6 December 2016 Asia Pacific/Japan Equity Research Semiconductor Equipment

Technology sector Research Analysts COMMENT Hideyuki Maekawa 81 3 4550 9723 [email protected] Asia feedback (semiconductor/SPE): NAND Akinori Kanemoto scarcity 81 3 4550 7363 [email protected] ■ Summary: As noted in our latest hardware outlook based on our Asia Market th Mika Nishimura Survey (Asia Market Survey (Hardware)) published on Dec 6 , smartphone 81 3 4550 7369 production remains strong, thereby reducing the risk of inventory corrections in [email protected] excess of normal seasonality in 1Q 2017. Supply chain inventories do not appear to be at risk of adjustment, allaying our previous concerns about inventory adjustments exceeding normal seasonal adjustments in 1Q 2017. Meanwhile, companies in the supply chain appear to have deeply rooted concerns about a gap between end demand and production volume, leading us to note the risk of inventory adjustments in May–Jun 2017 as cyclical demand for new products plateaus in early spring. In semiconductors, the NAND sufficiency ratio has dropped further, and lower planar NAND (2D) supply due to the shift to 3D NAND has spurred further market tightness, prompting NAND price rises for 1Q 2017. Regarding capex, we could not confirm any new concrete projects in our latest survey. In SPE orders, 3D NAND equipment should remain high through 1Q 2017. Foundry orders in the 10nm process are likely to peak out, but we anticipate investment in the 7nm process starting in late-1Q/early-2Q. Large orders from local Chinese chipmakers are still too far away to assume, in our view. ■ Key takeaways: (1) Logic/foundry: Taiwanese foundry 300mm line operating rates (input) dropped below 90% in Sep–Oct, but have rebounded to around 95% since November. We therefore think rates should remain high in 1Q 2017. Korean LSI advanced process lines are at full capacity, while Taiwanese foundries are roughly one year ahead of Korean LSI makers in 7nm mass production. (2) DRAM: The shift in production from mobile RAM to PC DRAM poses a risk of excess supply in 1H 2017, but DRAM companies are likely to control prices through inventory increases. Adoption of LP DDR4 in mid-range Chinese smartphones is likely to come in 2H 2017 at the earliest. (3) NAND: Declining planar NAND bit growth starting in 4Q is deepening supply tightness. We expect major NAND maker sufficiency ratios were 70–80% in 4Q and to be 70–80% in 1Q 2017. NAND prices were up 10–25%QoQ in 4Q 2016, and we foresee them rising 5–30% in 1Q 2017. Some NAND makers are discontinuing 4GB and 8GB supply. (4) Front-end SPE equipment: Korean 3D NAND investment is to begin with equipment delivery in December (continue through 1H 2017). We expect Taiwanese foundries to continue expanding 10nm process capabilities through 1H 2017 and start investing in 7nm trial production lines, followed by mass-production investment in 2H 2017. It is still too early to expect memory investment by Chinese local makers, in our view. (5) Backend SPE: OSAT operating rates are high, but capex is slowing. We see no forward progress in FO-WLP/PLP. ■ Stock calls: Our top picks are Toshiba (as NAND supplies remain tight in 1Q 2017), Tokyo Electron and Hitachi Kokusai Electric (owing to full-scale 3D NAND investment), and Towa (as investment in fingerprint authentication sensors remains high). With utilization rates at Taiwanese foundries coming off their recent lows, we see few near-term negatives overall for the technology sector.

DISCLOSURE APPENDIX AT THE BACK OF THIS REPORT CONTAINS IMPORTANT DISCLOSURES, ANALYST CERTIFICATIONS, LEGAL ENTITY DISCLOSURE AND THE STATUS OF NON-US ANALYSTS. US Disclosure: Credit Suisse does and seeks to do business with companies covered in its research reports. As a result, investors should be aware that the Firm may have a conflict of interest that could affect the objectivity of this report. Investors should consider this report as only a single factor in making their investment decision.

6 December 2016

Executive summary and investment implications Logic/memory: NAND prices poised to rise more in 1Q 2017. Taiwanese foundry utilization rates hitting lows in Sep–Oct Foundry/logic utilization rates are high Taiwanese foundry 300mm line capacity utilization (input basis), which we focus on as a leading indicator for the overall technology sector, started to show signs of adjustment at the time of our September survey. However, the results of our most recent survey suggest that the adjustments were completed in a short period of time, with utilization rates hitting lows in Sep–Oct and turning upward since November. Previously at risk for oversupply, 28nm lines remain at full capacity, and we see no concerns about adjustments in 1Q 2017. However, there is a lingering risk of excess inventories given smartphone production volume, the production shift from 28nm to 14nm, and market share trends in the application processor (AP) maker market (Qualcomm upbeat, Mediatek and Spreadtrum slowing). Chinese smartphone production is unlikely to correct sharply in 1Q 2017, and because production of new models looks to start in early spring as usual, we believe the risk of inventory adjustments has been pushed back to May–Jun. Consequently, overall semiconductor supply chain adjustments are unlikely to exceed normal seasonality. We therefore see few share price negatives in the market. Also, in our most recent survey, we heard for the first time of demand-supply tightness in discrete devices (power semiconductors, small signal processors). Memory demand-supply remains tight, but watch for DRAM price risk in late-1Q 2017; NAND demand-supply remains tight DRAM and NAND prices continue to rise amid tightness in the demand-supply balance, but we expect prices to diverge in 1Q 2017. DRAM supply balance could crumble in late-1Q 2017; themes for 2017 are 1xnm and LP DDR4 DRAM production is tilted toward mobile RAM, causing the demand-supply balance to worsen. Consequently, the balance for PC DRAM is particularly tight. However, the major Korean DRAM maker appears to be shifting back to PC DRAM, which means that rising PC DRAM supply from late-1Q onward could worsen the demand-supply balance for a time. On the other hand, DRAM makers are curtailing supply volume by increasing inventories, so prices appear to be stabilizing. We therefore see little risk of a sharp drop in DRAM prices. There are two key themes for 2017: (1) miniaturization on the 1xnm process and (2) expanded adoption of LP DDR4. We see no signs of aggressive production capacity expansion in 1xnm process migration, and heavy investment in this migration also appears to be lacking. In addition, LP DDR4 started to be adopted for high-end smartphones in 2016, and we expect adoption in mid-range models in 2H 2017, as there are orders for LP DDR4-compliant application processors (APs). We expect limited investment in desktop processors considering the benefits to the shift in the product mix.

Technology sector 2 6 December 2016

NAND supply tightness likely to persist in 1Q 2017. 3D NAND adoption expanding, including use in smartphones We expect NAND demand to remain upbeat due to continued growth in installed density. On the other hand, the shift in production toward 3D NAND has resulted in negative bit growth for planar NAND, so we expect supply to become even tighter. Major NAND makers' sufficiency ratios are around 70–80%, dropping from 80% in 3Q. In 1Q 2017, in addition to demand for increased installed density by Chinese smartphone makers, we expect reduced supply volume for planar NAND due to the production shift toward 3D NAND. We therefore expect the demand-supply balance to remain tight despite 1Q typically being a period of weak demand. Volumes and prices have been set for 1Q, and volume/price negotiations for 2Q appear to be getting underway. Prices differ in accordance with potential supply volume, and we expect prices to rise 10–25% QoQ in 4Q 2016 and 5–30% in 1Q 2017. In 3D NAND, the major Korean player is increasing 48-layer capacity to meet current demand (by shifting investment from planar NAND and by making new investments). Meanwhile, Japanese makers are steadily preparing for greater 64-layer demand, where they are ahead on shipping samples. However, the leading Korean maker is expected to start shipping samples in December as well. This manufacturer is scheduled to start shipping 64-layer SSD samples in 1Q 2017, followed by the Japanese makers in 2Q, which points to the start of full-scale shipments in 2H. As such, we do not expect a narrowing of the technology gap between the top- and bottom-tier makers. Nevertheless, the demand-supply balance could tighten further in 2H 2017 if 64-layer 3D NAND start-up does not go smoothly. We are therefore keeping a close watch on yield trends. Toshiba still our top pick Toshiba remains our top pick. We see upside to company guidance based on higher average selling prices amid demand-supply tightness—even during the traditionally weak demand period of 1Q 2017. If the start of 64-layer mass production goes well, we foresee further improvement in operating margins becoming a catalyst for more share price upside. SPE We expect front-end process SPE demand to be driven by 3D NAND investment and foundry 7nm process investment over the next six months. It is still too early to expect investment by local Chinese firms, in our view. For backend process SPE, we anticipate no major investment moves aside from that related to fingerprint authentication sensors. Front-end SPE: 3D NAND investment by major NAND makers should continue through 1Q 2017; next major foundry investment the 7nm process No new semiconductor capex could be confirmed in our most recent survey. At this point, we expect foundries to increase 10nm process capacity and the major Korean maker to start delivery of 3D NAND equipment at new fabs in December 2016. The major upcoming investments look to be 64-layer 3D NAND by Japanese makers in 1Q 2017 and 7nm lines by Taiwanese foundries in 2Q. In NAND capex, the major Korean NAND maker has basically completed visual inspection of its new fab and plans to start delivery of 48-layer 3D NAND equipment in December 2016, with production start expected in late 1Q or 2Q 2017. Japanese NAND makers’ capex budgets are skewed toward Jan–March, with major investment in 64-layer 3D NAND likely to start. The largest foundry in Taiwan is expanding its 10nm process capacity and is in the process of delivering equipment for trial production for the development of 7nm process products. Its new 7nm process fab is basically finished, so we expect investment in 7nm

Technology sector 3 6 December 2016

process pilot lines for fab certification. Moreover, we believe the company will start placing orders for equipment for mass-production lines starting between end-1Q and 2Q 2017 We thus expect front-end SPE firms to see strong orders (although with some change in the order mix) from memory chipmakers and foundries in 4Q 2016 and 1–2Q 2017. Backend SPE: Capacity utilization remains high, but investment is down due to normal seasonality; fingerprint authentication sensor investment remains strong. No major progress in FO-WLP/PLP OSAT capacity utilization remains high due to high output at foundry chipmakers. Still, from a capex perspective, we do not see any major moves overall. We expect that under these conditions, investment in fingerprint authentication sensors will remain high in 2017 after very brisk activity in 2016. We do not expect major investment in FO-WLP/PLP (Fan-Out Wafer Level Package/Panel Level Package) in 2017. The InFO-WLP specifications for 2017 models are the same as for 2016 models, which limits investment. We also believe the adoption of new Korean LSI technologies for use in new smartphone models in 2017 has been postponed. The major investment themes for test process-related equipment are new 3D NAND fabs in 2017 and the adoption of LP DDR4 in mid-range smartphones. But our impression is that investment will be limited to small orders. Still recommend Tokyo Electron, Hitachi Kokusai Electric, and Towa We continue to recommend front-end process SPE makers, favoring memory over foundry and logic. Our top picks are Tokyo Electron and Hitachi Kokusai Electric. SCREEN, which we started to focus on in our last survey, is still putting through the evaluation process its wafer cleaning equipment for use by Korean 3D NAND makers. Assuming full-scale adoption occurs, we see the potential for SCREEN to separate itself from the foundry/logic-related SPE names. For this reason, we are keeping an eye on SCREEN. Backend-process SPE firms are in an order deceleration cycle for assembly-related equipment, meaning that share price catalysts are lacking. However, investment in fingerprint authentication sensors remains upbeat, so we like TOWA from a relative perspective due to its strength in compression molding technology. Test-related firms lack catalysts due to limited 1xnm process migration investment in the DRAM field and because the adoption of LP DDR4 in mid-range Chinese smartphones is likely to come in 2H 2017 at the earliest. Investment in 3D-related test process equipment is also small. Memory test-related Advantest and Micronics Japan appear to be in a somewhat harsh phase. However, mass production of APs for the 10nm process has started, leading us to expect Advantest’s orders to rebound in Oct–Dec 2016 and Jan–Mar 2017. We therefore see limited share price downside catalysts in the near term.

Technology sector 4 6 December 2016

Figure 1: Semiconductor/SPE valuations (as of 5 December) Closing Closing Price: 12/05/16 Price: Mkt Price EPS Dvd EV / EV / Net Absolute Relative performance ROE PBR Yld Sales EBITDA D/E performance % % Cap Cur TP % to TP CS EPS P/E, x Company Code Rating 3/16 3/17 3/18 3/16 3/17 3/18 3/17 3/17 3/17 3/17 3/17 3/17 JPY, bn JPY JPY % 1M 3M 12M 1M 3M 12M A E1 E2 A E1 E2 % % x x x % Semiconductor Toshiba 6502 OUTPERFORM 1,918 436 460 5.4 -108.6 48.8 47.7 -2.0 9.3 9.5 0.0 47.9 3.6 0.5 5.9 96.7 24.6 34.0 50.0 15.0 24.6 56.8 SPE Hitachi Kokusai Electric 6756 OUTPERFORM 224 2,176 1,980 -9.0 126.5 144.1 151.9 10.7 15.2 14.4 1.4 14.5 2.1 0.9 6.3 -53.1 2.7 19.7 18.4 -7.0 10.3 25.2 Tokyo Electron 8035 OUTPERFORM 1,633 10,000 10,200 2.0 461.1 577.6 668.5 15.9 17.2 14.9 2.9 16.2 2.7 1.9 8.6 -43.0 3.9 7.0 22.0 -5.7 -2.4 28.8 SCREEN Holdings 7735 NEUTRAL 296 6,400 5,000 -21.9 396.8 485.0 522.9 11.2 12.9 12.0 1.1 17.6 2.1 1.1 8.7 -4.1 -9.3 -5.1 36.3 -18.9 -14.6 43.1 Disco 6146 NEUTRAL 470 13,050 9,450 -27.6 646.1 484.6 476.2 14.8 27.1 27.6 2.0 10.1 2.7 3.4 12.9 -37.7 6.1 10.8 8.2 -3.6 1.3 15.0 Advantest 6857 UNDERPERFORM 268 1,536 765 -50.2 45.5 20.7 5.2 22.9 74.2 296.6 1.0 2.7 2.0 1.2 17.4 -72.7 4.9 3.0 35.3 -4.8 -6.4 42.1 JEOL 6951 OUTPERFORM 47 487 510 4.7 42.3 -9.0 12.9 13.5 -53.7 37.6 1.4 -2.8 1.6 0.6 21.3 38.0 15.2 27.7 -39.3 5.6 18.2 -32.5 NuFlare Technology 6256 NEUTRAL 67 5,510 5,820 5.6 767.7 803.2 259.4 6.5 6.9 21.5 2.2 17.0 1.1 1.3 4.2 -1.3 -1.4 11.5 -9.9 -11.1 2.1 -3.1 Lasertec 6920 NEUTRAL 47 1,990 1,710 -14.1 143.2 134.2 136.6 8.7 14.8 14.6 2.4 13.3 1.9 2.1 8.4 -26.9 8.4 7.2 44.5 -1.2 -2.2 51.3 TOWEA 6315 OUTPERFORM 33 1,341 1,680 25.3 71.6 129.9 173.8 10.0 10.3 7.7 0.8 14.4 1.4 1.1 5.8 -11.3 16.4 7.7 54.8 6.8 -1.8 61.6 Micronics Japan 6871 UNDERPERFORM 42 979 775 -20.8 144.7 13.2 27.6 7.5 78.5 37.6 1.4 2.4 1.9 1.2 10.0 -45.5 -11.0 -23.6 -19.6 -20.6 -33.1 -12.8 Tokyo Seimitsu 7729 NEUTRAL 131 3,165 2,740 -13.4 234.6 223.4 224.7 9.4 14.2 14.1 2.1 11.3 1.5 1.4 6.5 -37.4 14.7 17.4 10.9 5.1 7.9 17.7 Source: Thomson Reuters, Credit Suisse estimates

Technology sector 5 6 December 2016

Figure 2: Japanese technology stocks' performance over the past six months

6 month stock returns

FUJITSU 62.2% TOSHIBA 60.7% TOWA 58.5% HITACHI KOKUSAI 49.6% LASERTEC 44.5% MITSUMI 39.4% CLARION 38.6% SHARP 36.2% ALPINE ELECTRONICS 32.1% ADVANTEST 31.1% ALPS 30.7% JAPAN DISPLAY 29.1% TDK 28.4% TOKYO SEIMITSU 28.0% DISCO 27.4% TOKYO ELECTRON 26.8% HITACHI 25.6% NIDEC 25.2% PANASONIC 25.1% KONICA MINOLTA 24.1% NEC 22.7% JVC KENWOOD 22.4% MITSUBISHI ELECTRIC 21.8% SCREEN HOLDINGS 21.6% MURATA 20.5% NGK SPARK PLUG 19.1% CITIZEN HDG. 18.4% NIKKEI 225 16.9% IBIDEN 16.0% SHINKO 14.2% NUFLARE TECHNOLOGY 11.3% NISSHA PRINTING 11.0% PIONEER 10.4% TOPIX ELECTRIC APPLIANCES 9.8% NIKKEI 225 STOCK AVERAGE 9.7% TAIYO YUDEN 9.5% NIKON 6.4% CANON 5.5% SEIKO HOLDINGS 4.2% KYOCERA 3.7% SONY 2.9% HIROSE 1.3% RICOH 1.0% MICRONICS JAPAN -1.9% FUJIFILM HOLDINGS -2.3% JAPAN AVNS.ELTN.IND. -3.5% CASIO COMPUTER -7.5% OLYMPUS -9.7% JEOL -9.8% WACOM -24.6% -40% -20% 0% 20% 40% 60% 80%

Source: Thomson Reuters

Technology sector 6 6 December 2016

Semiconductor market outlook DRAM: Tight supply–demand due to maker production mix; risk of oversupply from end-1Q through 2Q 2017, but companies will likely stabilize prices by building up inventories Key takeaways:

■ Production shift to mobile RAM is creating tight supply–demand in 2H 2016, mainly due to supply-side factors. However, we expect oversupply on a production capacity basis in 1H 2017.

■ We expect mobile RAM overproduction in 4Q 2016 due to competition for share of supply to the major North American smartphone maker. For 1Q 2017, we look for oversupply, as demand from smartphone makers has typically been weak in the first quarter of recent years. We also see risk of PC DRAM swinging to oversupply from Mar–Apr due to increased production resulting from changes in the production mix.

■ However, we expect companies to stabilize market prices, controlling supply by building up inventories to meet demand in 2H 2017.

■ On the demand side, installed memory has continued to rise. More PCs are being equipped with 8GB of memory. In mobile RAM, LP DDR4 is being adopted for high-end smartphones and we expect it to be fitted in midrange units in 2H 2017. Major Chinese smartphone makers are expected to adopt LP DDR4X (high-speed version).

■ On the supply side, from this month (December), we expect a production shift toward PC DRAM due to margins and seasonality. However, as regards market bit growth, we anticipate a slight QoQ decline due to changes in the production mix.

■ Margins on PC DRAM have started to exceed those on mobile RAM and server DRAM, reducing supply preference for server DRAM. We expect this to result in short supply of server DRAM and to limit the fill rate at major DRAM makers to around 60%.

■ DRAM inventories continued to fall in November, but a production shift to these products and seasonality should eliminate shortages. Mobile RAM inventories are likely to build up in 1Q 2017 due to overproduction and weaker demand due to seasonality.

■ We expect PC DRAM price increases of 20% or more in 1Q 2017.

■ In advanced process investment, we expect the Korean major to start using 1xnm process for PC memory (previously server and mobile RAM; changed to PCs due to a focus on margins and to certification). It plans to increase the proportion of production capacity that uses 1xnm process to 30–40% by end-2017, but aggressive investment looks unlikely due to the heavy investment burden. We expect US makers plan to step up introduction from mid-1Q 2017.

■ We believe there are concerns about supply of next-generation memory 3D Xpoint, and requests for aggressive investment in ReRAM.

Technology sector 7 6 December 2016

DRAM demand–supply outlook: Tightness due solely to product mix changes; expecting mobile RAM oversupply in 1Q 2017, PC DRAM oversupply from spring due to higher production Supply-demand tight in Chipmakers have shifted toward mobile RAM production in 2H 2016 to meet rising 2H CY16 due to change smartphone installed density, resulting in tight PC DRAM supply. As major Korean DRAM in DRAM production makers are lagging in 20nm process production, they shifted sharply to mobile RAM mix, but capacity production to regain supply share to the major North American smartphone maker in 4Q. remains in excess We think this helped tighten PC DRAM and mobile RAM for Chinese smartphone supply. We think some makers would still have excess production capacity versus DRAM demand even if they optimized production mixes and yields. However, US DRAM makers, which caught up the lag have not cut mobile RAM production in 4Q 2016. We think this, coupled with increased mobile RAM production at Taiwanese fabs and slow adoption in the key Chinese market, is raising concerns of general mobile RAM oversupply. Some makers Major DRAM makers are considering shifting from mobile RAM to PC DRAM, and we see considering shift to PC this happening in December 2016 or January 2017. We attribute this to generally lower DRAM; excess supply demand for mobile RAM caused by seasonal production adjustments, one of the major could be absorbed by DRAM makers having a monopoly of supply for 3GB chips to the North American major, stockpiling inventory and the possibility of later-than-planned acquisition of parts for companies’ own new smartphone models. In addition, PC DRAM margins have improved and are now higher than margins on other products, encouraging a production shift. On the demand side, PC DRAM continues to see growth in 8GB (4GB + 4GB) installs, reflecting better performance and GB costs over 6GB (4GB + 2GB). Smartphone installed density has continued to rise, and there appear to be no particularly new developments outside of LP DDR4, which we discuss later. Assuming excess supply in 1H 2017, we expect chipmakers to control supply–demand by building up inventories, thereby stabilizing market prices. We therefore do not expect DRAM prices to fall sharply in 1H 2017.

Figure 3: DRAM supply–demand model

15% 12% 12% Oversupply 11%

10% 7%7% 6% 5% 5% 5% 3% 2% 1% 0%

-5% -4% -4% -5%

-10% -9% Supply shortage

-15%

3Q05 3Q08 2Q10 2Q13 2Q04 3Q04 4Q04 1Q05 2Q05 4Q05 1Q06 2Q06 3Q06 4Q06 1Q07 2Q07 3Q07 4Q07 1Q08 2Q08 4Q08 1Q09 2Q09 3Q09 4Q09 1Q10 3Q10 4Q10 1Q11 2Q11 3Q11 4Q11 1Q12 2Q12 3Q12 4Q12 1Q13 3Q13 4Q13 1Q14 2Q14 3Q14 4Q14 1Q15 2Q15 3Q15 4Q15 1Q16 2Q16 3Q16 4Q16 1Q17 2Q17 3Q17 4Q17 1Q04 Source: Credit Suisse estimates

Technology sector 8 6 December 2016

Figure 4: Mobile RAM supply–demand model Figure 5: PC DRAM supply–demand model

14,000 24% 25% 14,000 40%

12,000 20% 12,000 24% 30% 22% 14% 13%14% 15% 16% 10,000 12% 15% 10,000 13% 13%13% 20% 11% 10% 11%10% 6% 8% 8,000 10% 8,000 5% 4% 4% 10% 6% 0% 5% 4% 0% -2% 4% 3% 3% -2% 6,000 2% 5% 6,000 -6% 0% 1% -9% 0% -1% 0% 0%0% -13%

1Gb equiv. mn units mn equiv. 1Gb -15% 4,000 -3% 0% units mn equiv. 1Gb 4,000 -10% -4% -6% 2,000 -9% -8% -5% 2,000 -26% -20%

0 -10% 0 -30%

1Q12

3Q12

1Q13

3Q13

1Q14

3Q14

1Q15

3Q15

1Q16

3Q16

1Q17

3Q17

1Q12

3Q12

1Q13

3Q13

1Q14

3Q14

1Q15

3Q15

1Q16

3Q16

1Q17 3Q17

Demand Supply Demand/supply sufficiency rate Demand Supply Demand/supply sufficiency rate

Source: Credit Suisse estimates Source: Credit Suisse estimates

Figure 6: Chinese smartphone DRAM density trend

100% 3.00 90% 2.50 80% 70% 2.00 60% 50% 1.50 40% 1.00 30% 20% 0.50 10% 0% 0.00 2014Q1 2014Q2 2014Q3 2014Q4 2015Q1 2015Q2 2015Q3 2015Q4 2016Q1 2016Q2 2016Q3

512MB 1GB 2GB 3GB 4GB 6GB Average DRAM (GB)

Source: IDC LP DDR4 adoption trend LP DDR4 could be used Regarding LP DDR4, which has been increasingly adopted in high-end smartphones, we in midrange think adoption in Chinese midrange smartphones (priced from RMB2,000 = above smartphones from 2H US$290) will likely increase alongside the launch of products feature LP DDR4-compatible CY17; high-speed application processors. However, we do not anticipate their market debut until 2H 2017 version set for use in (we see this as a key data point with respect to forecasting DRAM tester demand, high-end Chinese discussed below). However, we do not yet see any signs of LP DDR4 adoption in low-end phones smartphones. Application processor compatibility appears to be generally limited/slow in the transition to LP DDR4 compared with the move from LP DDR2 and LP DDR3. A number of Chinese smartphone makers have expressed interest in the high-speed version (3GHz band to 4GHz band) of LP DDR4 (LP DDR4X), and we expect adoption to start in the future. However, we think North American/Korean smartphone makers do not yet have plans for adoption.

Technology sector 9 6 December 2016

Figure 7: Chinese smartphone breakdown by ASP 100%

80% $0<$50 $50-$100 60% $100-$200 40% $200-$300 $300-$400 20% $400-$500 0% $500-$600

$600<

2015Q2 2007Q1 2007Q4 2008Q3 2009Q2 2010Q1 2010Q4 2011Q3 2012Q2 2013Q1 2013Q4 2014Q3 2016Q1

Source: IDC

DRAM inventories: Downtrend continues in November PC DRAM inventory The largest Korean maker and other major Korean manufacturers expect overall continues to shrink November DRAM inventories including PC and mobile RAM to decline by roughly another week of sales (from around three weeks currently; even shorter for PC DRAM) compared with end-September. Meanwhile, we expect PC DRAM supply to rise due to seasonally depressed demand from 1Q 2017 and a change in the product mix. We believe that individual chipmakers aim to stabilize prices by boosting inventories amid excess supply in 1H 2017. DRAM prices: Price increases again in 1Q 2017 Prices could rise nearly PC DRAM prices continue to rise amid currently tight supply, and the outlook is for a 20% in 1Q CY17 despite nearly 20% price increase (QoQ) in 1Q 2017 to $22–23 per 4GB. Outside of PC DRAM, shift to PC DRAM we expect similar price increases for server DRAM and DRAM used in consumer production; we see 1Q electronics to match the price level for PC DRAM. as likely peak We think prices are likely to peak in 1Q, as it appears the largest Korean maker is considering shifting production to PC DRAM.

Technology sector 10 6 December 2016

Figure 8: DDR4 4GB module large lot price trend US$ 22.00

20.00

18.00

16.00

14.00

12.00

10.00

Oct

Apr

Oct

Jan

Jun

Mar

Feb

Aug

Sep

Nov

Dec

July

Aug

Sep

Nov

Dec May 2015 2016

Source: Company data, Credit Suisse estimates DRAM 1xnm process migration Chipmakers cool In our previous report, we had expected Korea’s largest DRAM maker to prioritize mobile toward investment in RAM and server DRAM for the 1xnm process. However, in our latest survey, we confirmed 1xnm node a change of policy to put priority on PC DRAM. One reason for this is that rising prices have put PC DRAM margins above those for mobile RAM and server DRAM. We estimate end-2016 production capacity at around 10%, but this is because PC ODM certification has just been completed, so we believe production volume is still low. As major Korean players are focusing on 20nm start-up, we could not confirm concrete plans for 1xnm mass production. Meanwhile, US makers are prototyping 1xnm process in Hiroshima and Taiwan; we believe they plan to start mass production in 1Q 2017. We understand there is no sign yet that any DRAM maker has aggressive plans for process migration, and that their process migration plans will depend on demand. We believe this is because, as yield is not stable yet, DRAM makers put more priority on near- term profitability than on making risky plays. Next-generation memory development (3D Xpoint, ReRAM, MRAM) Next-generation Next-generation memory has been a recent topic of conversation, but many market memory will take time participants have concerns about 3D Xpoint mass production, and some makers are seeing requests from US MPU makers for expedited ReRAM development. However, we expect full-scale ReRAM rollout to take 3–4 years. Also, as MRAM requires an expensive carbon film shield to protect against malfunctions due to external magnetic waves, we believe its prospects as a next-generation memory candidate are receding from the viewpoints of cost and thickness. An easy-to-understand example are disruptions in the magnetic function of the PiP Elekiban device.

Technology sector 11 6 December 2016

NAND: 2D supplies drying up Main points from our research

■ We expect the NAND supply–demand balance to remain tight in both 4Q and 1Q. NAND sufficiency ratios were about 80% of demand in 3Q, and we estimate they will be about 70–80% in 4Q and 1Q.

■ Although we expect seasonal production volume corrections in Chinese smartphones, for the NAND supply–demand balance, we also expect continued growth in memory demand per phone, up 5–10% QoQ. In procurement for North American smartphones, we expect a narrow decline in memory demand in step with phone production volume.

■ In Korean smartphones, we expect the spring 2017 Galaxy S8 to offer a choice between 64GB or 128GB of NAND (planar NAND rather than 3D NAND; the Galaxy S7 series had 32GB and 64GB versions). In mid-range to low-end Chinese eMCPs, the most common configuration is 16+16 (2GB DRAM + 16GB NAND), and we now expect the volume shift to 32+32 (4GB DRAM + 32GB NAND) to arrive only in 2017.

■ SSD demand is brisk in notebook, enterprise, and data center segments. Supplies of SSDs for notebooks (2D planar NAND) are not keeping up with demand and we understand some PC makers are thinking of shifting back to HDDs for some models.

■ NAND majors have been shifting production capacity toward 3D NAND, and 2D NAND bits supplied is expected to decline in 4Q and then decline further in 1H 2017. Thus, 2D supply shortages are set to become even more serious not only in 4Q, but also for the entire first half of 2017.

■ Some NAND makers are planning to stop making lower-density 4GB and 8GB units.

■ We expect NAND prices to rise 10–25% in 4Q and 5–30% in 1Q. Price increases at makers with surpluses are especially large. Price negotiations for 1Q have already concluded and price and volume negotiations for 2Q are underway.

■ Chinese smartphone makers in strong financial positions are taking an attitude of minimal concern about price if they can secure the volumes they seek.

■ Leading 64-layer 3D NAND makers are planning to supply eMMC and SSD samples in 1H 2017, moving to mass-production and shipping in 2H. If 64-layer products do not reach expected yields, however, the risk of supplies becoming even tighter is high.

■ Chinese PC and server makers are planning to start manufacturing their own notebook SSDs and have set up new companies for this. The plan is to procure 50% of SSDs from in-house production and the remaining 50% from usual outside channels.

■ Our impression is that Toshiba’s market share is higher than expected in supplying chips to local Chinese SSD makers and data centers.

Supply–demand outlook NAND sufficiency likely Major NAND maker sufficiency ratios (supply volume versus total demand, including fell further in 4Q temporary demand) worsened from 80% in 3Q to 70–80% in 4Q, and we expect them to remain at that level in 1Q 2017. We therefore expect the demand–supply balance to tighten further.

Technology sector 12 6 December 2016

Bit demand from In 1Q 2017 NAND bit demand from smartphone makers, we expect declines in North Chinese smartphone America in step with unit declines, but in China, despite seasonal production corrections, makers could rise in 1Q we expect NAND demand to rise 5–10% QoQ with the continued increase in average CY17 despite seasonal density per phone. weakness Given the expected supply constraints in 2H 2017, in 1H, we expect inventory increases among smartphone makers in Chinese smartphone makers for mid-/low-end smartphone makers that do not procure in 4Q and good financial positions and module and SSD makers with low inventories. Therefore, we think decreasing demand at North America can offset the declines. Supply shortage On the supply side, 2D NAND bit supply started declining in 4Q due to a shift in production impeding planned to 3D NAND by manufacturers leading the switch. We therefore expect NAND supply to increases in memory remain fairly tight until 2H 2017, when 3D NAND supply picks up. In 1H 2017, we forecast capacity in low-end lower NAND bit growth YoY at the major NAND makers as they continue to shift Chinese smartphones production from 2D to 3D. Rising output in 2H 2017 after the switch has been completed should lift supply, but the tight supply-demand conditions could persist into 2H if yields fail to rise as planned, particularly for 64-layer 3D NAND.

Figure 9: NAND supply–demand model

25% Oversupply 20% 18% 13% 15% 13% 11% 12% 8% 10% 8% 8% 5% 4% 5% 5% 4%

0%

-5% -6% -6% -6% -10% -8% -9%-9% -15% -14% -20% -17% Supply shortage -19%

-25%

2Q11 4Q15 1Q17 1Q07 2Q07 3Q07 4Q07 1Q08 2Q08 3Q08 4Q08 1Q09 2Q09 3Q09 4Q09 1Q10 2Q10 3Q10 4Q10 1Q11 3Q11 4Q11 1Q12 2Q12 3Q12 4Q12 1Q13 2Q13 3Q13 4Q13 1Q14 2Q14 3Q14 4Q14 1Q15 2Q15 3Q15 1Q16 2Q16 3Q16 4Q16 2Q17 3Q17 4Q17

Source: Credit Suisse estimates

Technology sector 13 6 December 2016

Figure 11: 2D/3D NAND production capacity Figure 10: 2D/3D NAND bit supply forecasts forecasts

45,000 1,800 70%

40,000 1,600 60% 35,000 1,400 50% 30,000 1,200

25,000 1,000 40%

20,000 800 30%

15,000 600

Production (mn GB) (mn Production 20% 22,186 22,436 23,098 21,870

10,000 20,140 400 totalof%capacity NAND 3D 18,531 17,710 16,520 3D Mass productionCapacity (K wpm) (K productionCapacity Mass 3D 10% 5,000 200

0 0 0% 1Q16E 2Q16E 3Q16E 4Q16E 1Q17E 2Q17E 3Q17E 4Q17E 1Q16E 2Q16E 3Q16E 4Q16E 1Q17E 2Q17E 3Q17E 4Q17E

2D NAND 3D NAND 2D Capacity 3D Capacity % of total NAND capacity

Source: Credit Suisse estimates Source: Credit Suisse estimates

In this supply–demand environment, we confirmed (1) a leading Korean maker is proposing contracts that guarantee an increase in bit supply on condition that 3D NAND products pass certification, (2) a leading Korean maker plans to terminate supplies of low- density eMCP (4GB, 8GB), and (3) bit supply to some Chinese smartphone makers will be cut more than half in 4Q (rather than 2Q) due to the decline in planar NAND production. Demand for NAND, particularly planar NAND, is rising among smartphone makers, but the drop in planar NAND supply is leading to unprecedented supply–demand conditions. In our June report, we forecast a “NAND boom” and a recovery in demand. In September, amid news that some suppliers had put customers on allocation, our view was that NAND distribution structure would be the key for companies in the supply chain, but 2D NAND supplies now seem to be exhausted. Supply shortage Chinese smartphone makers also appear to be behind schedule in increasing NAND impeding planned density. Under our previous outlook, we expected mid-range and low-end Chinese makers increases in memory to upgrade to 32+32 eMCP (DRAM 4GB, NAND 32GB) by end-2016, but 16+16 (DRAM capacity in low-end 2GB, NAND 16GB) is still the mainstream configuration due to tight supplies and rising Chinese smartphones prices for NAND. We expect the shift to 32+32 eMCP to happen in 2017 if 3D NAND volume rises. CY17 could see global In global high-end smartphone models, we still expect US makers to adopt 3D NAND for smartphone makers 256GB and 128GB 2017 models. However, we understand that Korea’s leading NAND add more NAND maker may not supply 3D NAND for 128GB, partly because it is focusing on high-margin capacity per phone and business in customer allocations, meaning smartphone makers could be forced to stick also use 3D NAND with 2D. Despite that risk, we see a strong possibility that 3D NAND will be adopted for 128GB models, as Japanese NAND suppliers are planning production mixes to take on their rivals in 64-layer 3D NAND. We also expect the major Korean smartphone maker to increase NAND density in its spring 2017 flagship models. Surprisingly, the 2016 range was limited to 32GB and 64GB, with no 128GB model on offer, but we expect density to increase to 64GB and 128GB for 2017 models. Chinese smartphone makers are likely to expand 3D NAND use, but the Korean smartphone maker is likely to adopt Planar NAND. In the SSD market, NAND demand is strong for cSSD and eSSD, but some notebook PC makers have decided to revert to HDDs temporarily due to the lack of planar NAND supply.

Technology sector 14 6 December 2016

We had expected average Chinese Figure 12: Chinese smartphone NAND density ratio trend (planned) smartphone NAND 100% 100.0 density above 40GB in 90% 90.0 4Q2016 in September, but, the density rose to 80% 80.0 41GB. We expect 70% 70.0 density will not rise in 60% 60.0 4Q due to NAND dearth 50% 50.0 40% 40.0 30% 30.0 20% 20.0 10% 10.0 0% 0.0 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 2013 2014 2015 2016 2017 <1GB 1GB 4GB 8GB 16GB 32GB 64GB 128GB+ Ave density (GB)

Source: Credit Suisse estimates

Figure 13: NAND density: iPhone, Samsung, Huawei, Chinese smartphone average

70.0

60.0

50.0

40.0

30.0

20.0

10.0 Average Average NAND Density GB/unit

0.0

1Q13

2Q13

3Q13

4Q13

1Q14

2Q14

3Q14

4Q14

1Q15

2Q15

3Q15

4Q15

1Q16

2Q16 3Q16 Huawei Samsung iPhone China average

Source: Credit Suisse estimates Supplies of low-density NAND (4GB, 8GB) set to end Largest Korean NAND We understand that Korea’s leading NAND maker told customers in early November that it maker could stop plans to discontinue low-density eMCPs (NAND density of 4GB or 8GB). This came as a supplying 4GB/8GB to major surprise to many customers, who appear to be pressuring the company to maintain China supplies, but we see a strong possibility that low-density eMCP devices will be discontinued. These low-density packages contain 16Gb and 32Gb chips, which are fairly old designs compared with the current mainstream chip (14nm TLC 128Gb). Given current supply-demand conditions and the shift to 3D NAND, the decision to end supplies of low- density NAND is probably inevitable.

Technology sector 15 6 December 2016

That move is likely to impact users such as Chinese smartphone makers, independent design houses (IDH; producers of export models) and retail suppliers. The decision to end supplies will likely force makers to increase density in low-end devices and retail memory cards to 16GB or higher, which will be positive for NAND demand. That said, the impact is likely to be limited, as Chinese smartphones with density of 8GB or less accounted for only 10% of total shipments in 3Q 2016, down from 37% a year earlier. Moves by Chinese smartphone makers to increase NAND density are likely to play a greater role in lifting NAND demand. Use of 3D NAND will The shift to 3D NAND will require a minimum chip spec of 128Gb (=16GB) for 48-layer boost minimum and 256Gb (=32GB) for 64 layer, leading to increases in smartphone density regardless of required density to smartphone maker needs. 16GB/32GB Figure 14: Chinese smartphone units by NAND density, ratio of below 8GB

140 105%

120 90%

100 75%

80 60%

mn units mn 60 45% 59 67 40 65 30% 64 53 54 20 40 38 15% 26 17 12

0 0%

2014Q1

2014Q2

2014Q3

2014Q4

2015Q1

2015Q2

2015Q3

2015Q4

2016Q1

2016Q2 2016Q3

0GB<1GB 1GB<4GB 4GB<16GB 16GB<32GB 32GB<64GB 64GB<128GB 128GB<256GB 256GB+ % of < 8GB

Source: IDC

Figure 15: IDH shipment units ranking (1H 2016)

35 29.8 30 27.7 25

20

15 mn units mn 12.1

10 9.3 8.1 6.4 5.7 5 4.5 3.5 2.3

0

CK

CCI

Wind

Huiye

TINNO

Huaqin

Wingtech

Longcheer

Sprocomm Chino/OnTim

Source: Company data

Technology sector 16 6 December 2016

NAND prices raised for 1Q 2017, 2Q prices now being negotiated Prices set to rise by NAND users have already completed bit allocation and price negotiations for 1Q 2017 and double digits QoQ in have recently started 2Q negotiations. We expect some Chinese smartphone makers, 4Q and 1Q CY17 which have strong finances, to resist supplier pricing if they can secure sufficient volumes. NAND prices continue to rise amid tight supply–demand, but the size of price hikes varies between companies due to different supply conditions. Korea’s leading supplier, which currently has spare 3D NAND capacity, apparently raised prices by 20−30% QoQ in each quarter from 3Q 2016 through 1Q 2017. Chinese smartphone makers are not happy about the latest price hikes. They have no other option but to accept the large increases, but they seem to harbor a grudge about the situation. Japanese NAND suppliers will probably struggle to significantly hike prices, as bit supply is tight now due to the shift from planar NAND to 3D NAND production. However, we estimate they pushed through hikes of around 10% QoQ for 4Q 2016 and 5−10% QoQ for 1Q 2017. The decline in planar NAND bit supply is likely to continue into 2Q, so we anticipate flat prices or a continued upward trend.

Figure 16: 128Gb TLC NAND contract price trend

8.0

7.0

6.0

5.0

4.0 USD 3.0

2.0

1.0

0.0

Oct-12

Apr-13

Oct-13

Apr-14

Oct-14

Apr-15

Oct-15

Apr-16

Oct-16

Jun-13

Jun-14

Jun-15

Jun-16

Feb-13

Feb-14

Feb-15

Feb-16

Dec-12

Aug-13

Dec-13

Aug-14

Dec-14

Aug-15

Dec-15 Aug-16 NAND contract price TLC 128Gb 16384Mx8

Source: DRAMeXchange, Credit Suisse NAND inventory NAND manufacturers’ wafer sales traditionally go to module manufacturers, but tighter NAND supply has prompted the majors to lower their wafer sales ratios as far as possible, making sourcing difficult for Taiwanese module manufacturers and SSD companies that have close relationships with them. The situation appears to have worsened since our previous survey. However, the poor-quality portion of US-manufactured 3D NAND appears to be going to Taiwanese module manufacturers, resulting in pronounced distinctions among sector players.

Technology sector 17 6 December 2016

Figure 17: Phison Electronics' inventory Figure 18: A-Data Technology's inventory

350 120 7,000 140

300 100 6,000 120 250 5,000 100 80 200 4,000 80 60 150 3,000 60 40

100 days Inventory 2,000 40 days Inventory

Inventory (NT$ mn) (NT$ Inventory Inventory (US$ mn) (US$ Inventory 50 20 1,000 20

0 0 0 0

3Q 2Q 1Q 2Q 3Q 4Q 1Q 2Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 3Q 4Q 1Q 2Q 3Q

2Q 1Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 2Q 3Q 2011 2012 2013 2014 2015 2016 2011 2012 2013 2014 2015 2016

Inventory Inventory days Inventory Inventory days

Source: Company data Source: Company data

3D NAND demand, development, and production trends 3D NAND being In addition to 256GB, North American smartphone manufacturers are expected to use 3D increasingly used in NAND for 128GB in their 2017 models. However, some NAND manufacturers appear Chinese smartphone reluctant to supply 3D NAND for 128GB because they want to prioritize more profitable SSDs and other SSD applications. applications Elsewhere, tighter supply of planar NAND gives Chinese smartphone manufacturers no option but to use 3D NAND, and North American PC manufacturers are apparently considering 3D NAND (MLC) for SSDs in 2017. Japanese makers lead At the same time, the development roadmap for 64-layer NAND indicates that major in 64-layer 3D NAND; manufacturers will likely ship engineering samples to clients from 1Q 2017 at the earliest. Korean rivals set to Korean manufacturers plan to ship SSD samples in 1Q, and Japanese companies in 2Q. ship samples no later than 1Q CY17; SSD In addition to switching their investment from planar to 48-layer NAND, South Korean samples likely to be manufacturers are adding new capacity, so we see little 3D NAND supply risk in 2017. In supplied in 1H CY17 contrast, other companies are limiting the investment switch from planar NAND, and their planned mass-production of 64-layer NAND suggests a possible acute supply shortage in 2H 2017 if yields at ramp-up do not improve as planned. Large gap between top- The major and lower-tier NAND manufacturers have very different views on 3D NAND and bottom-tier makers development. US manufacturers plan mass-production in 2017, but have apparently yet to in developing 3D NAND decide on floating gate or charge trapping technology. Korean manufacturers plan to with 64 or more layers convert 10% of their capacity to 3D NAND at end-2016, but technical issues with 48-layer samples (chip base) make evaluation difficult, and we expect actual output to remain at pilot-line level. Moving to next-generation product, manufacturers plan to complete development of 72-layer NAND in 1H 2017. However, as the process appears to involve two 36-layer stacks, manufacturing costs will increase and we think the product will be unable to generate cost benefits from layer expansion. We thus expect the Korean majors to increase supply of 48-layer NAND in 2017 and ramp up 64-layer NAND in 2H and estimate that Japanese manufacturers also plan to launch 64-layer NAND in 2H. We see these developments further widening the gap with lower-tier manufacturers. China server, data center demand China starts producing servers domestically in 2017, and is apparently putting in place an investment incentive scheme. Companies such as Lenovo, which begin with in-house manufacture of SSD for PCs, are also setting up SSD production operations (described later). On the demand side, the market for enterprise SSD in China, where public cloud services are restricted, is still only about one-third that in the US, confirming that conditions differ

Technology sector 18 6 December 2016

little from the picture we formed earlier. China demand for enterprise SSD does not appear to be substantial at this stage. Toshiba has strong Anecdotal evidence of a Toshiba presence in China’s enterprise SSD market also came presence in supplying as a surprise. Beijing Memblaze Technology and Shannon Systems are the two majors in chips for Chinese China’s flash storage market, with Toshiba apparently supplying mainly chips. At the same enterprise SSDs time, the picture for data center demand (Tencent Holdings, Alibaba Group Holding, Baidu) appears to be one of Intel supporting Tencent Holdings and Baidu while Shannon Systems provides mainly SSD support to Alibaba Group Holding. Huawei is supported by Toshiba and Micron Technology. As an SSD manufacturer, Samsung Electronics apparently provides negligible chip supply to these rival storage manufacturers.

In-house SSD production Chinese PC makers to A major Chinese PC and server manufacturer plans to establish a new company to make SSDs in house produce proprietary SSD for notebook PCs. It targets increasing the in-house production ratio of cSSD for proprietary use to 50%. The controller will apparently be from Marvell, and we understand that NAND will be supplied exclusively by Toshiba. The operation is to initially use planar TLC NAND, followed by 3D NAND in the future. We have confirmed that Toshiba has some share of eSSD and cSSD chip supply in China.

Figure 19: Enterprise storage market (worldwide) Figure 20: Enterprise storage market (US)

40.0 40.0 0.66 0.80 35.0 0.41 35.0 0.36 30.0 0.31 0.69 30.0 25.0 25.0 0.21 0.32 0.42 20.0 20.0 0.15 0.16

0.31 Exabytes 15.0 Exabytes 15.0 10.0 10.0 5.0 5.0

0.0 0.0

2013Q1

2013Q2

2013Q3

2013Q4

2014Q1

2014Q2

2014Q3

2014Q4

2015Q1

2015Q2

2015Q3

2015Q4

2016Q1

2016Q2

2013Q1

2013Q2

2013Q3

2013Q4

2014Q1

2014Q2

2014Q3

2014Q4

2015Q1

2015Q2

2015Q3

2015Q4

2016Q1 2016Q2

HDD Internal Storage Capacity Flash HDD Internal Storage Capacity Flash

Source: IDC, Credit Suisse Source: IDC, Credit Suisse

Figure 21: Enterprise SSD shipments by capacity Figure 22: Enterprise SSD shipment mix by capacity Enterprise shipment by capacity Enterprise shipment by capacity 1,200 100%

1,000 80% 800 60% 600 40%

400

Units (k pcs)Units(k Units (k pcs) (kUnits 200 20%

0 0%

Jul Jul

Jul Jul

Jul Jul

Oct Oct

Oct Oct

Oct Oct

Oct Oct

Apr Apr

Apr Apr

Apr Apr

Apr Apr

Jan Jan

Jun Jun

Jan Jan

Jun Jun

Jan Jan

Jun Jun

Jan Jan

Mar Mar

Mar Mar

Mar Mar

Mar Mar

Feb Feb

Feb Feb

Feb Feb

Feb Feb

Aug

Aug Sep

Sep

Nov

Nov Dec Dec

Aug

Aug Sep

Sep

Nov

Nov Dec Dec

Aug

Aug Sep

Sep

Nov

Nov Dec Dec

July

July Aug

Aug Sep Sep

May May

May May

May May

May May

June June 2013 2014 2015 2016 2013 2014 2015 2016 <100GB 101-199GB 200-399GB 400-599GB <100GB 101-199GB 200-399GB 400-599GB 600GB< 600-999GB 1TB-1.9TB >2TB 600GB< 600-999GB 1TB-1.9TB >2TB

Source: TSR, Credit Suisse Source: TSR, Credit Suisse

Technology sector 19 6 December 2016

Figure 23: Enterprise SSD shipments by maker Figure 24: Enterprise SSD share by maker Enterprise shipment by supplier Enterprise shipment by supplier 1,200 100%

1,000 80% 800 60% 600 40%

400

Units (k pcs) (k Units Units (k pcs) (k Units 200 20%

0 0%

Jul Jul

Jul Jul

Jul Jul

Apr Apr

Oct Oct

Apr Apr

Oct Oct

Apr Apr

Oct Oct

Apr Apr

Oct

Oct

Jan

Jan

Jun Jun

Jan Jan

Jun Jun

Jan Jan

Jun Jun

Jan

Jan

Mar Mar

Mar Mar

Mar Mar

Mar

Mar

Feb

Feb

Feb Feb

Feb Feb

Feb

Feb

Nov

Nov

Dec Dec

Nov Nov

Dec Dec

Nov

Nov

Dec Dec

July

July

Aug

Aug

Sep Sep

Aug Aug

Sep Sep

Aug

Aug

Sep Sep

Aug Aug

Sep

Sep

May May

May May

May May

May May

June June 2013 2014 2015 2016 2013 2014 2015 2016 Samsung Intel Sandisk (Fusion-io) WDC (STEC) Others Samsung Intel Sandisk (Fusion-io) WDC (STEC) Others

Source: TSR, Credit Suisse Source: TSR, Credit Suisse

Figure 25: PC SSD shipments by capacity Figure 26: PC SSD shipment mix by capacity PC by density PC by density 6,000 100% 90% 5,000 80% 4,000 70% 60% 3,000 50% 2,000 40% 30% Units (k pcs) (k Units 1,000

Units (k pcs) (k Units 20%

0 10%

Jul Jul

Jul 0%

Oct

Oct

Oct

Oct

Apr

Apr

Apr

Apr

Jan

Jun

Jan

Jun

Jan

Jun

Jan

Mar

Mar

Mar

Mar

Feb

Feb

Feb

Feb

Nov

Dec

Nov

Dec

Nov

Dec

July

Aug

Sep

Aug

Sep

Aug

Sep

Aug

Sep

May

May

May

May

June

Jul Jul Jul

Apr Apr Oct Oct Apr Oct Apr Oct

Jan Jan Jun Jan Jun Jun Jan

Mar Feb Feb Feb Mar Feb Mar Mar

Dec Nov Nov Dec Nov Dec July

Aug Sep Aug Sep Aug Sep Aug Sep

May May May May

2013 2014 2015 2016 June <32GB 33-64GB 65-128GB 129-256GB 256GB 512GB >1TB 2013 2014 2015 2016 <32GB 33-64GB 65-128GB 129-256GB 256GB 512GB >1TB Source: TSR, Credit Suisse Source: TSR, Credit Suisse

Figure 27: PC SSD shipments by maker Figure 28: PC SSD share by maker PC shipment by supplier PC shipment by supplier 6,000 100%

5,000 80% 4,000 60% 3,000 40%

2,000

Units (k pcs) (k Units Units (k pcs) (k Units 1,000 20%

0 0%

Jul

Jul

Jul Jul

Jul Jul

Apr

Apr

Oct

Oct

Apr Apr

Oct Oct

Apr Apr

Oct Oct

Apr Apr

Oct Oct

Jan

Jan

Jun

Jun

Jan Jan

Jun Jun

Jan Jan

Jun Jun

Jan Jan

Feb

Mar

Feb Mar

Feb

Mar

Feb Mar

Feb

Mar

Feb Mar

Feb

Mar Feb Mar

Aug

Sep

Aug

Nov Sep

Dec

Nov Dec

Aug

Sep

Aug Sep

Nov

Dec

Nov Dec

Aug

Sep

Aug Sep

Nov

Dec Nov Dec

July

Aug July

Sep Aug Sep

May

May

May May

May May

May May

June June 2013 2014 2015 2016 2013 2014 2015 2016 SanDisk Intel Toshiba Samsung Micron Lite-On IT Others SanDisk Intel Toshiba Samsung Micron Lite-On IT Others

Source: TSR, Credit Suisse Source: TSR, Credit Suisse Chinese manufacturer's entry into NAND A number of media sources have reported on capital alliances with and technology licensing to Yangtze River Storage Technology (YRST). However, commenting on technology licensing talks with Micron Technology reported by the Nikkei on 18 November, YRST has denied that its chairman has signed any agreement. This is an unusual step, and we can only attribute it to political factors. Negotiations continue with a number of

Technology sector 20 6 December 2016

Japanese and US manufacturers, but we think more time will be needed owing to authorization issues. Local Chinese makers A groundbreaking ceremony for Wuhan Xinxin Semiconductor Manufacturing’s (XMC’s) have yet to begin Wuhan plant, which is to be YRST’s manufacturing base, was held in March 2016, but a building factories subsequent location change has stalled all progress on construction. Work is now to start after a second groundbreaking ceremony, but seasonal factors put this in spring 2017 at the earliest, suggesting mass-production is unlikely to ramp up before 2H 2018 even if alliances with NAND manufacturers and other partners go ahead.

Figure 29: NAND makers' technology roadmap: We had expected three companies to start mass production in 2Q 2017, but US makers are likely to delay

Mass Production Samsung Toshiba Micron Hynix Commercial Sample

1Q 14 2Q 14 3Q 14 4Q 14 1Q 15 2Q 15 3Q 15 4Q 15 1Q 16 2Q 16 3Q 16 4Q 16 1Q 17 2Q 17 3Q 17 4Q 17

3D NAND 24 Layers

32 Layers

48 Layers

64 Layers

Source: Credit Suisse estimates Logic makers, foundries, discretes: Foundry utilization bottomed in Sep–Oct Key takeaways

■ Capacity utilization at Taiwanese foundries reached a low at 85–90% in Sep–Oct, started to pick up in November and has now risen to around 95%. We expect utilization to stay high in 1Q 2017. Taiwanese foundries are operating at full capacity in 28/40nm and other process, and 16nm utilization has been picking up due mainly to additional demand for North American smartphones. Mass-production at 10nm has started.

■ 300mm capacity utilization at Korean LSI makers has been tracking at 100% or higher, supported by growth in smartphones equipped with in-house-manufactured 14nm process APs and the start of mass-production at 10nm.

■ Yields using 10nm process at Taiwanese and Korean makers have risen to 40–50%.

■ Taiwanese foundries have already taken delivery of equipment for test production lines for 7nm process, with mass-production due to start in 2H 2017. Korean LSI makers plan to initiate mass-production at 7nm around a year later at end-2018.

■ Supply of LSIs for power and small-signal applications is tight. We understand that some Japanese makers, in their shipments to electronics manufacturing services (EMSs) and independent design houses (IDHs), have started to focus supply on white- label brands, where the final customer is unknown.

Technology sector 21 6 December 2016

Utilization at Taiwanese foundries: a trough but no major correction Taiwanese foundries’ Taiwanese foundries’ capacity utilization at 300mm started to correct in September, 300mm capacity dropping to 85% in 4Q. However, utilization this turned out to be a trough for Sep–Oct, and utilization rebounded in recovering resumed in November without a major correction. We expect utilization to stay November; likely to high in 1Q 2017 and later. remain high in 1Q The largest correction was in 16nm process, where utilization fell to 65–70% in Sep–Oct. However, we gather additional orders for North American smartphones and shipments to Hisilicon, Nvidia, and others have been firm since November. We understand capacity utilization in 28nm process, products mainly shipped to Chinese smartphone makers, stayed at 100%. This was particularly true of shipments to Mediatek, and there were no signs of a correction. Correction risk remains a concern for all suppliers. However, as forecasts are calling for high inputs, we can say 2016 had virtually no corrections. Assuming smartphone volume will have increased by 100mn units in 2016, an increase in wafer production on the order of 15–20k/month would have been sufficient. However, we think concerns about overproduction of APs will not disappear as Taiwanese foundries have operated at full capacity throughout the year and Korean LSI makers have been running at full capacity in 2H to make APs at 14nm process. We are also concerned about utilization remaining high in Taiwanese foundries, as Spreadtrum's share among Korean smartphone makers has decreased. As the risk has receded of a substantial correction in production of Chinese smartphones until production gets underway for spring 2017 new models, we intend to focus on this risk from May 2017 onward. There have tended to Since 2010, there have been major cyclical corrections, perhaps coincidental, in utilization be major corrections in at Taiwanese foundries in odd-numbered years. First in 2011, Chinese notebook PC odd-numbered years volume growth slowed. Then in 2013, there were inventory adjustments in Chinese 3G and Samsung Electronics (005930 KS) high-end smartphones. This was followed in 2015 by wide-ranging inventory adjustments in Chinese LTE smartphones, and in consumer electronics and industrial products. This end demand for smartphones is not strong enough to warrant upward revisions to existing forecasts, while production of some advanced smartphone APs has shifted to Korean LSI makers. In view of these developments, the potential for utilization adjustment at Taiwanese foundries, including distributor inventories, appears to be building.

Technology sector 22 6 December 2016

Figure 30: Projected capacity utilization for Taiwanese foundry 300nm/200nm lines (input basis)

1Q10

2Q10

3Q10

4Q10

1Q11

2Q11

3Q11

4Q11

1Q12

2Q12

3Q12

4Q12

1Q13

2Q13

3Q13

4Q13

1Q14

2Q14

3Q14

4Q14

1Q15

2Q15

3Q15

4Q15

1Q16

2Q16

3Q16

4Q16 1Q17 120% 120% FORECAST

110% 110%

100% 100%

90% 90%

80% 80%

70% 70% Rates Utilization Quarterly

60% 60% TSMC Monthly Utilization rate (300/200mm)rate Utilization Monthly

50% 50%

Jul.

Jul.

Jul.

Jul.

Jul.

Jul.

Jul.

Apr.

Oct.

Apr.

Oct.

Apr.

Oct.

Apr.

Oct.

Apr.

Oct.

Apr.

Oct.

Apr.

Oct.

Jan.

Jun.

Jan.

Jun.

Jan.

Jun.

Jan.

Jun.

Jan.

Jun.

Jan.

Jun.

Jan.

Jun.

Jan.

Feb.

Mar.

Feb.

Mar.

Feb.

Mar.

Feb.

Mar.

Feb.

Mar.

Feb.

Mar.

Feb.

Mar.

Feb.

Mar.

Aug.

Sep.

Nov.

Dec.

Aug.

Sep.

Nov.

Dec.

Aug.

Sep.

Nov.

Dec.

Aug.

Sep.

Nov.

Dec.

Aug.

Sep.

Nov.

Dec.

Aug.

Sep.

Nov.

Dec.

Aug.

Sep.

Nov.

Dec.

May.

May.

May.

May.

May.

May. May. 2010 2011 2012 2013 2014 2015 2016 2017

300mm 200mm TSMC quarterly Utilization rates

Source: Credit Suisse estimates

Figure 31: Projected quarterly utilization rates for Figure 32: Projected quarterly utilization rates for major Taiwanese foundry 300/200nm lines (input major Taiwanese foundry 300/200nm lines, by basis) process (input basis)

110% 110% 99% 100% 100% 91% 100%97% 99% 89% 90% 90% 95% 96% 91%92% 91%91% 78% 80% 80% 88% 75% 78% 70% 70% 76% 60% 60% 50% 50% 40% 40% 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2013 2014 2015 2016 2017

2013 2014 2015 2016 2017 90nm 65nm 40nm 300mm Fab 200mm Fab 28nm Poly-SiN 28nm HKMG 20/16nm 10nm

Source: Credit Suisse estimates Source: Credit Suisse estimates The major Korean LSI maker: Still operating above full capacity The major Korean LSI maker's 300mm lines are even busier than those of Taiwanese foundries. Korean LSI makers’ This manufacturer has been operating at full capacity, especially in 10nm/14nm. Input has 10/14nm process are been tracking at around 115% of capacity, and we expect utilization to stay high in 1Q going strong 2017. We attribute this to production by Qualcomm and the company’s production of the Exynos (its proprietary 10nm process AP) for the Galaxy S8, and increased adoption Exynos for the 14nm process in its mid-range and low-end smartphone models. We

Technology sector 23 6 December 2016

thought the company would use its own APs in around 33% of its 2016 smartphone models. However, this proportion is now expected to reach 50% for the year (vs. around 22% in 2015). We expect this figure to surpass 60% in 2017. Meanwhile, low-end Spreadtrum chipset adoption has sharply declined, which we see as a factor depressing utilization at Taiwanese foundries (in practice, 40nm process utilization has remained high, possibly due to offsetting elsewhere). We think there could be a shortage of 14nm/10nm process production capacity if adoption of the company's Exynos AP in Chinese smartphones increases (it is currently used in only some models, such as those made by Lenovo and Meizu) or if there is volume growth at Qualcomm. As Line 17, the next base for LSI production capacity, has been returned to DRAM/NAND production, we understand the company is considering setting up its next LSI line at its Pyeongtaek fab in Korea (now under construction). We think this decision will be determined by how various factors impact production capacity, including 14/10nm volume growth, foundry customer base expansion, and Qualcomm switching to a different foundry for 7nm process production (10nm production is currently handled by the major Korean LSI maker). Mediatek has started to use its APs in its smartphones, and while contracted AP production (on condition of use in the company's own smartphones) looks possible, this has not yet been decided. As adoption in its own smartphones in 2017 will be limited to 10mn units, which means volume of 2k/month will be sufficient, given currently tight production capacity, we think contracting out production to foundries still looks unlikely.

Figure 33: Korean LSI 14/20nm process capacity Figure 34: Samsung Electronics' share of utilization smartphone AP

140% 100% 2% 3% 6% 3% 90% 19% 120% 80% 34% 31% 70% 100% 60% 36% 80% 50% 40% 38% 60% 30% 63% 20% 44% 40% 10% 22% 20% 0% 2015 2016 2017

0% Samsung Qualcomm Spreadtrum Marvell Mediatek

Jul

Jul

Apr

Oct

Apr

Oct

Jun

Jan

Jun

Jan

Mar

Mar Feb

Feb

Aug

Sep

Nov

Dec

Aug

Sep

Nov

Dec

May May 2015 2016 2017

Source: Credit Suisse estimates Source: Credit Suisse estimates

Regarding the major Korean LSI maker's CIS line, the survey conducted three months ago led us to believe the company planned to expand 300mm production capacity by around 15%. However, as 10/14nm process AP production has been very tight recently and the company has decided not to adopt dual cameras in its spring model smartphones, the company plans to expand production capacity for CIS via 200mm wafers, where utilization is low, rather than via 300mm wafers. For the same reasons, the company plans to expand display driver IC (DDI) production capacity via 200mm wafers. However, due to sensor size, high-end units cannot be made using 200mm wafers. We understand that the company is therefore considering using existing memory lines and its Pyeongtaek fab among possible options to expand CIS production capacity using 300mm wafers. The company's LSI division could split in 2017 into separate LSI manufacturing and foundry divisions. We think foundry operations would expand the customer base from Qualcomm and nVidia, who we believe are current customers, to include companies such as Mediatek and AMD (ARM servers).

Technology sector 24 6 December 2016

7nm process development/mass-production plans Taiwanese foundries Taiwanese foundries are already setting up 7nm test lines. They plan to set up mini lead Korean LSI makers production lines (several thousand wafers/month) for certification purposes in 1H 2017 by around a year for also at mass-production plants. Mass-production is slated to commence from end-2017 7nm process through 1H 2018. For the major Korean LSI maker, 2016 was a significant year with respect to the continuation of LSI business, which explains why it focused resources on near-term business. We believe it is lagging rivals in 7nm process development and is therefore planning mass-production at 7nm at end-2018, around a year later than Taiwanese foundries. If so, in addition to 7nm process APs for the major North American smartphone maker’s 2018 models, we see a risk that orders could flow to Taiwanese foundries if the timing of mass-production at the company did not match that of the shift to 7nm process production at Qualcomm, which has secured orders from foundries at 10nm. We therefore intend to focus on the start of 7nm production at the major Korean LSI maker and Qualcomm's 7nm process roadmap.

Figure 35: Smartphone application processor maker technology roadmaps

2014 2015 2016 2017 2018 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q

Apple A8 A9 A10 A10X A11 A11X? A12 (20nm) (16nm) (16nm) (10nm) (10nm) (7nm) (7nm)

MSM8994 Qualcomm MSM8996 MSM8998 (20nm) (14nm) (10nm)

Exynos 8890 Exynos XXXX Samsung Exynos 5422 Exynos 7420 Exynos (7nm) (28nm) (14nm) (14nm) (10nm)

X30 Mediatek X10/MT6795 X20/MT6797 (28nm) (20nm) (10nm)

Kirin 920 Kirin 930 Hi Silicon Kirin 950 Kirin 960 (28nm) (28nm) (16nm) (10nm)

Source: Credit Suisse estimates

Tight supply of power and small-signal discretes Small signal discretes We understand that for discretes, the supply-demand balance has started to tighten in short supply due to across the board due mainly to falling supply capacity caused by semiconductor supply-side factors; restructuring and shortages in supply from foundries at some power makers. We get the EMS firms unable to impression that shortages in TVs and other consumer electronics have started to stand out. procure enough Makers of discretes have started to focus supply to white-label brands, where the final customers (EMSs and IDHs) are invisible and where we understand they are allocating supply. In pricing, we have started to see prices being left unchanged whereas previously price cuts came as a matter of course.

Technology sector 25 6 December 2016

Semiconductor capex Front-end SPE investment: Taiwanese foundries continued to invest in 10nm while starting to fund development of 7nm lines; delivery of equipment to the Korean 3D NAND maker's new line is due to start in December Key takeaways

■ In leading-edge investment by Taiwanese foundries, we expect expansion of 10nm process production will continue through 2Q 2017. Companies are installing equipment on test 7nm process lines. Investment in mini production lines at plants that will manufacture at 7nm process is due in 1H 2017.

■ The major Korean LSI maker has decided to invest $1bn in its Austin, Texas plant (S2 line). The investment is mainly in conversion to 10nm process production.

■ Equipment delivery to the largest Korean NAND maker's new 3D NAND plant is due to start in December. We gather concrete plans for delivery have been drawn up through mid-2017.

■ The largest Korean maker's Pyeongtaek plant was initially planned to start up 3D NAND production. However, as LSI production capacity shortages and tight DRAM supply-demand are expected, we gather the company is considering DRAM or perhaps LSI lines in Phase 2.

■ The Korean NAND maker is considering two proposals: transfer of the M12 line from planar and a new M14 line. However, 3D NAND acceptance is still outstanding and the technology is lagging; therefore it has not adopted an aggressive investment stance.

■ For investment in China, Powerchip plans to deliver equipment to its plant in Hefei in 2Q 2017. Equipment will start being delivered to TSMC's Nanjing fab in August 2017. However, it mainly plans to transfer 20/16nm equipment from Taiwan (production is due to start in November). A groundbreaking ceremony was held at XMC Wuhan in March. However, as the actual location of the plant was changed, we gather a second ceremony is due to be held. Capex at Taiwanese foundries: Active investment in 10nm mass- production, 7nm test lines Taiwanese foundries to Among Taiwanese foundries, companies such as Mediatek and Hisilicon are due to start invest in 10nm process mass-production at 10nm in 1H 2017, beginning with iPad APs. Production lines providing expansion through 1H capacity of around 35k wafers/month have already been built. Capacity, including that for CY17; also initiating APs for 2017 North American smartphones, is due to be expanded by around 20k investment in wafers/month in 1H 2017. We expect expansion of 10nm process production line capacity prototyping lines for to come to a halt at around 55k wafers/month. 7nm process For next-generation 7nm process, equipment is currently being delivered for test lines with capacity of a few thousand wafers/month, which will be used as development lines. TSMC's Fab 15 P/5/6 in Taichung, Taiwan (7nm process mass-production line) is very near completion. The company is expected to invest in a mini-line (capacity of several thousand wafers/month) for factory-certification purposes in 1H 2017. As delivery of equipment for 10nm mass-production process will continue until 2Q 2017, we think delivery for 7nm process mass-production lines will not take place until 3Q 2017 at the earliest. Thus, as with 10nm process whose equipment orders started to be placed

Technology sector 26 6 December 2016

in March 2016, we expect orders for 7nm mass-production process equipment to start being placed in early spring 2017. Compatibility of 10nm process equipment with 7nm devices is said to be 95%. However, we think that when 7nm process enters mass production, all equipment will be new and equipment orders will be on par with initial 10nm investment.

Figure 36: Taiwanese foundry 7nm process mass Figure 37: Taiwanese foundry 7nm process mass production factory (as of end-Nov 16) production factory (as of end-Nov 16)

Source: Credit Suisse Source: Credit Suisse Capex at the major Korean LSI maker: 10nm expansion, fresh expansion of CIS lines also needed Korean LSI makers As noted above, the major Korean LSI maker has continued to operate at full capacity in expanding 10nm leading-edge process due mainly to upping use of its own Exynos APs in its smartphones process; more capacity and securing Qualcomm 14/10nm process orders. In addition to the 300mm lines in Korea, also needed for CIS the company is due to invest in converting its Austin, Texas fab to 10nm process. We think the possibility of capacity shortages could arise if the company beefs up foundry business and secures customers. We understand that because the company is already using Line 17, which was originally built to make LSIs, for DRAM/NAND, it is also considering using Phase 2 at its Pyeongtaek fab for LSI production. CIS demand is expected to increase due to growing adoption of dual cameras in smartphones, and we gather the company is also considering partly converting existing memory lines to CIS production. 3D NAND capex at the largest Korean NAND maker The largest Korean NAND maker is investing in the conversion of its current Line 16 to 3D NAND, and taking delivery of equipment for 3D NAND process at Line 17. We expect equipment to start being delivered to the company's Pyeongtaek fab in December. Apart from equipment delivery being brought forward by around a month, we believe everything, including the scale of investment, is running according to plan. At end-November, external walls were not yet completed, as were facilities including the vendor rooms where equipment makers maintain a permanent presence. However, equipment delivery is due to start nevertheless. A 48-layer 3D NAND line with capacity of 30–40k wafers/month is due to be built in 1H 2017, and we understand that plans for equipment delivery have already been drawn up. We think that if 64-layer 3D NAND development is completed, equipment for 64-layer lines will be delivered in 2H 2017.

Technology sector 27 6 December 2016

Figure 38: Pyeongtaek plant under construction – Figure 39: Pyeongtaek plant under construction – image from end-November image from end-November

Source: Credit Suisse Source: Credit Suisse

Figure 40: Pyeongtaek plant under construction – Figure 41: Pyeongtaek plant under construction – image from end-August image from end-August

Source: Credit Suisse Source: Credit Suisse

Capex in China New Chinese There are three major investments in mainland China by Taiwanese makers. At UMC's investment includes Amoy fab, equipment delivery for a 40nm process line with capacity of around 10k equipment installation wafers/month is almost complete, and we expect production to start. by two Taiwanese chipmakers from CY17; At Powerchip's Hefei fab, where it makes LCD driver chips, equipment delivery is planned local Chinese memory for 2Q 2017. maker held Equipment for TSMC's Nanjing plant is due to start being delivered in August 2017. The groundbreaking company plans mainly to relocate 20nm/16nm process equipment from its plant in Taiwan. ceremony in March for Production is due to commence in November. If it is targeting Chinese smartphone new plant, but site was makers, as there are almost no 20nm process APs apart from those at Apple, it would only subsequently changed be able to transfer two-generation-old process technology. We therefore think the and ceremony will be held again company will get 16nm process production off the ground together with 7nm process mass-production on the Taiwan side. XMC (holding company Yangtze River Storage Technology) held a groundbreaking ceremony in March 2016 in Wuhan. Construction has not yet started, however, due to a change in the fab's planned location. Since construction is unlikely to start in winter, we

Technology sector 28 6 December 2016

think it will commence in spring next year at the earliest. In this case, we think the fastest schedule would be completion in 1H 2018, with mass-production starting at end-2018. Assuming XMC's technology partner is decided upon, we think equipment orders would start to be placed in 2018. We believe investment in Chinese local makers, which SPE makers were expecting, is taking slightly longer to come on stream than expected. At Intel Dalian, Phase 1, which commenced in spring 2016, is expected to be following by Phase 2 investment in 1H 2017. SMIC has continued to expand 28nm process production at its Beijing fab. Front-end SPE maker orders Front-end SPE orders In front-end SPE order dates and projects during 4Q 2016, we expect 10nm process likely to remain high in mass-production lines and 7nm process test lines at Taiwanese foundries, and 3D NAND Oct–Dec and Jan–Mar production by the largest Korean maker at its Pyeongtaek fab. centered on 3D NAND; 7nm process could We think orders will remain firm as we anticipate ongoing 3D NAND production at come into play from Pyeongtaek, investment in 3D NAND by Intel Dalian and Japanese NAND makers, and Apr–Jun completion of Powerchip's Hefei fab in 1Q 2017. After that, we see 7nm process orders from Taiwanese foundries. However, we see little likelihood of other major investments until specific plans emerge in China.

Technology sector 29 6 December 2016

Figure 42: Investment in new semiconductor lines 2016 2017 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q

Foundry/Logic 10nm FinFET Fab12 P7 10nm FinFET Fab15 P5 7nm FinFET Fab15 P6 16nm FinFET+ Nanjing

Samsung 10nm FinFET S2 (Austin)

Intel 10 nm Israel

UMC 40/28nm XiaMen

Powerchip 90nm Hefei

DRAM Samsung Line 17 * 1xnm process migration in 2H 2016

SK Hynix M14

Micron 1xnm ex-Elpida Hiroshima * 1xnm migration pushed out to 1Q17 ex-Rexchip * 1xnm process migration in 1Q17

Nanya Fab3AN

Winbond ?

NAND Samsung 3D NAND 48 Layers Xi-an Line 16 ** Conversion from DRAM to Planer/3D. 3D NAND 48 Layers Line 17 3D NAND 64 Layers Line 18 (Pyongteak)

Toshiba 3D NAND 48 Layers Fab2 3D NAND 64 Layers

Micron 3D NAND 32 Layers Singapore 3D NAND 64 Layers Singapore ? (maybe floating gate structure)

SK Hynix 3D NAND 48 Layers ? 3D NAND 72 Layers ? 14nm M12 * 14nm process migration

Intel 3D NAND, 3D X-Point Fab68, Dalian

Timing of order placement Equipment installation R&D / Pilot Production Mass production Source: Credit Suisse estimates

Technology sector 30 6 December 2016

Figure 43: WFE capex forecasts (out Japan Team forecast)

$40.0 34.5 $35.0 33.3 32.1 31.9 6.5 $30.0 28.1 6.7 27.7 5.7 6.0

$25.0 6.1 5.4 12.5 11.0 11.9 $20.0 14.3

$15.0 14.4 13.1 7.0 $10.0 10.0 5.9 9.2 5.3 $5.0 4.7 7.9 6.2 5.5 5.5 3.0 3.9 $0.0 2012 2013 2014 2015 2016E 2017E

DRAM NAND Foundry Logic/Other

Source: Company data, Credit Suisse estimates Back-end SPE investment Key takeaways

■ Capacity utilization has stayed high at Taiwanese OSATs. Consequently, competition has been absent for orders arising from memory assembly demand, as seen in 3Q, and capex has stalled.

■ We anticipate DRAM tester demand due to higher speeds in mobile DRAM from 3Q 2016 and adoption of LP DDR4 in mid-range smartphones in 2H 2017.

■ In probers and testers for the largest Korean NAND maker's new fab, we expect order placement to start in 1Q 2017 with back-end equipment orders coming in from 2Q.

■ We expect Shinko Electric's MCeP technology to continue to be used in Qualcomm's 10nm process AP packages. In Fan-Out technology, adoption in 7nm process (mass- production due to start at end-2018) is under consideration. We expect the direction to be determined in 1H 2017.

■ We understand that no progress has been made in FO-PLP (Fan-Out Panel-Level Packaging) development at the major Korean LSI maker and that adoption of in-house FO-PLP in 2017 10nm process APs will not meet requirements. We also gather that if there are senior management changes due to personnel moves at SEMCO in December, strategy could also change. Adoption of FO-PLP in in-house APs will take place in 2018 at the earliest. Initially, we expect mass-production for devices with small chip sizes such as power management ICs and modems.

■ We expect virtually no changes in TSMC's InFO specification in 2017, but do look for changes in 2018. Several development projects are currently under way.

■ We think fingerprint scanner sensor-related investment will continue in 2017 as major Taiwanese OSATs plan to increase such investment during the year. We expect a partial shift from 200mm to 300mm in foundry production.

■ In system-in-package (SiP) modules, metal mold manufacturing is currently difficult and expensive, and module manufacturing process have not been established. However, we expect these process to start getting off the ground in 2017.

Technology sector 31 6 December 2016

Capex at Taiwanese OSATs Taiwanese OSATs With regard to assembly capacity utilization at Taiwanese OSATs, foundry maker 300mm refraining from new capacity utilization (output) has stayed high, coming off a low in November, leading us to investment despite expect utilization to remain high in 4Q, and do not expect utilization to decline as much as high capacity it normally would for the season in 1Q 2017. However, we were unable to confirm a strong utilization; sole bright enough investment commitment that would result in capacity expansion. spot is fingerprint sensors In September, OSATs, which had been facing the risk of falling foundry capacity utilization starting in November, scrambled to meet memory-related assembly demand, which led to capital spending. However, memory-related investment has stalled of late. The only bright spot is ongoing investment in fingerprint scanner sensor-related assembly. This is due to the shift from 200mm to 300mm production lines and to an ongoing steady growth in demand. We expect major OSATs to invest more in fingerprint scanner sensors in 2017 than in 2016, making them the only application with a secure outlook.

Figure 44: Capacity utilization of Taiwanese OSAT utilization

Taiwan Assemblyprocess utilization

100%

90%

80%

70%

60%

50%

40%

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q 4Q 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 E 2017 E Assembly

Source: Company data, Credit Suisse estimates

Technology sector 32 6 December 2016

Figure 45: Forecast for ratio of smartphones with fingerprint recognition sensors

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0% 2014 2015 2016E 2017E 2018E

iPhone fingerprint penetration China and others smartphone fingerprint penetration

Source: Company data, Credit Suisse estimates

Fan-Out WLP/PLP: No major new developments; focusing on Korean LSI maker's package development strategy Few developments in For InFO-WLP at Taiwanese foundries, we expect 2017 mass-produced 10nm process FO-WLP/PLP; Korean APs to have broadly the same specification as before. We see little prospect of demand LSI makers to use for back-end SPE in particular. We expect major specification changes in 2018, with existing package several projects under consideration at the moment. We believe the projects being technology in CY17 considered include not only process migration (the shift from 5um to 2um in redistribution layers), but also other projects such as (1) InFO + TSV technology in which DRAM is incorporated in APs, not in PoPs (package on package), and connected via TSV and (2) technology for making packages thinner by mounting APs and DRAM horizontally rather than using PoP structures. We believe the Korean LSI maker is engaged in two projects in developing next- generation packages: both by itself and at subsidiary SEMCO. We think the company on its own is developing interposer PoP packages with structures similar to MceP from Shinko Electric Industries. SEMCO is engaged in FO-PLP development, but we understand that it could change its plans depending on developments within senior management. We think mass-production of FO-PLP APs will not start until 2018 at the earliest. We think the company will start of by accumulating experience in mass-production of devices with small chip sizes, such as power management ICs and modems. We gather that larger chips present technological issues. However, the group's smartphone division has shown no interest in adopting FO- WLP/PLP for APs used in premium models. Even if the group's smartphone division adopted FO-WLP/PLP, we would expect it to be in mid-range models, meaning that volume would be limited. However, if this technology were adopted in mid-range models, FO-PLP costs would have to be cut, leading us to the opinion that large PLP will be needed. However, there is currently no equipment compatible with large PLP, and we believe developing such equipment could take as long as two years).

Technology sector 33 6 December 2016

Qualcomm is also reluctant to adopt FO-WLP/PLP in APs as it is now using Shinko Electric's current MCeP technology in 10nm process packages. We understand the outlook for adoption of FO-WLP/PLP in 7nm process will solidify in 1H 2017. SiP modules: Expecting investment if adoption in smartphones increases We believe Amkor and STATS ChipPAC are handling the manufacturing of SiP modules for Korean LSI makers and that mass-production at STATS’s Korea base is already established. We expect mass-production at Amkor to commence in 2018. Passive component space for North American smartphones has contracted by 5% in 2016, while adoption of SiP modules has increased (in passive modules, the number of types increased from just one in 2015 models to four in 2016’s). In SiP modules, manufacturing of current dies for molding is difficult and expensive, making it difficult to establish process in module manufacturing. This manufacturing is therefore limited to certain customers’ process, but we expect a further take-off in 2017.

Figure 46: SiP market – Trends by application 2015-2017 2015E 2016E 2017E Volume (M unit) CAGR

Mobile Phones & Tablets Camera 3,700 3,900 4,200 6.5%

Fingerprint 400 700 800 41.4%

Connectivity 400 500 500 11.8%

Audio AMP 100 200 -

LCD Driver 100 200 -

PMIC 200 -

Front-end Module 5,700 5,900 6,200 4.3%

IoT & Wearable Connectivity 200 600 1,200 144.9%

PMIC 100

Total 10,000 11,500 13,400 15.8%

Source: Company data, Credit Suisse estimates

Memory tester market: We expect LP DDR4-related demand in 2H 2017, 3D NAND demand in Korea from 1Q Investment in DRAM In DRAM demand and technology changes, manufacturers are migrating Mobile RAM testers being driven by from LP DDR3 to LP DDR4, principally in flagship models, and we expect adoption in mid- high-speed LP DDR4 range units as well in 2H 2017. We also expect Chinese smartphone makers to adopt and LP DDR4’s use in high-speed LP DDR4, as they migrate from 3Gbps to 4Gbps models. For the twin reasons midrange Chinese of stronger demand and the shift to higher speeds, we thus expect DRAM tester demand smartphones to emerge starting in 3Q. However, despite our expectations for a peak-out of LP DDR4 volume in flagship models and for bit growth, driven by adoption of 8GB by Chinese smartphone makers and the equipping of 3GB in all 2017 models from the major North American smartphone maker, we generally believe that tester demand is unlikely to accelerate until adoption in 2H 2017 mid-range models takes off.

Technology sector 34 6 December 2016

As companies are reluctant to invest in migration to 1xnm process in DRAM, we see little prospect of process migration driving tester demand. Investment in NAND However, the correlation of bits and test time for NAND means that any bit growth due to testers low versus bit 3D NAND mass-production will likely lead to an increase in total test time. However, growth (shows comments by NAND makers leave us with the impression that they are reluctant to invest chipmakers using in testers in response to stronger bit growth. Nevertheless, we see demand for wafer existing testers); new probe test equipment in response to increased NAND production capacity. Specifically, we Korean NAND factories expect investment in expanding new capacity at the largest Korean NAND maker's likely to add little Pyeongtaek fab. As delivery of key front-end SPE equipment to this plant will start in demand December, we think equipment orders related to probe testers will start to be placed in 1Q 2017. Among NAND tester makers, local Korean company YIK, a former Korean subsidiary of Yokogawa Electric (6841), is also a presence, and we think it will secure nearly half the market for front-end testers with Advantest (6857) taking the remainder. We expect volume of around 40–50 systems. Since the scale of orders should be limited to around several billion yen, we think a substantial order boost looks unlikely. Teradyne, which has a solid track record, should win virtually all orders for back-end testers.

Figure 47: Disco – Quarterly order forecasts Figure 48: Advantest – Quarterly order forecasts

Jan-12

Apr-12

Jul-12

Oct-12

Jan-13

Apr-13

Jul-13

Oct-13

Jan-14

Apr-14

Jul-14

Oct-14

Jan-15

Apr-15

Jul-15

Oct-15

Jan-16

Apr-16

Jul-16

Oct-16

Jan-17

Apr-17

Jul-17

Oct-17

Jan-12

Apr-12

Jul-12

Oct-12

Jan-13

Apr-13

Jul-13

Oct-13

Jan-14

Apr-14

Jul-14

Oct-14

Jan-15

Apr-15

Jul-15

Oct-15

Jan-16

Apr-16

Jul-16 Oct-16 45.0 2.25 60.0 3.0 40.0 35.5 2.00 49.2 50.2 35.034.5 50.0 46.2 45.0 2.5 35.0 33.4 33.3 32.0 31.3 32.1 44.0 43.8 30.8 31.0 1.75 39.5 39.840.6 28.4 27.5 39.138.2 38.7 30.0 26.626.1 26.1 28.0 28.0 40.0 35.037.0 2.0 24.1 23.4 34.3 25.2 25.0 25.0 1.50 29.8 32.0 25.0 25.3 27.4 28.5 21.8 19.3 30.0 27.1 28.0 1.5 20.0 24.4

1.25 21.4 Orders (¥ bn) (¥ Orders

Order (¥ (¥ bn) Order 15.0 20.0 1.0

1.00 (x) P/B rel TOPIX TOPIX rel P/B (x) P/B rel TOPIX 10.0 10.0 0.5 0.75 5.0 0.0 0.0

0.0 0.50

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

4QE

1QE

2QE

3QE

4QE

1Q

1Q

2Q

3Q

4Q

2Q

2Q

3Q

4Q

4Q

1Q

2Q

3Q

4Q

1Q

2Q

3Q

1Q

3Q

4Q E

1Q E

2Q E 3QE 4Q E CY2012 CY2013 CY2014 CY2015 CY2016 CY2017 CY2012 CY2013 CY2014 CY2015 CY2016 CY2017 Order TOPIX rel P/B Order TOPIX rel P/B

Source: Company data, Credit Suisse estimates Source: Company data, Credit Suisse estimates

Figure 49: Tokyo Seimitsu – Quarterly SPE order forecasts Figure 50: TOWA– Quarterly order forecasts

14,000 Consolidated order

12,000 10,000 9,000 10,000 8,000 7,000 8,000 6,000 5,000

6,000 mn JPY mn million JPY million 4,000 4,000 3,000 2,000 2,000 1,000 0

0

1Q12 1Q14 4Q14 4Q15 2Q12 3Q12 4Q12 1Q13 2Q13 3Q13 4Q13 2Q14 3Q14 1Q15 2Q15 3Q15 4Q15 1Q16 2Q15 3Q15 1Q16 2Q16 3Q16

1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q

4Q16E 1Q17E 2Q17E 3Q17E 4Q17E

4Q E 4Q E 1Q E 2Q E 3Q E 4Q CY2012 CY2013 CY2014 CY2015 CY2016 CY2017 Consolidated order

Source: Company data, Credit Suisse estimates Source: Company data, Credit Suisse estimates

Technology sector 35 6 December 2016

Figure 51: Micronics Japan – order forecasts

35,000

30,000 4,420 4,233 25,000 7,926 6,000 6,000 4,690 5,088 20,000 3,550

mn JPY mn 15,000 27,618 25,460 21,159 21,600 21,900 10,000 19,017 20,275 16,829 5,000

0 11/09 12/09 13/09 14/09 15/09 16/09 17/9 18/9 CSE CSE CSE

Probe card TE

Source: Company data, Credit Suisse estimates OLED investment

Canon Tokki’s vapor ■ We see little in the way of packaged sales of steppers and vapor deposition equipment. deposition systems not the industry standard ■ Chinese display makers CSOT and Tianma plan to install Tokki and Ulvac's vapor deposition solutions to one production line each for evaluation. Korean makers will have to adopt local solutions as they cannot expect supply from Canon Tokki in 2017.

■ Korean makers are struggling to bring Canon Tokki’s vapor deposition systems on line, and Chinese display makers do not regard them as the standard process either.

Technology sector 36 6 December 2016

Companies Mentioned (Price as of 05-Dec-2016) Adata (3260.TWO, NT$53.9) Advanced Micro Devices, Inc. (AMD.OQ, $8.68) Advantest (6857.T, ¥1,536) Alibaba Group Holding Limited (BABA.N, $90.99) Amkor Technology Inc. (AMKR.OQ, $11.24) Apple Inc (AAPL.OQ, $109.11) Baidu Inc (BIDU.OQ, $164.32) Canon Tokki (Unlisted) China Star Optoelectronics Technology (Unlisted) DISCO (6146.T, ¥13,050) Hi Silicon (Unlisted) Hitachi Kokusai Electric (6756.T, ¥2,176, OUTPERFORM, TP ¥1,980) Huawei Technologies (Unlisted) Intel Corp. (INTC.OQ, $34.39) JEOL (6951.T, ¥487) Lasertec (6920.T, ¥1,990) Lenovo (LNVGY.PK, $30.685) Lenovo Group Ltd (0992.HK, HK$4.77) Lite-On Technology (2301.TW, NT$49.3) Marvell Technology Group Ltd. (MRVL.OQ, $14.03) MediaTek Inc. (2454.TW, NT$223.5) Meizu (Unlisted) Memblaze (Unlisted) Micron Technology Inc. (MU.OQ, $18.61) Micronics Japan (6871.T, ¥979) Mizu (Unlisted) NVIDIA Corporation (NVDA.OQ, $91.88) Nanya Technology (2408.TW, NT$44.6) NuFlare Technology (6256.T, ¥5,510) Phison Electronics (8299.TWO, NT$253.5) Powerchip (5346q.L, $0.0) QUALCOMM Inc. (QCOM.OQ, $66.36) SCREEN (7735.T, ¥6,400) SK Hynix Inc. (000660.KS, W44,400) SMIC (SMI.N, $6.44) Samsung Electronics (005930.KS, W1,718,000) SanDisk Corp. (SNDK.OQ^E16) SanDisk Corp. (SNDK.OQ^E16) SanDisk Corp. (SNDK.OQ^E16) Semiconductor Manufacturing International Corp. (0981.HK, HK$1.0) Shannon Systems (Unlisted) Shinko Electric Industries (6967.T, ¥701) Spreadtrum Communications (Unlisted) Stats Chippac (STTS.SI^J15) Stats Chippac (STTS.SI^J15) Stats Chippac (STTS.SI^J15) Stats Chippac (STTS.SI^J15) TOWA (6315.T, ¥1,341, OUTPERFORM[V], TP ¥1,680) Taiwan Semiconductor Manufacturing (2330.TW, NT$179.0) Tencent Holdings (0700.HK, HK$190.0) Tenma (7958.T, ¥1,910) Teradyne Inc. (TER.N, $24.27) Tokyo Electron (8035.T, ¥10,000, OUTPERFORM, TP ¥10,200) Tokyo Seimitsu (7729.T, ¥3,165) Toshiba (6502.T, ¥436, OUTPERFORM[V], TP ¥460) Ulvac (6728.T, ¥3,360) United Microelectronics (2303.TW, NT$11.2) Western Digital Corp. (WDC.OQ, $62.09) Winbond (2344.TW, NT$9.69) XMC (Unlisted) Yangtze River Storage Tech (Unlisted) Yokogawa Electric Corp (6841.T, ¥1,595) Please see figure 2 and 15 for other companies mentioned.

Disclosure Appendix Analyst Certification Hideyuki Maekawa, Akinori Kanemoto and Mika Nishimura each certify, with respect to the companies or securities that the individual analyzes, that (1) the views expressed in this report accurately reflect his or her personal views about all of the subject companies and securities and (2) no part of his or her compensation was, is or will be directly or indirectly related to the specific recommendations or views expressed in this report.

Technology sector 37 6 December 2016

3-Year Price and Rating History for Hitachi Kokusai Electric (6756.T)

6756.T Closing Price Target Price Date (¥) (¥) Rating 18-Dec-13 1,409 1,470 O 28-Mar-14 1,211 1,560 08-Oct-14 1,546 1,370 N 26-Jan-15 1,811 1,650 18-Feb-15 1,564 1,640 17-Aug-15 1,675 1,550 01-Dec-15 1,850 1,650 28-Jun-16 1,719 1,980 O * Asterisk signifies initiation or assumption of coverage. OUTPERFORM NEUTRAL

3-Year Price and Rating History for TOWA (6315.T)

6315.T Closing Price Target Price Date (¥) (¥) Rating 07-Nov-16 1,153 1,580 O * 16-Nov-16 1,326 1,680 * Asterisk signifies initiation or assumption of coverage.

OUTPERFORM

3-Year Price and Rating History for Tokyo Electron (8035.T)

8035.T Closing Price Target Price Date (¥) (¥) Rating 28-Apr-14 5,827 6,150 O 08-Oct-14 6,775 7,300 N 17-Aug-15 6,496 6,170 01-Dec-15 8,427 7,610 28-Jun-16 8,079 10,200 O * Asterisk signifies initiation or assumption of coverage.

OUTPERFORM NEUTRAL

Technology sector 38 6 December 2016

3-Year Price and Rating History for Toshiba (6502.T)

6502.T Closing Price Target Price Date (¥) (¥) Rating 07-May-14 390 610 O 27-May-14 401 645 27-Aug-14 462 650 09-Feb-16 168 R 12-Feb-16 158 650 O 02-Mar-16 185 265 17-Jun-16 281 360 12-Oct-16 355 460 * Asterisk signifies initiation or assumption of coverage. OUTPERFORM REST RICT ED

The analyst(s) responsible for preparing this research report received Compensation that is based upon various factors including Credit Suisse's total revenues, a portion of which are generated by Credit Suisse's investment banking activities As of December 10, 2012 Analysts’ stock rating are defined as follows: Outperform (O) : The stock’s total return is expected to outperform the relevant benchmark* over the next 12 months. Neutral (N) : The stock’s total return is expected to be in line with the relevant benchmark* over the next 12 months. Underperform (U) : The stock’s total return is expected to underperform the relevant benchmark* over the next 12 months. *Relevant benchmark by region: As of 10th December 2012, Japanese ratings are based on a stock’s total return relative to the analyst's coverage universe which consists of all companies covered by the analyst within the relevant sector, with Outperforms representing the most attractive, Neutrals the less attractive, and Underperforms the least attractive investment opportunities. As of 2nd October 2012, U.S. and Canadian as well as European ra tings are based on a stock’s total return relative to the analyst's coverage universe which consists of all companies covered by the analyst within the relevant sector, with Outperforms representing the most attractive, Neutrals the less attractive, and Underperforms the least attractive investment opportunities. For Latin American and non-Japan Asia stocks, ratings are based on a stock’s total return relative to the average total return of the relevant country or regional benchmark; prior to 2nd October 2012 U.S. and Canadian ratings were based on (1) a stock’s absolute total return potential to its current share price and (2) the relative attractiv eness of a stock’s total return potential within an analyst’s coverage universe. For Australian and New Zealand stocks, the expected total return (ETR) calculation includes 12-month rolling dividend yield. An Outperform rating is assigned where an ETR is greater than or equal to 7.5%; Underperform where an ETR less than or equal to 5%. A Neutral may be assigned where the ETR is between -5% and 15%. The overlapping rating range allows analysts to assign a rating that puts ETR in the context of associated risks. Prior to 18 May 2015, ETR ranges for Outperform and Underperform ratings did not overlap with Neutral thresholds between 15% and 7.5%, which was in operation from 7 July 2011. Restricted (R) : In certain circumstances, Credit Suisse policy and/or applicable law and regulations preclude certain types of communications, including an investment recommendation, during the course of Credit Suisse's engagement in an investment banking transaction and in certain other circumstances. Not Rated (NR) : Credit Suisse Equity Research does not have an investment rating or view on the stock or any other securities related to the company at this time. Not Covered (NC) : Credit Suisse Equity Research does not provide ongoing coverage of the company or offer an investment rating or investment view on the equity security of the company or related products. Volatility Indicator [V] : A stock is defined as volatile if the stock price has moved up or down by 20% or more in a month in at least 8 of the past 24 months or the analyst expects significant volatility going forward. Analysts’ sector weightings are distinct from analysts’ stock ratings and are based on the analyst’s expectations for the fundamentals and/or valuation of the sector* relative to the group’s historic fundamentals and/or valuation: Overweight : The analyst’s expectation for the sector’s fundamentals and/or valuation is favorable over the next 12 months. Market Weight : The analyst’s expectation for the sector’s fundamentals and/or valuation is neutral over the next 12 months. Underweight : The analyst’s expectation for the sector’s fundamentals and/or valuation is cautious over the next 12 months. *An analyst’s coverage sector consists of all companies covered by the analyst within the relevant sector. An analyst may cov er multiple sectors. Credit Suisse's distribution of stock ratings (and banking clients) is:

Global Ratings Distribution Rating Versus universe (%) Of which banking clients (%) Outperform/Buy* 44% (63% banking clients) Neutral/Hold* 38% (60% banking clients) Underperform/Sell* 15% (55% banking clients) Restricted 3% *For purposes of the NYSE and NASD ratings distribution disclosure requirements, our stock ratings of Outperform, Neutral, and Underperform most closely correspond to Buy, Hold, and Sell, respectively; however, the meanings are not the same, as our stock ratings are determined on a relative basis. (Please refer to definitions above.) An investor's decision to buy or sell a security should be based on investment objectives, current holdin gs, and other individual factors. Important Global Disclosures Credit Suisse’s research reports are made available to clients through our proprietary research portal on CS PLUS. Credit Suisse research products may also be made available through third-party vendors or alternate electronic means as a convenience. Certain research products are only made available through CS PLUS. The services provided by Credit Suisse’s analysts to clients may depend on a specific client’s preferences regarding the frequency and manner of receiving communications, the client’s risk profile and investment, the size and scope of the overall client relationship with

Technology sector 39 6 December 2016 the Firm, as well as legal and regulatory constraints. To access all of Credit Suisse’s research that you are entitled to receive in the most timely manner, please contact your sales representative or go to https://plus.credit-suisse.com . Credit Suisse’s policy is to update research reports as it deems appropriate, based on developments with the subject company, the sector or the market that may have a material impact on the research views or opinions stated herein. Credit Suisse's policy is only to publish investment research that is impartial, independent, clear, fair and not misleading. For more detail please refer to Credit Suisse's Policies for Managing Conflicts of Interest in connection with Investment Research: http://www.csfb.com/research-and- analytics/disclaimer/managing_conflicts_disclaimer.html . Credit Suisse's policy is only to publish investment research that is impartial, independent, clear, fair and not misleading. For more detail please refer to Credit Suisse's Policies for Managing Conflicts of Interest in connection with Investment Research:

Target Price and Rating Valuation Methodology and Risks: (12 months) for Hitachi Kokusai Electric (6756.T) Method: Our ¥1,980 target price for Hitachi Kokusai Electric is based on our end-FY3/17E EPS of ¥144.09 and a forward P/E of 13.7x, which is two standard deviations above the average since October 2013 when memory makers restarted investment. We think the company is well- placed to benefit from 3D NAND investment. Our OUTPERFORM rating is based on a comparison of expected 12-month total returns with our coverage universe. Risk: Downside risks to our ¥1,980 target price and OUTPERFORM rating for Hitachi Kokusai Electric include falling NAND prices, cutbacks to NAND investment and unprofitable projects in the Video & Communication solutions business. Target Price and Rating Valuation Methodology and Risks: (12 months) for TOWA (6315.T) Method: Our ¥1,680 target price for Towa is derived by applying our fair-value P/B of 1.51x to our FY3/18E BPS of ¥1,109. The P/B is calculated by applying a premium of one standard deviation to the TOPIX-relative P/B, based on our FY3/18E ROE estimate of 16.9%. The three main points are: (1) NAND makers accelerating their investment in 3D and (2) adoption of compression packaging in North American smartphones. We rate the company OUTPERFORM on the basis of 12-month expected total return relative to our sector coverage. Risk: Upside catalysts to our ¥1,680 target price and OUTPERFORM rating for Towa include (1) NAND makers accelerating their investment in 3D and (2) adoption of compression packaging in North American smartphones. Risks include (1) a slowdown in semiconductor investment owing to market deterioration and (2) limited uptake of compression packaging. Target Price and Rating Valuation Methodology and Risks: (12 months) for Tokyo Electron (8035.T) Method: We base our ¥10,200 target price for Tokyo Electron reflects our normalized FY3/17E EPS of ¥623.05 (which excludes from NP a quake- related ¥10bn extraordinary loss) and a target P/E of 16.3x (the average P/E since the merger talks with Applied Materials broke off, plus a one-standard-deviation premium). We think the company should benefit from new investment in 3D NAND equipment by major Korean manufacturers as well as from increased market share in etchers for the 3D NAND manufacturing process. As visibility on 2017 investment has improved, expectations for a buyback are growing. Our OUTPERFORM rating is based on a comparison of the company's 12-month potential total return versus our coverage universe. Risk: Downside risks to our ¥10,200 target price and OUTPEFORM rating for Tokyo Electron include:a curtailing of investment owing to deterioration in NAND market conditions as well as a continuation of the decline in SPE orders in 2Q and beyond owing to fewer orders from logic manufacturers and foundries. Target Price and Rating Valuation Methodology and Risks: (12 months) for Toshiba (6502.T) Method: We derive our ¥460 target price for Toshiba by applying a SOTP-derived P/E of 14.3x to our FY3/17 EPS forecast of ¥31.86.(excluding the ¥90bn gain on the sale of the home appliance business). The SOTP-based P/E references the recent valuations of the top 5–10 global competitors (Hitachi, GE etc) in each of Toshiba’s businesses. We also assign a 20% premium to the NAND business.We think the recovery of NAND margin as favorable NAND market is not priced in the current share price. Our OUTPERFORM rating is based on a comparison of the company's 12-month potential total return versus our coverage universe. Risk: Risks that could impede achievement of our ¥460 target price and OUTPERFORM rating for Toshiba include equity financing and the booking of unprofitable projects and write-downs if NAND supply-demand collapses due to smartphone production cutbacks.

Please refer to the firm's disclosure website at https://rave.credit-suisse.com/disclosures for the definitions of abbreviations typically used in the target price method and risk sections. See the Companies Mentioned section for full company names The subject company (6502.T, 6756.T) currently is, or was during the 12-month period preceding the date of distribution of this report, a client of Credit Suisse. Credit Suisse expects to receive or intends to seek investment banking related compensation from the subject company (6502.T, 6756.T) within the next 3 months.

Technology sector 40 6 December 2016

For date and time of production, dissemination and history of recommendation for the subject company(ies) featured in this report, disseminated within the past 12 months, please refer to the link: https://rave.credit-suisse.com/disclosures/view/report?i=273907&v=7bgbx1vvla5gsf4g2322p4j2g . Important Regional Disclosures Singapore recipients should contact Credit Suisse AG, Singapore Branch for any matters arising from this research report. The analyst(s) involved in the preparation of this report may participate in events hosted by the subject company, including site visits. Credit Suisse does not accept or permit analysts to accept payment or reimbursement for travel expenses associated with these events. Restrictions on certain Canadian securities are indicated by the following abbreviations: NVS--Non-Voting shares; RVS--Restricted Voting Shares; SVS--Subordinate Voting Shares. Individuals receiving this report from a Canadian investment dealer that is not affiliated with Credit Suisse should be advised that this report may not contain regulatory disclosures the non-affiliated Canadian investment dealer would be required to make if this were its own report. For Credit Suisse Securities (Canada), Inc.'s policies and procedures regarding the dissemination of equity research, please visit https://www.credit- suisse.com/sites/disclaimers-ib/en/canada-research-policy.html. Principal is not guaranteed in the case of equities because equity prices are variable. Commission is the commission rate or the amount agreed with a customer when setting up an account or at any time after that. This research report is authored by: Credit Suisse Securities (Japan) Limited ...... Hideyuki Maekawa ; Akinori Kanemoto ; Mika Nishimura To the extent this is a report authored in whole or in part by a non-U.S. analyst and is made available in the U.S., the following are important disclosures regarding any non-U.S. analyst contributors: The non-U.S. research analysts listed below (if any) are not registered/qualified as research analysts with FINRA. The non-U.S. research analysts listed below may not be associated persons of CSSU and therefore may not be subject to the NASD Rule 2711 and NYSE Rule 472 restrictions on communications with a subject company, public appearances and trading securities held by a research analyst account. Credit Suisse Securities (Japan) Limited ...... Hideyuki Maekawa ; Akinori Kanemoto ; Mika Nishimura For Credit Suisse disclosure information on other companies mentioned in this report, please visit the website at https://rave.credit- suisse.com/disclosures or call +1 (877) 291-2683.

Technology sector 41 6 December 2016

This report is produced by subsidiaries and affiliates of Credit Suisse operating under its Global Markets Division. For more information on our structure, please use the following link: https://www.credit-suisse.com/who-we-are This report may contain material that is not directed to, or intended for distribution to or use by, any person or entity who is a citizen or resident of or located in any locality, state, country or other jurisdiction where such distribution, publication, availability or use would be contrary to law or regulation or which would subject Credit Suisse or its affiliates ("CS") to any registration or licensing requirement within such jurisdiction. All material presented in this report, unless specifically indicated otherwise, is under copyright to CS. None of the material, nor its content, nor any copy of it, may be altered in any way, transmitted to, copied or distributed to any other party, without the prior express written permission of CS. All trademarks, service marks and logos used in this report are trademarks or service marks or registered trademarks or service marks of CS or its affiliates.The information, tools and material presented in this report are provided to you for information purposes only and are not to be used or considered as an offer or the solicitation of an offer to sell or to buy or subscribe for securities or other financial instruments. CS may not have taken any steps to ensure that the securities referred to in this report are suitable for any particular investor. CS will not treat recipients of this report as its customers by virtue of their receiving this report. The investments and services contained or referred to in this report may not be suitable for you and it is recommended that you consult an independent investment advisor if you are in doubt about such investments or investment services. Nothing in this report constitutes investment, legal, accounting or tax advice, or a representation that any investment or strategy is suitable or appropriate to your individual circumstances, or otherwise constitutes a personal recommendation to you. CS does not advise on the tax consequences of investments and you are advised to contact an independent tax adviser. Please note in particular that the bases and levels of taxation may change. Information and opinions presented in this report have been obtained or derived from sources believed by CS to be reliable, but CS makes no representation as to their accuracy or completeness. CS accepts no liability for loss arising from the use of the material presented in this report, except that this exclusion of liability does not apply to the extent that such liability arises under specific statutes or regulations applicable to CS. This report is not to be relied upon in substitution for the exercise of independent judgment. CS may have issued, and may in the future issue, other communications that are inconsistent with, and reach different conclusions from, the information presented in this report. Those communications reflect the different assumptions, views and analytical methods of the analysts who prepared them and CS is under no obligation to ensure that such other communications are brought to the attention of any recipient of this report. Some investments referred to in this report will be offered solely by a single entity and in the case of some investments solely by CS, or an associate of CS or CS may be the only market maker in such investments. Past performance should not be taken as an indication or guarantee of future performance, and no representation or warranty, express or implied, is made regarding future performance. Information, opinions and estimates contained in this report reflect a judgment at its original date of publication by CS and are subject to change without notice. The price, value of and income from any of the securities or financial instruments mentioned in this report can fall as well as rise. The value of securities and financial instruments is subject to exchange rate fluctuation that may have a positive or adverse effect on the price or income of such securities or financial instruments. Investors in securities such as ADR's, the values of which are influenced by currency volatility, effectively assume this risk. Structured securities are complex instruments, typically involve a high degree of risk and are intended for sale only to sophisticated investors who are capable of understanding and assuming the risks involved. The market value of any structured security may be affected by changes in economic, financial and political factors (including, but not limited to, spot and forward interest and exchange rates), time to maturity, market conditions and volatility, and the credit quality of any issuer or reference issuer. Any investor interested in purchasing a structured product should conduct their own investigation and analysis of the product and consult with their own professional advisers as to the risks involved in making such a purchase. Some investments discussed in this report may have a high level of volatility. High volatility investments may experience sudden and large falls in their value causing losses when that investment is realised. Those losses may equal your original investment. Indeed, in the case of some investments the potential losses may exceed the amount of initial investment and, in such circumstances, you may be required to pay more money to support those losses. Income yields from investments may fluctuate and, in consequence, initial capital paid to make the investment may be used as part of that income yield. Some investments may not be readily realisable and it may be difficult to sell or realise those investments, similarly it may prove difficult for you to obtain reliable information about the value, or risks, to which such an investment is exposed. This report may provide the addresses of, or contain hyperlinks to, websites. Except to the extent to which the report refers to website material of CS, CS has not reviewed any such site and takes no responsibility for the content contained therein. Such address or hyperlink (including addresses or hyperlinks to CS's own website material) is provided solely for your convenience and information and the content of any such website does not in any way form part of this document. Accessing such website or following such link through this report or CS's website shall be at your own risk. This report is issued and distributed in European Union (except Switzerland): by Credit Suisse Securities (Europe) Limited, One Cabot Square, London E14 4QJ, England, which is authorised by the Prudential Regulation Authority and regulated by the Financial Conduct Authority and the Prudential Regulation Authority. Germany: Credit Suisse Securities (Europe) Limited Niederlassung Frankfurt am Main regulated by the Bundesanstalt fuer Finanzdienstleistungsaufsicht ("BaFin"). United States and Canada: Credit Suisse Securities (USA) LLC; Switzerland: Credit Suisse AG; Brazil: Banco de Investimentos Credit Suisse (Brasil) S.A or its affiliates; Mexico: Banco Credit Suisse (México), S.A. (transactions related to the securities mentioned in this report will only be effected in compliance with applicable regulation); Japan: by Credit Suisse Securities (Japan) Limited, Financial Instruments Firm, Director-General of Kanto Local Finance Bureau ( Kinsho) No. 66, a member of Japan Securities Dealers Association, The Financial Futures Association of Japan, Japan Investment Advisers Association, Type II Financial Instruments Firms Association; Hong Kong: Credit Suisse (Hong Kong) Limited; Australia: Credit Suisse Equities (Australia) Limited; Thailand: Credit Suisse Securities (Thailand) Limited, regulated by the Office of the Securities and Exchange Commission, Thailand, having registered address at 990 Abdulrahim Place, 27th Floor, Unit 2701, Rama IV Road, Silom, Bangrak, Bangkok10500, Thailand, Tel. +66 2614 6000; Malaysia: Credit Suisse Securities (Malaysia) Sdn Bhd; Singapore: Credit Suisse AG, Singapore Branch; India: Credit Suisse Securities (India) Private Limited (CIN no.U67120MH1996PTC104392) regulated by the Securities and Exchange Board of India as Research Analyst (registration no. INH 000001030) and as Stock Broker (registration no. INB230970637; INF230970637; INB010970631; INF010970631), having registered address at 9th Floor, Ceejay House, Dr.A.B. Road, Worli, Mumbai - 18, India, T- +91-22 6777 3777; South Korea: Credit Suisse Securities (Europe) Limited, Seoul Branch; Taiwan: Credit Suisse AG Taipei Securities Branch; Indonesia: PT Credit Suisse Securities Indonesia; Philippines:Credit Suisse Securities (Philippines ) Inc., and elsewhere in the world by the relevant authorised affiliate of the above. Additional Regional Disclaimers Hong Kong: Credit Suisse (Hong Kong) Limited ("CSHK") is licensed and regulated by the Securities and Futures Commission of Hong Kong under the laws of Hong Kong, which differ from Australian laws. CSHKL does not hold an Australian financial services licence (AFSL) and is exempt from the requirement to hold an AFSL under the Corporations Act 2001 (the Act) under Class Order 03/1103 published by the ASIC in respect of financial services provided to Australian wholesale clients (within the meaning of section 761G of the Act). Research on Taiwanese securities produced by Credit Suisse AG, Taipei Securities Branch has been prepared by a registered Senior Business Person. Australia (to the extent services are offered in Australia): Credit Suisse Securities (Europe) Limited (“CSSEL”) and Credit Suisse International (“CSI”) are authorised by the Prudential Regulation Authority and regulated by the Financial Conduct Authority (“FCA”) and the Prudential Regulation Authority under UK laws, which differ from Australian Laws. CSSEL and CSI do not hold an Australian Financial Services Licence (“AFSL”) and are exempt from the requirement to hold an AFSL under the Corporations Act (Cth) 2001 (“Corporations Act”) under Class Order 03/1099 published by the Australian Securities and Investments Commission (“ASIC”), in respect of the financial services provided to Australian wholesale clients (within the meaning of section 761G of the Corporations Act). This material is not for distribution to retail clients and is directed exclusively at Credit Suisse's professional clients and eligible counterparties as defined by the FCA, and wholesale clients as defined under section 761G of the Corporations Act. Credit Suisse (Hong Kong) Limited (“CSHK”) is licensed and regulated by the Securities and Futures Commission of Hong Kong under the laws of Hong Kong, which differ from Australian laws. CSHKL does not hold an AFSL and is exempt from the requirement to hold an AFSL under the Corporations Act under Class Order 03/1103 published by the ASIC in respect of financial services provided to Australian wholesale clients (within the meaning of section 761G of the Corporations Act). Credit Suisse Securities (USA) LLC (CSSU) and Credit Suisse Asset Management LLC (CSAM LLC) are licensed and regulated by the Securities Exchange Commission of the United States under the laws of the United States, which differ from Australian laws. CSSU and CSAM LLC do not hold an AFSL and is exempt from the requirement to hold an AFSL under the Corporations Act under Class Order 03/1100 published by the ASIC in respect of financial services provided to Australian wholesale clients (within the meaning of section 761G of the Corporations Act). Malaysia: Research provided to residents of Malaysia is authorised by the Head of Research for Credit Suisse Securities (Malaysia) Sdn Bhd, to whom they should direct any queries on +603 2723 2020. Singapore: This report has been prepared and issued for distribution in Singapore to institutional investors, accredited investors and expert investors (each as defined under the Financial Advisers Regulations) only, and is also distributed by Credit Suisse AG, Singapore Branch to overseas investors (as defined under the Financial Advisers Regulations). Credit Suisse AG, Singapore Branch may distribute reports produced by its foreign entities or affiliates pursuant to an arrangement under Regulation 32C of the Financial Advisers Regulations. Singapore recipients should contact Credit Suisse AG, Singapore Branch at +65-6212-2000 for matters arising from, or in connection with, this report. By virtue of your status as an institutional investor, accredited investor, expert investor or overseas investor, Credit Suisse AG, Singapore Branch is exempted from complying with certain compliance requirements under the Financial Advisers Act, Chapter 110 of Singapore (the “FAA”), the Financial Advisers Regulations and the relevant Notices and Guidelines issued thereunder, in respect of any financial advisory service which Credit Suisse AG, Singapore Branch may provide to you. UAE: This information is being distributed by Credit Suisse AG (DIFC Branch), duly licensed and regulated by the Dubai Financial Services Authority (“DFSA”). Related financial services or products are only made available to Professional Clients or Market Counterparties, as defined by the DFSA, and are not intended for any other persons. Credit Suisse AG (DIFC Branch) is located on Level 9 East, The Gate Building, DIFC, Dubai, United Arab Emirates. EU: This report has been produced by subsidiaries and affiliates of Credit Suisse operating under its Global Markets Division This research may not conform to Canadian disclosure requirements. In jurisdictions where CS is not already registered or licensed to trade in securities, transactions will only be effected in accordance with applicable securities legislation, which will vary from jurisdiction to jurisdiction and may require that the trade be made in accordance with applicable exemptions from registration or licensing requirements. Non-US customers wishing to effect a transaction should contact a CS entity in their local jurisdiction unless governing law permits otherwise. US customers wishing to effect a transaction should do so only by contacting a representative at Credit Suisse Securities (USA) LLC in the US. Please note that this research was originally prepared and issued by CS for distribution to their market professional and institutional investor customers. Recipients who are not market professional or institutional investor customers of CS should seek the advice of their independent financial advisor prior to taking any investment decision based on this report or for any necessary explanation of its contents. This research may relate to investments or services of a person outside of the UK or to other matters which are not authorised by the Prudential Regulation Authority and regulated by the Financial Conduct Authority and the Prudential Regulation Authority or in respect of which the protections of the Prudential Regulation Authority and Financial Conduct Authority for private customers and/or the UK compensation scheme may not be available, and further details as to where this may be the case are available upon request in respect of this report. CS may provide various services to US municipal entities or obligated persons ("municipalities"), including suggesting individual transactions or trades and entering into such transactions. Any services CS provides to municipalities are not viewed as "advice" within the meaning of Section 975 of the Dodd-Frank Wall Street Reform and Consumer Protection Act. CS is providing any such services and related information solely on an arm's length basis and not as an advisor or fiduciary to the municipality. In connection with the provision of the any such services, there is no agreement, direct or indirect, between any municipality (including the officials,management, employees or agents thereof) and CS for CS to provide advice to the municipality. Municipalities should consult with their financial, accounting and legal advisors regarding any such services provided by CS. In addition, CS is not acting for direct or indirect compensation to solicit the municipality on behalf of an unaffiliated broker, dealer, municipal securities dealer, municipal advisor, or investment adviser for the purpose of obtaining or retaining an engagement by the municipality for or in connection with Municipal Financial Products, the issuance of municipal securities, or of an investment adviser to provide investment advisory services to or on behalf of the municipality. If this report is being distributed by a financial institution other than Credit Suisse AG, or its affiliates, that financial institution is solely responsible for distribution. Clients of that institution should contact that institution to effect a transaction in the securities mentioned in this report or require further information. This report does not constitute investment advice by Credit Suisse to the clients of the distributing financial institution, and neither Credit Suisse AG, its affiliates, and their respective officers, directors and employees accept any liability whatsoever for any direct or consequential loss arising from their use of this report or its content. Principal is not guaranteed. Commission is the commission rate or the amount agreed with a customer when setting up an account or at any time after that. Copyright © 2016 CREDIT SUISSE AG and/or its affiliates. All rights reserved. Investment principal on bonds can be eroded depending on sale price or market price. In addition, there are bonds on which investment principal can be eroded due to changes in redemption amounts. Care is required when investing in such instruments. When you purchase non-listed Japanese fixed income securities (Japanese government bonds, Japanese municipal bonds, Japanese government guaranteed bonds, Japanese corporate bonds) from CS as a seller, you will be requested to pay the purchase price only.

Technology sector 42