe anyone here would have predicted the events of this week with a huge downward and upward swings.

Forecasting for our industry is the same in that it is difficult to predict. That said, normally, you would expect that there would be a pickup in 2025 with a positive impact from contracts with higher price points. But to be clear, many of the LTAs are 5-year contracts.

There are LTAs where the price is fixed for the life of the contract, but others where prices step up. I don't think I have said that prices on the step-up contracts are set to rise 10% a year.

However, there were customers that have committed to step-ups, but others where the price is fixed. So the impact of price increases is not necessarily 10%, although there have been some who have imagined that, that is the case. In terms of the overall impact, it isn't a straight increase across the board. Instead, the overall price point should be viewed as a weighted average.

[Yu Yoshida, analyst]

Can you provide some color on the sales-related variance in terms of volume and price?

[Shinichi Kubozoe, executive]

Almost all of the sales-related variance is the impact of volume. As we've mentioned earlier, there was some impact from adjustments to LTAs, but almost all of the negative impact is due to volume.

[Takayuki Komori, executive]

Next is Mr. Omura of UBS.

[Shunta Omura, analyst]

My question is also about sales-related variance. Can you tell me what your capacity utilization is currently? I believe sales-related variance reflects not only sales, but the impact of production utilization. For instance, if utilization is low, there would be an impact on sales-related variance. If we take the year-on-year waterfall chart on the left-hand side of the page for the 9-month forecast, how much of an impact is there from utilization?

Also, earlier, you mentioned that the greenfield investments would start to contribute in earnest from 2026, but you expect 2025 is likely to be tough. Can you also share your current outlook for utilization for the greenfield investments as well?

[Mayuki Hashimoto, executive]

First, our greenfield investments will complete in the first half of next year. Although depreciation has already begun, the full-scale impact will kick in from next year, even if the facilities have yet to become operations at that time, reflecting accounting practice.

In terms of when the new capacity will be fully utilized, I can share with you projection, but it is only a projection. I note that our original forecast had been for full utilization in the current fiscal year based on our customers' forecast at the ton. However, much has changed derailing our original plan. At this stage, we don't have a lot of clarity.

That said, we tapped the markets and formulated a number of backup plans as well as putting LTAs into place. Even so, although I can't say for sure, based on our experience to date, we should see a recovery in 2026, and our greenfield investments should be fully utilized.

Almost all of our greenfield investments are for leading-edge logic use wafers, so I expect that we can fill the new capacity in 2026. But depreciation will kick in from 2025, even though volume is unlikely to have fully recovered at that point. I think 2025 will be a tough year, but this is only a forecast. The operating environment has become very volatile, which makes forecasting very challenging.

[Shinichi Kubozoe, executive]

With regard to the first part of your question, we do not comment on utilization. With regard to the numbers on the right-hand side of the page, for the same period in the previous fiscal year, production levels were relatively high so utilization was running at relatively high levels.

What this means is that there would have been significant wafer volume in progress across the process. When sales fall off and production levels decline, we would be reducing work in progress inventory across the process in order to optimize, which would lead to a decline in overall volume.

So when we are operating at lower production levels, production-related losses tied to utilization tend to be large. This is a factor in the relatively large negative number we show here.

[Shunta Omura, analyst]

Understood. For Q3, you are not anticipating that there will be a significant improvement in production. Is that correct?

[Mayuki Hashimoto, executive]

Yes.

[Shinichi Kubozoe, executive]

The gap between the first half numbers and the 9-month numbers, apart from the impact of factoring in an additional quarter, reflects the continued low level of production for 200-millimeter and smaller diameters. 300-millimeter volume is gradually improving, but does not come through that strongly in Q3.

[Takayuki Komori, executive]

Understood. Next is Mr. Miyamoto of SMBC Nikko.

[Go Miyamoto, analyst]

I would like to ask about the competitive environment. First, with regard to leading-edge logic use wafers, as shown on Page 16, SUMCO has a majority share of the market, and I believe your main competitor is a Japanese player. But the third and fourth ranked players have been beefing up investments in epitaxial wafers. I would like to know if there is a possibility that the laggards could take share?

Secondly, on Page 17, with regard to HBM-used wafers, the leading HBM player is based in South Korea and has a well-positioned wafer maker within its larger corporate group. I believe that your Japanese peer is relatively well positioned as well.

Can you talk about your current position in HBM-used wafers? Do you expect to be able to take share going forward? Finally, you touched upon the Buy China policy earlier. You indicated that this was having an impact on 200-millimeter wafers, but is this policy also having an impact on 300-millimeter wafers?

Last year, on a full year basis, your sales to China accounted for 14% of total sales. How much do you think it will fall to this fiscal year?

[Mayuki Hashimoto, executive]

I can't say how much the share of China sales will fall to, but it is true that the Buy China policy has led to a decline in sales of silicon wafers to China. So likely that the China share of sales will drop below 14%.

We are monitoring 300-millimeter, but in many cases, it doesn't turn up in the statistics. It does appear that the Chinese players sell around 1 million wafers monthly, half that of chest wafers between 30% to 40% is prime and the remaining 20% or so is simple epitaxial. This is a rough image.

Even if yields are low, the Chinese chip makers are apparently buying from Chinese wafer makers without question. With regard to whether others can take share from SUMCO in leading-edge wafers, obviously, if we don't stay humble and focused, there is always the possibility that others may find opportunities to win share.

Currently, the customer is very busy, running at full capacity. I would question whether customers would choose to take the risk of taking on another supplier in such circumstances. Given that prices are already fixed under LTAs, I would be skeptical that a customer would feel the need to take risk by bringing on a different supplier. I think we are in a phase where barriers to entry are extremely high.

[Go Miyamoto, analyst]

Are you positioned in HBM-used wafers? And is there a potential to grow share in future?

[Mayuki Hashimoto, executive]

Recently, we have been able to develop a deeper relationship with the HBM player, although we didn't have much of a relationship in the past. We had some wafers that showed good performance for the customer. Obviously, given they have a subsidiary within the group, they will procure from the subsidiary, but I feel we are increasingly well positioned for HBM-used wafers.

[Takayuki Komori, executive]

Next is Mr. Okazaki of Nomura Securities.

[Shigeki Okazaki, analyst]

You have indicated that demand for leading edge is strong. Would it be possible to have the greenfield capacity ramp up normally? Are your existing leading-edge lines not yet at full utilization. Also in terms of depreciation for the greenfield capacity, you had previously indicated you expected a half-on-half increase of around JPY 10 billion. Is this still the case? If possible, could you provide an image of what you expect depreciation to be in 2025 and 2026.

[Mayuki Hashimoto, executive]

With regard to greenfield capacity, we still have room on our existing lives so there would be no need to front-load the ramping up of new capacity. However, our greenfield capacity is all equipped for leading edge.

So given the growth we are seeing in leading edge, I would expect we can fill the new capacity. I have more of a concern about the older existing capacity, which is being used for legacy product. Filling this capacity will be challenging for a while. I will hand over to CFO, Kubozoe, for the depreciation question.

[Shinichi Kubozoe, executive]

For this year, we continue to expect that depreciation will increase JPY 10 billion half-on-half into second half. First half was JPY 35.9 billion. So second half depreciation would be -- the first half amount plus roughly JPY 10 billion.

As we have been saying on the outlook for depreciation for 2025 and beyond, given that the greenfield capacity will come online next year, I would expect next year's depreciation to be around JPY 150 billion. The image for the following year would be flat to slightly higher year-on-year.

[Shigeki Okazaki, analyst]

So this fiscal year, likely to be JPY 85 billion on a full year basis and next fiscal year, around JPY 150 billion. Is that correct?

[Shinichi Kubozoe, executive]

For 2024, probably around JPY 83 billion or JPY 84 billion.

[Takayuki Komori, executive]

Thank you. Next is Mr. Yamada of Mizuho Securities. Mr. Yamana.

[Mikiya Yamada, analyst]

I would like to ask about wafer unit prices in Q3 using the presentation material.

Earlier, I understood your comments on LTA prices to be that while LTA prices were being respected, the volume levels are so low that the blended average price is not rising. If I am incorrect, please say so.

If that is the case, if we compare the chart on the right-hand side of Page 23, to the chart on the right-hand side of Page 27, it implies there is a year-on-year negative in sales-related variance for Q3 of JPY 9.4 billion. If we then look at the 300-millimeter wafer trend on Page 9, and assume that volume in Q3 is unchanged from Q2, it suggests that volume for Q3 would be largely unchanged year-on-year.

If we compare Q3 sales for the current year to the previous year, setting aside variance related to ForEx, then optically, sales appear to be largely unchanged. On this basis, the implied year-on-year Q3 negative sales-related variance appears to be too large, suggesting that prices have not risen at all. How should we think about this? Can you explain how all of these elements are consistent?

[Shinichi Kubozoe, executive]

Much of the difference between the first half figures and the 9-month forecast is explained by the decline in utilization for 200-millimeter as discussed earlier. On a year-on-year basis, there was a significant decline in utilization.

If you go back to Page 8, the decline in volumes for 200-millimeter will carry over through Q3 based on our assumptions that we won't see much of a recovery in Q3. This accounts for the vast majority of the figure you cited earlier.

[Mikiya Yamada, analyst]

If we look at Page 8, 200-millimeter wafers fell in Q3 2023, so extrapolating Q3 2024 volume to be largely unchanged Q-on-Q would not appear to result in a significant drop in volume. Are you saying that as a result of puts and takes in inventory and other factors, the 200-millimeter is the key factor behind the deterioration in sales-related variance, resulting in a more than JPY 9 billion year-on-year negative for OP in Q3. Is this the major factor?

[Shinichi Kubozoe, executive]

Yes.

[Mikiya Yamada, analyst]

If that's the case, that means that prices will not rise much in Q3. In other words, even at the end of Q3, does that mean that Q3 volume is effectively just the wafers that were pushed out last year?

[Mayuki Hashimoto, executive]

That is in fact the case.

[Mikiya Yamada, analyst]

So demand is really weak.

[Mayuki Hashimoto, executive]

Yes, demand is weak. Although the market has bottomed, we haven't seen a strong recovery.

[Mikiya Yamada, analyst]

I see.

[Mayuki Hashimoto, executive]

Unfortunately, that is the case. In particular, customer capacity utilization for 28-nanometer is only around 70%. Also, with regard to leading edge, there were variety of contract types.

The terms of the contract set out a fixed number of wafers for delivery each month, but there is a breakdown of wafer types within the specified volume. Prices are different for each type of wafer with leading edge to the highest and commodity the cheapest. The combination of wafer types has implications in terms of contribution to higher blended average price.

The contracts are quite complex. Also, a significant proportion of the LTAs are fixed price with no change over the life of the contract. There are contracts with step-up features, but many market participants have focused only on these contracts, suggesting that there is a 10% increase in price every year. That's not how I explain the contracts, but maybe my comments were misinterpreted. A lot of the LTAs are fixed price for 5 years. That's particularly true for polished wafers. For more technologically advanced wafers, we have asked for higher prices.

Customers have agreed to rising prices for leading edge, but much of this has gotten pushed out. Also, overall volumes are impacted by the decline in legacy wafers. So we don't have a sense that there has been a push-up effect from prices.

As well, although it isn't the case for SUMCO, our affiliate has seen some price erosion on 200-millimeter. So if you put all of this together, overall, it doesn't look like prices have risen, and they actually have not risen. On a blended basis, it's probably awash.

[Mikiya Yamada, analyst]

Understood. I had expected to see more of a push up to prices given the high proportion of leading-edge wafers for you.

[Mayuki Hashimoto, executive]

I also had similar expectations. It appears that customers can't really increase capacity.

[Takayuki Komori, executive]

Thank you for your questions. We will end the meeting here. Thank you to everyone for joining the Q2 fiscal 2024 results briefing. We are grateful for your participation today.

[Mayuki Hashimoto, executive]

Thank you.

[Shinichi Kubozoe, executive]

Thank you.