Sumco Corp

TSE:3436

Earnings Call Transcript 2021-Q1

[Hiroshi Shibaya, executive]

I am Hiroshi Shibaya of Sumco Corporation. Thank you for your participation today and for your continued support. This is the results briefing for first quarter 2021.

Before starting the presentation, allow me to confirm today's materials, which consist of 4 items: the brief statement on consolidated financial results for Q1 fiscal year 2021; the presentation deck entitled results for first quarter of fiscal 2021; the announcement concerning disparity between forecast and actual figures for the first quarter of fiscal 2021; and the announcement regarding revision to dividend forecast. The materials are available for download on our website at sumcosi.com.

We will have presentations today from Representative Director, Chairman and CEO, Mayuki Hashimoto; and Representative Director, Vice Chairman and CFO, Michiharu Takii. Hiroshi Ito, General Manager of Accounting, is also on hand. We have set aside 60 minutes for the briefing. The briefing will end at 5 p.m.

Finally, a disclaimer. The estimates, expectations, forecasts and other future information discussed here 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.

Chairman and CEO, Hashimoto, will discuss our forecast and operating environment, to be followed by an explanation of the earnings results by Vice Chairman Takii. 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. Please turn to Slide 5 of the presentation. This is the overview of our Q1 results. Relative to our forecast, sales were largely in line. Operating income exceeded plan by JPY 0.8 billion, ordinary income by JPY 1.4 billion and net income by JPY 1.8 billion.

Moving on to the forecast for second quarter, we are guiding for sales of JPY 81 billion, operating income of JPY 11.5 billion, ordinary income of JPY 11 billion and net income of JPY 8 billion. Our ForEx assumption is JPY 109 to the dollar.

Next, on to Slide 6 and shareholder returns. We are guiding for an interim dividend of JPY 16 for a dividend payout ratio of 30.4%. The year-end dividend is to be decided.

Please turn to Slide 7 for a discussion of the market environment for silicon wafers. Starting with the Q1 results, demand for 300-millimeter from logic was very strong. Demand was so strong that supply was not able to keep up. In our case in particular, half of our 300-millimeter capacity was off-line as a result of periodic maintenance, reducing our production levels and leading to acute shortages. On the memory side, DRAM has recovered. Customer inventory is getting close to appropriate levels. In 200-millimeter and smaller wafers, there was a sharp recovery in demand, primarily in automotive and consumer electronics applications. Supply-demand is extremely tight.

On prices, in hindsight, if we had known in 2020 that demand would tighten, we would have held out a little more. First half 2021 prices were largely set last year, so we haven't seen a rise in prices. The majority of prices on memory LTAs are unchanged from 2018 levels. We did see some pushouts in terms of delivery timing, but the LTA price points have been maintained. Spot prices were flat Q-on-Q as expected. This reflects the fact that LTAs include some contracts that are 6 or 12 months, as well as the fact that spot prices have stopped falling. Actually, spot prices on 300-millimeter epitaxial wafers and 200-millimeter are already starting to rise.

On the outlook for Q2 2021, we expect 300-millimeter logic demand will continue to expand. Conditions will tighten further, although we are already supplying wafers on an allocation basis. There won't be any periodic maintenance in Q2, but even so, we do not expect to be able to get close to keeping up with demand.

In memory, following the lead of DRAM, we expect NAND will start to show signs of recovery in Q2. For 200-millimeter and smaller wafers, demand remains very strong. We rushed to rehire staff to boost capacity, but even so, supply is not keeping up with demand. As we had previously reduced headcount for 200-millimeter, we have been unable to run at full capacity, but we are now virtually back to full capacity production. In spite of this, demand is outstripping supply.

On pricing, it appears price hikes for 300-millimeter logic wafers are starting to gain traction in the spot market for second half 2021. Near-term prices, of course, were already set last year and can't be changed suddenly. In 200-millimeter as well, there appears to be increased willingness to accept higher prices for second half 2021. In the longer term, 300-millimeter should continue to benefit from very strong 5G smartphone and data center demand. We expect shortages for logic to persist. On the memory side, we expect to see a tightening, supported by the recovery in demand.

Sumco is reaching the limits of brownfield capacity expansion. I believe our peers are similarly placed. If we try to fulfill our supply obligations to our customers by increasing supply beyond current levels, it will require greenfield expansion. We have already started looking at greenfield investment. Obviously, the economics do not work at all at current price levels, so any investment will depend on price trends going forward.

For 200-millimeter, the only way to increase supply volumes for us as well as our peers is to try to eke out more productivity gains. Given this, we expect strong demand and market tightness to persist over the longer term, at least to the end of 2022.

Please turn to Slide 8. This is the wafer trend for 200-millimeter by quarter, which we share every quarter. Generally speaking, we estimate the full capacity production for the industry in 200-millimeter wafers is around 5.7 million wafers per month. Everyone is running at maximum capacity and is supplying wafers on an allocation basis. This is the same chart for 300-millimeter. Here again, there was a significant increase, with industry demand at close to 7 million wafers near term. It does appear that there is sufficient capacity to accommodate this level, so no surprise. However, given brownfield expansion capacity has been fully tapped, the industry will not be able to immediately increase capacity in excess of this level.

If you look at this chart, you can see that in 2019 and 2020, demand was at a plateau. 2019 is the line with the blue Xs, and 2020 is the blue diamond. Although demand was at a plateau during this period, you can see that it started to pick up in the second half of 2020 as the recovery gained momentum. You can see here where we are with 2021. There are certainly positive factors like 5G, but PCs in particular have broken out of what was viewed as a slow erosion. PC demand has not only stopped falling but is actually on the rise on the prolonged impact of working from home and sheltering in place. PC consumption is supporting strong CPU production levels. Strong demand is also coming from games as well. All of this contributes to strong logic demand, which underpins the trend in the chart.

Slide 10 shows, as always, our estimates for customer inventory levels in 300 millimeter. You can see the sharp decline from the start of the calendar year with overall inventory levels at around 1.3 months. We have provided breakouts of this chart into logic and memory, as shown on Slide 11.

If you look at logic, it is now down to 1.11 month from the peak level of 1.68 months. You can also clearly see the imbalance between wafer input levels and purchase levels in March. This reflects the strong shortage of wafers. All of our customers are demanding more volume, but capacity is fixed, which regrettably makes it challenging to increase supply volumes. In contrast, our memory customers appear to have some breathing space on inventory.

Please turn to Slide 13. I would like to explain bit growth by application. I will start with how we are thinking about NAND use polished wafers. To think about bit growth, please first look at the bar chart on the left. You can see that mobile handsets and smartphones are the key drivers and that NAND bit growth is likely to continue to be slightly higher than 30%.

On 3D technology, we continue to see progress. The mainstays in 2021 will be 96 and 128 layers, but this will shift to 128 and 176 layers in 2022. On the upper right, we show trends in 3D NAND mass production technology by company. If we look at bit growth that results from the increase in layers, you can see that the impact of chip shrink largely offsets the impact of expected bit growth. By implication, I believe NAND demand for polished wafers is unlikely to grow significantly going forward.

Turning to Slide 14. We look at bit growth for DRAM. DRAM bit growth is estimated at a CAGR of 18%. However, in the case of DRAM, there is a limit to what can be achieved through scaling. The process moves in increments of 1 nanometer. Given this, the impact of chip shrink in offsetting expected bit growth is at most 10%. Relative to the 18% CAGR, this leaves you with 8% net bit growth. This is our view.

When we look at the all-important logic, we expect to see continued tightness in epitaxial wafers. If you look at Slide 15 and progress on scaling in logic, the rising tightness in wafer requirements is driving demand for further scaling. But in the case of logic, the wafer savings generated by chip shrink have been utilized to enhance functionality. Effectively, wafer die sizes have not changed. In 2018, it was 83 square millimeters; 2019, it was 98 square millimeters; and in 2021, 88 square millimeters.

In general, wafer consumption for logic has remained largely unchanged despite progress on chip shrink. However, the number of logic chips has continued to grow. This is true on a per capita basis, but the growth is further boosted by population growth, which is what is driving the curve shown here. So we expect logic growth to be around 13%. In the case of logic, the correlation to wafers is 1:1. We expect this trend to remain unchanged going forward.

I would now like to discuss our expectations for 300-millimeter wafers by application. Please turn to Slide 17. For smartphones, we do not expect wafer growth to be driven by NAND. However, logic and DRAM demand for wafers should increase, as you can see on the bar chart. We continue to see solid handset unit volume growth on the migration to 5G, which is the blue part of the bar. Both DRAM and logic demand for wafers should grow. But for NAND, the increase in layers will temper wafer consumption.

Next, let us look at PCs and tablet on Slide 18. The trend here has actually changed. The expectation for PCs and tablets to date was that unit volumes were in long term decline. However, in fact, the declines have halted, and we have been seeing relatively firm replacement demand for a solidly flat trend. If we then look at data center demand for 300-millimeter wafers on Slide 19, we continue to see aggressive investment in cloud facilities, which will drive both logic and DRAM. In the case of data centers, DRAM growth is particularly strong.

Finally, I will talk about our view of the supply-demand balance for 300 millimeters. Our view is that we will see acute shortages of 300-millimeter wafers in second half 2022. The shortages will only get worse in 2023 because although it takes 12 to 18 months to ramp new brownfield capacity, it takes 2 years plus to ramp greenfield capacity. In our case, even if we were to make an investment decision now, the capacity would only come online in 2024. So our only recourse for 2023 would be to increase output by improving productivity and tapping into what little brownfield capacity is left at this stage. Based on this, we believe there will be severe shortages in 2023. Obviously, we will do our best to fulfill our obligations as a supplier. But if we look at our peers, it appears that there isn't that much more that can be done through brownfield expansion. So we really have no choice but to build new facilities, which probably means a lengthy lead time.

This completes my remarks. I will now hand over to Vice Chairman Takii to talk about the earnings results in more detail.

[Michiharu Takii, executive]

I am Vice Chairman Takii. I will present the Q1 results. Please turn to Slide 23, where we show the Q1 results. We show both the year-on-year and Q-on-Q comparison. Sales were up on both a year-on-year and Q-on-Q basis. However, OP was down on a year-on-year basis but up Q-on-Q. At the nonoperating level, there were one-offs both last fiscal year and this fiscal year. Last year in Q1, we received a subsidy on investments in the Nagasaki Plant. In Q1 this year, we received a subsidy for the Imari Plant for energy saving initiatives of JPY 0.45 billion, for a net nonoperating gain this fiscal year. Ordinary income showed the same trend as OP. Net profit declined year-on-year but improved Q-on-Q.

Moving to CapEx. There was a concentration of CapEx in Q1 last year, leading to the year-on-year decline this year. Typically, CapEx is around the JPY 10 billion level, so last year's Q1 level is not really useful as a reference point. Depreciation reflects the impact of the start of a new fiscal year at JPY 11 billion. The resulting EBITDA level is down slightly on both a year-on-year and Q-on-Q basis. The OPM and EBITDA margins are as shown here. I will discuss this in more detail later, but the decline in depreciation on a Q-on-Q basis was a major contributor to the Q-on-Q improvement in profits.

Please turn to Slide 24. This is the waterfall chart for a sequential comparison of Q1 results. Sales rose JPY 3.3 billion Q-on-Q, while OP rose JPY 1.2 billion. This was largely in line with our forecast. Costs increased JPY 0.7 billion, consisting of increases in unit prices on materials, maintenance expenses and labor costs on the back of an increase in head count. Depreciation decreased JPY 1.9 billion on the transition into a new fiscal year. There was a positive contribution of JPY 0.2 billion from sales related variance and others. As I noted last quarter, this includes the negative impact of price declines of around JPY 1 billion, which was offset by a positive contribution from production and sales of slightly more than JPY 1 billion. The reason for the relatively small positive contribution from volume, despite solid sales, reflects higher levels of shipments from inventory rather than production during the quarter, as discussed last quarter. The small negative in terms of ForEx impact is related to Taiwan.

Please turn to the next page, Slide 25. This is the waterfall chart for year-on-year changes to Q1 operating income. The background to the year-on-year changes is similar to what was discussed previously. Costs increased JPY 1.2 billion, with increases of around JPY 0.5 billion each for labor and maintenance costs. Also, as a result of the pandemic, shipping costs were up slightly year-on-year. Depreciation was outweighed by CapEx, increasing JPY 1.5 billion. There was a positive contribution of JPY 2.2 billion from sales-related variance. While sales volume was up on a year-on-year basis, the proportion of LTAs declined slightly, leading to a lower unit price, for a negative impact of around JPY 1 billion. In addition, there were declines in spot prices in Taiwan as well as lower prices for the smaller diameter wafers produced at the Taiwanese plant for a total negative impact of JPY 2 billion within sales-related variance. This was offset by a JPY 4 billion-plus positive on higher volumes.

Please turn to Page 26. This is the balance sheet. Cash and cash equivalents increased JPY 6 billion. Product inventory and work in progress was largely unchanged. Raw materials and supplies fell JPY 1.8 billion. Of this, JPY 2.8 billion comes from declines in polysilicon inventory. However, this was offset by increases in some supply inventory as a part of efforts to enhance BCP preparedness. Interest-bearing debt declined JPY 2.2 billion. Under the capital account, the increase in retained earnings reflects the contribution of profits net of dividend payments. Under other, there was an increase of JPY 4.6 billion related to a ForEx translation adjustment as a result of a weaker yen as of the end of March. All of the financial metrics shown at the bottom of the table improved versus the end of December.

Slide 27 looks at the cash flow statement. After adding back depreciation expense to the pretax profits discussed previously and adjusting for the decline in inventories, operating cash flow was JPY 24.1 billion. After taking into account the outflow from investing cash flow, free cash flow was a positive JPY 6.6 billion. Although there were outflows for dividend payment and the repurchase of shares, this was offset by gains on ForEx translation adjustments. We also note that there was an outflow of JPY 6.3 billion under others within investing cash flow. This is actually not related to CapEx, but reflects an outflow into a term deposit. This is included under investing cash flow rather than finance cash flow. As a result, cash and cash equivalents increased by JPY 6 billion.

Next, I will discuss our forecast for Q2 2021. Please turn to Slide 29. As discussed by Chairman Hashimoto at the outset, we are projecting sales of JPY 81 billion and OP of JPY 11.5 billion for Q2, both up Q-on-Q. We expect net income of JPY 8 billion. For first half on a year-on-year basis, we expect sales to increase, but OP to decline.

Please turn to Page 30. This slide breaks out the major components of change for the second quarter forecast on a Q-on-Q basis. We expect sales to rise JPY 5.1 billion and OP to improve JPY 2.2 billion. We have seen a weakening of the yen and are assuming a dollar-yen rate of JPY 109 to the dollar. The forecast of a JPY 5.1 billion increase in sales includes the ForEx impact. This is the waterfall chart which breaks out the major components of the projected JPY 2.2 billion Q-on-Q improvement to second quarter operating income. We expect costs to increase slightly Q-on-Q. This includes a roughly equal increase in both labor and electric power costs of around JPY 0.2 billion to JPY 0.3 billion, contributing to the JPY 1 billion increase.

Depreciation is expected to increase by JPY 1.6 billion on the ramp-up of new facilities. The Q-on-Q change in sales related variance does not include an increase in prices. All of the increase comes from higher volumes for a positive contribution of JPY 3.7 billion. The ForEx impact reflects the impact of a weaker yen, as touched upon earlier. This gets us to the JPY 11.5 billion operating income forecast, up JPY 2.2 billion Q-on-Q.

Please turn to Page 31. This is the year-on-year waterfall chart for first half OP. We expect a JPY 9.8 billion year-on-year increase in sales, but a JPY 2.3 billion decline in OP. In terms of the components of the JPY 2.3 billion year-on-year decline in OP, first, we expect cost to increase JPY 3 billion. We have been stepping up maintenance, so expect maintenance costs to increase by slightly more than JPY 1 billion. On labor costs, we have increased contract workers as well as overtime for an expected year-on-year increase of slightly less than JPY 1 billion. We also expect increases in materials cost and shipping expenses as a result of the pandemic, for an overall combined increase of JPY 3 billion. Depreciation is expected to increase from last year.

On sales related variance, we are expecting a positive impact of JPY 5.7 billion year-on-year. This factors in the negative impact from lower prices on a year-on-year basis of around JPY 4 billion. Japan accounts for around JPY 3 billion of this JPY 4 billion as a result of a decline in the proportion of 300-millimeter LTAs and higher spot prices, for a mix impact of around JPY 2 billion. The remaining negative impact is due to lower spot prices on 200-millimeter and smaller diameters. We also expect a negative impact from lower prices of JPY 1 billion from Taiwan. However, we expect a positive of JPY 10 billion from the increase in volumes on the back of higher revenues. Relative to last year, the yen is still stronger on a year-on-year basis.

Next, I will comment briefly on some of the reference slides. You can see on Slide 33 that sales is gradually recovering, reflecting the higher production volumes. On the back of this, OP is also improving.

Slide 34 shows quarterly trends for EBITDA. This slide shows a similar trend to the previous slide. Last year, we saw a slight temporary deterioration in EBITDA, but we are guiding for an EBITDA margin of 30%.

Slide 35 shows the regional breakout for sales. As you can clearly see, the yellow part of the bars is increasing.

This completes my presentation.

[Hiroshi Shibaya, executive]

Thank you. We will now open the floor to questions. We ask you to kindly limit yourself to one question each. We will start with Mr. Enomoto of Bank of America Securities.

[Takashi Enomoto, analyst]

I have one question. You have indicated that you have started considering greenfield investment. Can you comment on what sort of specific requests you have received from your customers? If you do go ahead, what is the potential scale you are considering? What kind of price increase would be necessary to undertake such an investment? Also, if possible, could you comment on whether or not you already have contracts in place that are predicated on this investment?

[Unknown Executive]

First of all, with regard to greenfield investments, almost all of the major customers have expressed strong demand for more volume. However, we have said that we cannot increase volumes with the current capacity. It is not the case that we are particularly hung up on greenfield investments, but instead that if we are to increase capacity, the only choice is greenfield. At this stage, we have not engaged with our customers on volume. Any new capacity will only come online in 2024 or later, which is a long way away. It will be subject to the forward forecast of our customers, although some of the largest customers do already have numbers. In terms of engaging with customers on contracts, that is still to come.

The first thing we need to do is to establish the economics of a greenfield investment. Looking at the ceiling, I suggested that we would need a price increase of 50% to 60%, but there is also the question of whether customers will accept an increase of 50% or 60%. All of this needs to be taken into consideration. Obviously, we recognize that as a supplier, we do have supply obligations, but of course, we see absolutely no need for us to be out-of-pocket in order to fulfill this obligation. We are a for-profit organization with shareholders. As such, we will need to negotiate well in order to make an investment decision.

In terms of volume scale, given that it will be greenfield, which means we will be building a physical shell, we will be looking to build a large physical shell as would be normal, which we can then gradually populate with equipment over time. We do not want to repeat the experience of more than 10 years ago where everyone invested aggressively and doubled capacity. This led to a 10-year period where we had to endure having our customers beat down prices. We want to avoid this at all costs, so we are being very cautious. Does this answer your question?

[Takashi Enomoto, analyst]

Yes. I would just like to follow-up on one point. If you do go ahead, where would you consider as likely sites?

[Unknown Executive]

Potential sites would be Japan and Taiwan. Taiwan is a possibility.

[Hiroshi Shibaya, executive]

The next question is from Mr. [ Ikeda ] of Citigroup Securities.

[Unknown Analyst]

This follows on from the previous question. In anticipation of greenfield investment, when do you think we will see pricing resets on LTAs? In other words, in 2021, 70% of your business is LTAs, and I would expect the proportion will go down in 2022. When should we expect to see a reset on prices? Also, with regard to the need for a price increase of 50% to 60%, I believe that with previous brownfield investments, you were ultimately able to achieve price increases of 40% to 50%, but I would like to confirm that you are looking to increase prices by a further 50% to 60%.

[Unknown Executive]

Actually, after the increases you are referring to, prices actually declined in 2020. We did not choose to lower prices, but spot prices declined. Non-LTA prices also declined. Prices are now rising again, but what I am saying is that we would need to see an increase of 50% to 60% from our current selling prices in order for a greenfield investment to be viable.

[Unknown Analyst]

You are talking about the average selling price?

[Unknown Executive]

Yes.

[Unknown Analyst]

How should we think about the time frame for pricing resets? Will you be able to do some in 2022? Or will it be 2023, given that you expect 2023 to be very tight?

[Unknown Executive]

It will probably be from 2023. While I expect the shortages will be very acute in 2022, given current trends, we cannot change LTA prices mid-contract. That said, our LTA customers are clamoring for more volume. So the discussion with our customers will be about the incremental portion of volume and how they are thinking about 2024 and beyond. I would expect to see a rapid surge in prices in 2023. For spot, prices are likely to rise strongly in 2022 with the uptrend starting in the second half of 2021. Wafers are already in short supply with customers competing to lock in supply. As a result, customers have started to say that they are not concerned about prices, but they need to secure volume.

[Unknown Analyst]

So in 2022, will the proportion of LTAs decline significantly from 70% to 60%?

[Unknown Executive]

No, it won't fall that much. It should be similar to current levels.

[Hiroshi Shibaya, executive]

Next is Mr. Watanabe of Mitsubishi UFJ Morgan Stanley Securities.

[渡邉 亮一, analyst]

Can you expand on customer wafer inventory levels? Earlier, you referred to memory customers choosing to push out the timing of deliveries. Why did this happen? Also, in March, there were shortages of logic wafers. If that was the case, why were wafer purchase volumes down?

[Unknown Executive]

On the pushout of wafer deliveries to memory customers, there was a significant increase in customer inventories, particularly last year. In response to this, we could have chosen to be confrontational and aggressively demand that our customers respect contractual volume levels. Given that the customers would have no place to store such inventory, we did not feel that this was a sensible approach. Instead, we chose to engage with our customers. We asked that they respect the aggregate volume levels for the life of the contract, but indicated that we would be willing to allow them to make slight adjustments to the delivery schedule. All of our peers did the same thing.

On the logic side, I think our periodic maintenance may have been a contributing factor. Epitaxial accounts were a very high proportion at Sumco. So I believe our periodic maintenance led to a decline in supply in March. Roughly speaking, half of our capacity was off-line in February and March for periodic maintenance. As a result, quarterly production volume was off by about 30%. So I think this probably had a significant impact. For us as well, the timing was unfortunate.

[渡邉 亮一, analyst]

With regard to memory customer inventories, is it possible that this could go back to the levels seen at the end of December? How much customer demand is there?

[Unknown Executive]

Memory customers are not that desperate for wafers at this time. However, they are aiming to maintain the current level of inventory, so wafer input levels are up slightly. It appears that they are gradually increasing input levels. I would expect that polished wafer demand picks up into second half 2021 and 2022. That said, we are seeing more use of polished wafers as a substrate for epitaxial wafers. It is true for us as well. Consumption of polished wafers as epi substrates is rising, so production levels are falling. I would expect to see a gradual tightening. DRAM activity levels are relatively firm. We don't distinguish between DRAM and NAND, but we have the data for each of the individual players, although we cannot disclose this information. If you look at the individual company trends, you can see clear differences in the trends for each, depending on whether DRAM is strong or NAND is strong.

[Hiroshi Shibaya, executive]

The next question is from Mr. Watabe of Morgan Stanley MUFG.

[Takato Watabe, analyst]

Earlier, you indicated that industry capacity was close to JPY 7 million for 300-millimeter wafers and JPY 5.7 million for 200-millimeter. I have heard talk that shipments for 300-millimeter are running at JPY 7.2 million, and JPY 6.4 million for 200-millimeters. Is the difference a reflection of sales from inventory? Or is it possible that there is some capacity that you may have missed?

[Unknown Executive]

We don't have a lot of clarity on 200-millimeter capacity in China. For instance, Ferrotec and others make wafers, although some suggest that in terms of quality, they aren't able to even match the requirements for prime wafers yet. It isn't clear whether this capacity is all being included in LTAs. At a minimum, we have good visibility for the capacity of the 5 majors. On that basis, at JPY 5.7 million, I believe all of the majors are running at maximum capacity with no room to expand. My view would be that the JPY 6.4 million figure includes sales from inventory. In the first half of last year, we made significant changes to our workforce and reduced our production slightly in response to market conditions. My view would be that there is significant inventory. Even so, the numbers still don't add up. So that suggests that local Chinese players may be shipping some product, although it is probably only test or monitor wafers.

[Takato Watabe, analyst]

In other words, even if the JPY 6 million figure is valid, you feel that spot prices should improve going forward. Is that correct? Can you comment on 300-millimeter capacity as well?

[Unknown Executive]

We have good visibility into 300-millimeter capacity. Current capacity is JPY 7.2 million or JPY 7.3 million. There really isn't much more room to expand on a brownfield basis.

[Hiroshi Shibaya, executive]

The next question is from Mr. Yamada of Mizuho Securities.

[Mikiya Yamada, analyst]

You have talked about wafer supply demand, but with regard to logic for data center use. Data centers will clearly require high performance, low power consumption semiconductors. On Slide 19, I understand why NAND will not increase, but I would have thought that logic would grow more. If that were to happen, would there be more cannibalization of polished wafers? Do you think that there is a possibility that this leads to overall shortages in the first half of 2023 as a result? Also, is it still too early to expect higher memory densities in DRAM or logic as a result of 3D technologies?

[Unknown Executive]

It is true that if you look at polished wafers in isolation, it would appear as if there is still more capacity, but everyone is now trying to use polished wafers as epi substrates. However, in order to do this, you need to have epitaxial furnaces. So there are a lot of orders for epi furnaces. Given that overall capacity cannot be increased, if you try to increase epitaxial wafer output, you end up cannibalizing polished wafers. If you look at demand for polished wafers, we are now seeing a recovery. So as you have suggested, it seems likely we will see overall shortages in the second half of 2022.

At the moment, everyone is only talking about shortages in logic. But given that the lead time for epitaxial furnaces is around 12 months, it would be possible to produce epitaxial wafers using polished wafers as substrates within 18 months, which may happen as early as the end of this year. Even with this, however, it won't be enough to offset the shortages in epitaxial. But if this happens, then we would have a situation where it won't be possible to expand polished wafer capacity. On that basis, I would concur that we could see overall shortages in the second half of 2022.

On DRAM, it appears 3D technologies are very challenging. It may be technologically possible but unlikely to be immediately viable, and the pace is unlikely to match the pace we have seen with NAND.

[Mikiya Yamada, analyst]

Understood. With the integration of GlobalWafers and Siltronic, in the short term, I believe that GlobalWafers will be able to increase its epitaxial output by leveraging Siltronic's expertise, which will lead to an overall shortage. However, in the longer term, although I think that there may be some risk that competition at the leading edge heats up, I think that this combination will accelerate the trends we have been talking about and will contribute to an overall shortage. What do you think? Even if that happens, do you still believe the prices, with the exception of some spot prices, will not rise next year?

[Unknown Executive]

With regard to the impact of the integration, the decrease in the number of players from 4 to 3 means there will be one less player that we need to monitor. However, on the customer side, there will be some that will see their share with the new entity rise significantly. Normally, a customer would not choose to source 60% of its wafers from a single player. The maximum is typically 40%, so there will be some share that will be up for grabs. In addition, senior management throughout the industry have a deep understanding of how a strategy to win share by lowering prices means everyone loses. I think we won't see irrational behavior, as a result.

[Mikiya Yamada, analyst]

Even so, you do not expect to see LTA prices to reset in 2022?

[Unknown Executive]

The customers are probably pleased that they were able to sign LTAs at good price levels. Price resets before LTAs expire are very difficult. At the same time, increasing volume is also tough. Negotiations will focus on how to think about pricing in conjunction with the need to increase volumes in 2023 and 2024. For instance, if a customer wants more volume in 2024, you could argue that it wouldn't be possible unless the customer agrees to higher prices in 2023. But if you have some customers that agree to this but others that don't, the customer that does not agree may see their volumes in 2024 drop off significantly. At the moment, everyone is competing to lock in supply. So leverage plays a factor. Of course, there are many different types of LTAs, some that are set at market prices, some where only volume is fixed and price negotiated each time. But basically speaking, greenfield pricing will only definitely kick in from 2024.

That said, I would expect to see rumblings emerge in 2023. The question is what type of behavior we will see from customers after they go through a period of acute shortages in 2022. We think semiconductor growth is a very positive thing and that this is technology that is necessary for mankind. We would like to fulfill our obligations as a supplier, but there is no need for us to lose money in fulfilling our obligations. So it will hinge on the discussions we have with our customers going forward.

[Hiroshi Shibaya, executive]

Next is Mr. Okazaki of Nomura.

[Shigeki Okazaki, analyst]

Will we only see the output from the greenfield investment for 300-millimeter in 2024? I was under the impression that customers are in more of a hurry, but if you could comment on this, that would be great. Also, I believe that costs will increase. Can you talk about your expectation for returns, or how long it will take to recover your investment?

[Unknown Executive]

First, with regard to the start of output in 2024, we are still at the early stage of the process. We haven't made a decision on capacity yet. However, we will definitely need to build a physical shell, so it will likely be 2024 at the earliest. In the past, I have said that it would likely take 2.5 years, so first half 2024 would be a little challenging.

[Shigeki Okazaki, analyst]

In terms of what you are hearing from your customers, is it possible that you will make a decision soon?

[Unknown Executive]

Well, we probably need to make a decision by the end of the year. However, in terms of customer commitment, it isn't necessarily a yes, no decision at this stage. It should be possible to move forward even if customers haven't yet fully committed at the outset. As the process moves forward and becomes more concrete, we would look for more clarity from the customers. Otherwise, needing to have everything set in stone to make a decision will push back the timing significantly, which would be bad for customers. So we will be making decisions as the process moves forward. The customers are aware of the shortages and have already asked what they can do. We are focused on moving quickly and aim to place orders early for equipment and facilities with long lead times. We will focus on getting the physical shell in place and position ourselves so that we can increase capacity incrementally over time.

[Shigeki Okazaki, analyst]

How about expected return levels or time to recover your investment?

[Unknown Executive]

That is something we will be looking at going forward. We first need to estimate how much it will cost and then engage with our customers to get a sense for return levels. We don't have a specific ROE or IRR target, although we do have a general sense of levels, and we will look at it at the appropriate time. There are many things that must be taken into account. So it isn't just simply setting an IRR hurdle rate.

[Shigeki Okazaki, analyst]

You don't have the sense that we could see a recurrence of the oversupply situation that we saw 10-plus years ago?

[Unknown Executive]

No, I don't.

[Hiroshi Shibaya, executive]

The final question is from Mr. Azuma of Jefferies Japan.

[Yoshihiro Azuma, analyst]

The questions today have focused on the wafer price level necessary for committing to greenfield investments, but at a basic level, I believe there are differences in demand between polished wafers and leading-edge epi. With regard to leading-edge epi, I would think you would need to have agreement with the customers on the inspection setup or the level of quality warranties that the supplier commits to before agreeing to a price level. What do you think?

[Unknown Executive]

Yes, that is right. When you talk about epi, there isn't a single price point because there isn't a standard epi wafer. There are different price points for differing design rules. With design rules such as 5-nanometer, 10-nanometer, 28-nanometer, 40-nanometer or 90-nanometer, obviously, the higher the degree of difficulty, the higher the price point. This is also a factor in deciding the equipment requirements. Depending on the anticipated product mix, the investment amount will be completely different. So it isn't that simple. So as you have said, it will hinge, for instance, on the proportion of 5-nanometer. If the plan is to fabricate for 5-nanometer, the equipment required is different.

This may be too specialized of a discussion, but you absolutely need SP7s for leading-edge wafers, but not for mid-level grade product. But SP7 are extremely expensive. Even a single unit can drive up the investment amount significantly. So you are correct that the required price level will depend on the expected product mix, and this will also impact returns. Our policy is not to invest in expanding capacity for commoditized products, so we only invest for leading edge. On the SP7, I believe the next-gen product is the SP9. When you get to the SP9, it isn't just the footprint, but you also need to take into consideration the underlying bedrock.

[Yoshihiro Azuma, analyst]

This also needs to be factored into the structure of the physical shell as well or you may not be able to install?

[Unknown Executive]

Yes, you need to build a physical structure with full seismic isolation.

[Yoshihiro Azuma, analyst]

If that is so, then in fact, it isn't as simple as saying that you will just build a big box?

[Unknown Executive]

Right. So we will be building a physical structure to cutting-edge specs from the outset.

[Statements in English on this transcript were spoken by an interpreter present on the live call.]