Cloud Hosting

2020 hyperscale server demand to grow 11% YoY; DRAM contract prices to pick up from 2Q20

We visited major IT H/W companies in Taiwan and Japan from Nov 11 to 15 to examine the key issues of the IT H/W sector. Through the visits we found that: 1) 2020 demand for finished products is expected to stagnate but hyperscale demand should grow 11% YoY on expanding 5G services and data center demand, helping to drive DRAM prices higher; 2) Samsung Electronics (OTC:SSNLF) will gain more prowess in the smartphone space with its foldable display and high-resolution cameras (64MP or higher); Huawei's marginalized presence in the overseas markets also benefits Samsung; and 3) MLCC inventory has entered the normal range in 4Q19 and price declines have slowed considerably; price hikes of ultra-small, high capacitance MLCCs are under way.

Hyperscale's share of servers and server DRAM to grow up to 60% and 78%, respectively

Hyperscale servers used by data centers and telcos currently account for 38% of total server demand but we believe the number will increase up to 60% by 2025. Hyperscale server DRAM's percentage of total server DRAM is also expected to climb to 78% by 4Q20. As data center investments pick up pace led by North America's cloud providers, OVHcloud, and China's BAT, server DRAM is likely to represent 35% of total DRAM demand in 2020.

Korean IT H/W sector to grow more attractive next year, top picks Samsung and SEMCO

We believe Korea's IT H/W sector will continue to be an attractive investment next year given the expansion of 5G services, memory chip price hikes, and the launch of the foldable smartphones, the embodiment of the "super-gap" strategy. We present Samsung and SEMCO (OTC:SMSGF) as our top picks. We favor Samsung the most, followed by SEMCO, SK Hynix (OTC:HXSCL), LG Innotek, and LG Electronics (OTC:LGEIY).

5G and hyperscale servers to drive 2020 memory semicon prices higher

We visited major IT H/W manufacturers and research institutions such as DRAMeXchange, Stone Partners, and InnoResearch in Japan and Taiwan from Nov 11 to 15 to examine current industry issues such as: 1) the 2020 memory semiconductor market outlook; 2) the competitive focus of the smartphone market in 2020 in terms of specifications; and 3) the supply/demand dynamics of MLCCs.

DRAM contract prices are likely to start to rebound in 2Q20 after sliding until 1Q20. We project DRAM contract prices will rise more than 20% YoY in 4Q20. After a rebound in 4Q19, NAND prices should correct in 1H20 before regaining momentum in 3Q20. In 4Q20, we estimate TLC NAND wafer prices will have risen 17% YoY. In 2020, helped by efforts to contain supply, the DRAM and NAND markets are expected to grow 3.8% and 18% YoY to USD69.7bn and 53.3bn, respectively. Meanwhile, the 2020 hyperscale server shipment volume will likely increase 11% YoY to 5.2mn units, and hyperscale servers should account for 78% of server DRAM demand. 2020 will also see the debut of Intel's (INTC) 14nm-based Cooper Lake processors, and 1H21 will witness the launch of the 10nm-based Ice Lake CPU.

Samsung Electronics' Galaxy S11 is expected to come with LPDDR5 DRAM and rear cameras with no less than 64MP. The S11+ will offer an even higher resolution of up to 108MP. The one with 108MP cameras will have the folded camera module technology, and its mobile DRAM capacity will probably be over 12GB. In 2020, Samsung is expected to launch two foldable smartphone models and plans to widen the gap with competitors by selling at least 5mn foldable smartphones. Meanwhile, at Apple (AAPL), the strong sales of the iPhone 11, which offers an LCD display, appear to be hampering Apple's original plan of making all the 2020 models in OLED. There is a growing possibility that the new products to come out in 2H20 will include a 6.06-inch LCD model. As for Huawei, the inability to use the Google (NASDAQ:GOOG) (GOOGL) OS will erode its 2020 export volume by 50mn units, which bodes well for Samsung and Apple.

After our visit to a Taiwan-based MLCC maker, we now believe MLCC inventory that had increased to 150 days at one point will contract to a normal level of 70 days in 4Q19, which is expected to limit MLCC prices' QoQ decline in 4Q19 to 4%. In particular, the prices of ultra-small, high-capacitance MLCCs for smartphones are likely to increase from 4Q19 amid the rise of demand for 5G smartphones. An overall increase in MLCC prices will likely happen from 2Q20.

Korean tech shares have been outperforming the broader market since the beginning of the year. In 2020, we believe Korean IT H/W names will come into an even bigger spotlight compared with their global peers, considering: 1) favorable business conditions characterized by the expansion of 5G services and the rebound of memory semiconductor prices; 2) Huawei's dwindling presence in the smartphone space; and 3) BOE's weakening strength in OLED. Most importantly, the Samsung Galaxy Fold is the manifestation of the tech giant's "super-gap" strategy (a strategy of creating an insurmountable technological gap with competitors) and it is slated to enjoy bigger and bigger sales going forward, which will make Samsung an even more attractive investment. We present Samsung Electronics and SEMCO as our top picks for Korea's IT H/W sector. We like Samsung the most, followed by SEMCO, SK Hynix, LG Innotek, and LG Electronics.

We had a meeting with DRAMeXchange, which has been a good prognosticator of the DRAM market direction for the past two years. In terms of demand, it now expects no finished products to see meaningful growth in 2020 except for servers. However, it believes neither PCs nor smartphones will suffer sales declines. Despite the protracted US-China trade dispute, demand for finished products has kept up well, which in itself helps diminish uncertainties. In 2020, with the expansion of 5G services, it expects major public cloud service providers to scale up investments in data centers and telcos' mobile edge computing (MEC), which is in line with our view. That said, its forecast of supply bit growth seems too conservative. It estimates DRAM wafer capacity will be 1,288K wpm at end-2019 and 1,340K wpm (+4% YoY) at end-2020, but it expects YoY supply bit growth to be a meager 12.4% because it believes technology migrations will provide severe headwinds to all DRAM makers. However, in this era of 1X and 1Y, DRAMeXchange's supply bit growth forecast seems overly conservative. Aside from this, DRAMeXchange projects DRAM's demand bit growth to be 18%, backed by memory content growth in servers and smartphones. While it estimates overall ASP will slip 9.7% YoY in 2020, it believes the DRAM prices in 4Q20 will be more than 20% higher than DRAM prices in 4Q19. In short, it expects DRAM prices to soar as the DRAM sector faces supply shortages from 2Q20.

Meanwhile, Micron (MU), which currently operates two fabs in Taiwan, appears to be considering building a new fab in Taiwan after 2020. As seen by Samsung's Pyeongtaek #2 fab and SK Hynix's M16, Micron also expects DRAM demand to grow in the long term. We have a more positive view toward DRAM supply bit growth vs. DRAMeXchange. DRAMeXchange assumes Micron's bit growth in 2020 to be 10.3% while Micron itself projects 15%. DRAMeXchange expects the DRAM market to expand 3.8% YoY to USD69.7bn in 2020 with supply bit growth of 15%. Although DRAMeXchange has been correct many times recently in predicting the direction of DRAM market, its predictions were far from precise, because of the high volatility of the DRAM industry. Therefore, it would be prudent for investors to focus on the basic direction that the DRAM market will take.

Server demand was the key driver of 2018's historic-high DRAM demand. Amid an increase in public cloud companies' long-term agreements (LTAs) to build new data centers, the price of server DRAM (32GB) rose 139% from its trough. However, as the first round of investments by public cloud operators came to a close, server DRAM prices dropped 66% from the peak, breaking below the previous low. Server companies' DRAM inventories are likely to reach normal levels by end-2019, and server DRAM prices should start rebounding after bottoming in 1Q20. During the first server DRAM upcycle, DRAM prices rose for eight consecutive quarters; so it should be interesting to watch how long DRAM prices continue to rise. The investments by public cloud companies should not be as intense as the last time. However, considering the demand from private cloud companies and telcos' investments in edge servers, DRAM price increases may last longer than the last time, although the scale of increase should be weaker. We must not overlook the fact that, as witnessed during the first upcycle, once cloud companies decide to make an investment, they will invest even in the typical off-season for tech products. Taken together, we believe server DRAM will account for up to 34.9% of total DRAM shipments in 2020.

For mobile DRAM, LPDDR5 demand is expected to grow in earnest from 2020 along with 5G services, and a growing number of rear cameras on smartphones will further increase mobile DRAM content. In particular, Samsung's Galaxy S11 is expected to come with LPDDR5 DRAM and raise the rear camera pixels to over 64MP. The most expensive model is likely to incorporate the folded camera technology into its 108MP cameras. There are expectations that the said model will offer 12GB or higher mobile DRAM. If the Galaxy S11 manages to display a distinguished 5G speed by adopting LPDDR5 earlier than competitors, Chinese smartphone makers are also likely to rush to adopt LPDDR5.

Meanwhile, DRAMeXchange expects DRAM contract prices (in the case of PC DRAM) to climb 6% QoQ in 2Q20, 12% QoQ in 3Q20, and 8.8% QoQ in 4Q20. It believes server DRAM demand will display a similar level of increase. As for mobile DRAM, the price premium is more than 35% vs. PC DRAM, so the scale of increase will likely be lower than that of PC or server DRAM.

At Intel, 3Q19 results remained flat YoY as PC CPU shipments slid about 5% YoY, and operating profit fell 12% YoY amid the deterioration of NAND earnings. In 4Q19, Intel expects sales to grow 3% YoY to USD19.2bn and an operating margin of 2%p YoY. What merits our attention in Intel's earnings results are the Data Center Group indicators which are highly correlated with server CPU and the Client Computing Group indicators which are related with PC CPU. As for the Data Center Group, 3Q19 shipments declined 6% YoY but ASP rose 9%, which worked to boost sales by 4% YoY. The client breakdown of Data Center Group sales reveals that telco-bound sales rose for a second consecutive quarter with 3Q19 seeing an 11% YoY increase. The investments in 5G infrastructure by telcos in the US, Korea, and China are spurring demand for edge servers and hyperscale servers used in data centers. In particular, sales bound for cloud computing companies, which slipped 1% YoY in 2Q19, rose 3% YoY in 3Q19. Furthermore, the sales of server CPUs for enterprises and government offices increased 1%, a dramatic rise from the 31% decline seen in 2Q19. Meanwhile, Intel guides for YoY sales growth of 7% in 4Q19 for the Data Centric segment (server CPU + NAND + IoT + Mobileye + Security + others). Since server CPUs account for more than 65% of its Data Centric sales, server CPU sales are expected to grow at least 5% YoY.

We also visited Taiwan-based major server suppliers Aspeed and Inventec. Aspeed has an 80% share in the market for hyperscale server baseboard management controllers (BMC), so its sales lead the demand for memory semiconductors and CPUs for servers by more than two months. With that said, Aspeed's sales are at an all-time high from July to October. Although sales are likely to decrease MoM from November due to inventory adjustments by server companies, its annual sales are expected to increase 13% YoY in 2019. Contrary to some beliefs (orders are placed two or three times a year), ODM server companies place orders for BMCs on a monthly basis to meet client demand. In conclusion, Aspeed's robust earnings results in 3Q19 suggest that server companies' shipments will increase in 4Q19 on a meaningful scale. Meanwhile, server demand is expected to grow QoQ from 2Q20 after inventory controls in 1Q20. Meanwhile, this year's BMC shipments are expected to climb 6.3% YoY, and ODM companies' server shipments are likely to increase 5.1% YoY. Of note, Aspeed's sales growth is more than double the growth of BMC shipments because of growing demand for 28nm products.

Currently, hyperscale servers account for around 40% of total server sales. However, by 2025 when demand from cloud companies and telcos increase, hyperscale servers' portion of sales is expected to rise up to 60%, which is positive for memory semiconductor demand. Meanwhile, Inventec, which has the largest sales exposure to servers among Taiwanese ODMs, expects server demand to grow in the high-single digits from 2020, cloud server demand 10% and enterprise server demand around 5%. At present, among Taiwanese ODMs, Quanta Computer (OTC:QUCCF) appears to have 25% of sales generated from servers while for Inventec the number is estimated to be 34%. As for public cloud computing companies, dependence on such companies as HP, Dell, Lenovo, and Huawei has plummeted, and instead they get most of their servers from Taiwanese ODM/OEM companies (Inventec, Quanta, QCT, and Wiwynn). It is believed that China's Inspur and Lenovo supply most of hyperscale servers to the BAT (Baidu (BIDU), Alibaba (BABA), and Tencent (OTCPK:TCEHY)). On the other hand, Chinese telcos are believed to buy their servers mostly from Huawei while US telcos obtain their servers from Ericsson (ERIC) and Nokia (NOK). Meanwhile, Taiwanese ODMs' servers for telcos appear to be supplied to Cisco (CSCO) Systems and Juniper Networks (JNPR). Ericsson and Nokia are assumed to provide Cisco's optimized products to telcos' edge infra network. Edge computing servers have different specifications by base station and switching center. High-capacity servers are mostly installed in telcos' data centers, or switching centers that are adjacent to data centers. In contrast, base stations are believed to be equipped with low-capacity servers. In all, if data traffic increases exponentially through edge computing, the capacity of public cloud servers will naturally increase, which means additional investments in servers. 5G services are expected to drive server demand by increasing data consumption and creating fresh demand (e.g., autonomous vehicles and robots). Against this backdrop, we believe server DRAM will represent 48% of total DRAM demand by 2025.

Globally, the number of data centers is increasing rapidly amid explosive growth in demand for public and private clouds. By the end of 2019, the number of cities with data centers (based on public clouds) is expected to rise to 144, and the number should continue to grow, led by Google which is strengthening its gaming cloud business and its rivals such as Amazon and Microsoft. Currently, the company that has established the largest number of data centers worldwide is Europe's OVHcloud. It has data centers in 30 different cities worldwide and built 380,000 physical servers. Google, Microsoft (MSFT), and Amazon (AMZN) are also building more than eight data centers this year. Google, which is striving to strengthen its gaming cloud business, plans to establish 15 data centers in various cities including Seoul, Jakarta, Warsaw, LA, and Salt Lake City. Since speed is the key to gaming clouds, data centers that are nearby are always preferred over data centers that are far away. For Amazon and Alibaba, additional investments in data centers are a must as they lead the innovations in delivery and logistics.

The long-term direction of data center demand is positive but in the short term, what is important is that Intel delivers its server CPUs in a timely manner. We note that Intel has more than 95% market share in hyperscale servers. The delayed launch of Cascade CPUs by more than 10 months has led to a delay of cloud computing companies' server investments. While Intel is scheduled to release 10nm-based Ice Lake CPUs in 2H20, the actual release will probably be 1H21 considering circumstances. Having said that, the 2020 launch of the 14nm-based Cooper Lake with eight channels is a certainty. Accordingly, demand should improve compared to this year.

With the launch of Cooper Lake CPUs, hyperscale servers' per device DRAM memory content is expected to continue rising. Accordingly, hyperscale servers' share of server DRAM demand will jump from 56% in 4Q18 to 78% in 4Q20. With the expansion of 5G and various cloud computing services (public, private, hybrid), hyperscale servers' share of the server market is continuing to grow and hyperscale servers' memory content per device is increasing. As such, the direction of server DRAM demand looks positive in the long term.

In 4Q19, NAND contract prices are expected to rebound around 5% QoQ due to reduced supply following Kioxia's (formerly Toshiba Memory) Yokkaichi fab outage and production cuts by major NAND makers. As seen by Intel and Aspeed's guidance, the server demand in 4Q19 should work to boost enterprise SSD prices most sharply, and client SSD prices are also likely to rise more than 10% QoQ thanks to burgeoning notebook demand. However, a seasonal demand decline expected in 1H20 will probably send down NAND contract prices again in 1H20. That said, the pace of decline in 1Q20 and 2Q20 will be in the low single digits thanks to production cuts, and the prices are expected to rise again from 3Q20 on strong seasonality. In the case of smartphone storage, UFS 2.1 has grown to represent 20% of total demand while eMMC demand falters. UFS 3.0 is expected to account for 10% of total demand in 2020 on the increased use in premium smartphones.

Meanwhile, SSD's penetration is rapidly increasing, especially in client PCs amid the plunge in NAND prices. The market for enterprise SSD is also growing on the back of increasing demand from newly built data centers and telcos. We expect SSD to account for 44.3% of total NAND demand in 2019, and surpass smartphones to become the top application (accounting for 44.3% of demand) for NAND by 2023. In 2019, we believe the NAND market will contract 28% YoY to USD53.3bn due to steep price corrections. However, in 2020, NAND price declines are expected to decelerate, and the market should expand 18% YoY to USD53.3bn. The size is still low compared with 2018 but we still find it positive that growth will turn around.

The visit to Taiwanese companies gave us a chance to examine the current conditions at Chinese memory semiconductor manufacturers. It appears that JHICC has almost given up its memory semiconductor business due to US sanctions. ASML stopped supplying photolithography solutions to JHICC. On the other hand, it was reported that CXMT (formerly Innotron) will start the mass-production of 19nm DRAM later this year. However, the visit to the Taiwanese firms confirmed that CXMT's actual capacity is 20K wpm and that the company has not qualified any of the standards required by its clients. At the very least, as of end-October, it was still unable to deliver even PC DRAM samples to Huawei, Lenovo, and Asus. Its official plan is to expand the capacity to 40K wpm by mid-2020, but we believe the primary purpose of the plan is to receive funding from the Chinese local government.

YMTC, a subsidiary of China's Tsinghua Unigroup, has completed the development of 64-layer TLC and is now supposedly developing 128-layer TLC, which is expected to be used to make SSDs for client PC. Currently, controller IC is known to have been developed by Taiwan-based Silicon Motion and Phison; other Chinese local controller IC companies like Koke also appear to be working on the development of related applications. Meanwhile, YMTC currently has a plant in Wuhan and is preparing to build a new plant in Chengdu. The Chinese local government is trying hard to allure the company to come to the city by offering various incentives. Having said that, the reliability of YMTC's products needs to be verified in the future. Additionally, Tsinghua Unigroup announced a plan to build a DRAM plant in Chongqing, China, with a target to complete construction by 2021. The local government of Chongqing, like the Chengdu government, is also striving to have a memory semiconductor factory in the city but again it remains unclear as to whether technological hurdles are all overcome. To sum up, concerns about the growth of China's DRAM market seem to be exaggerated in many ways.

Apple's iPhone shipments in 2H19 are better than expected. Especially, the response to the iPhone 11, which is attractively priced and upgraded with a dual camera, has been more favorable than expected. Indeed, after visiting Sharp which supplies most of the dual cameras for the iPhone 11, we found that its camera module sales in 4Q19 would grow more than twofold QoQ, and Apple's orders have been more than 10% higher than expected. It is estimated that currently 90% of the dual cameras for the iPhone 11 are supplied by Sharp and 10% by OFilm. OFilm suffered some financial difficulties, but the worst seems to be over after the local government of Fujian took over its debt. In the meantime, the iPhone 11 series' production volume in 2H19 confirmed by Taiwanese EMS companies is estimated to be 77.6mn units, with LCD models making up close to 50%. The reason why the sales of LCD models are better than their predecessors is attractive price first and foremost. The naming (from iPhone XR to iPhone 11) also seems to have helped boost consumer sentiment.

Apple is expected to release the iPhone SE2, a 4.7-inch LCD model in 1H20. The size is easy to handle with one hand and the price should also be attractive. Since Huawei is bound to take a hit in the overseas market due to the lack of Google's Android OS, Apple will likely accelerate efforts to expand its market share. Meanwhile, Apple's original plan for 2H20 was to discontinue LCD models and have all three differently sized models (5.42-, 6.06- and 6.68-inch) in OLED. However, as LCD models are selling better than expected, it is considering an early release of the 6.06-inch SE3 in 2H20, which was initially planned for a 1H21 release, and delaying the launch of its 6.06-inch OLED model. If Apple makes such a decision, Samsung Display's flexible OLED shipments to Apple may stay flat YoY in 2020. What is clear now is that BOE, which is struggling with the Y-Octa technology, will be unable to supply Apple even in 2H20.

Huawei smartphones did quite well even after it was blacklisted by the US government. However, losing access to Google's OS in new smartphones dealt a devastating blow to its shipments. That YouTube and Google Maps are not available in overseas markets is especially critical. This has already forced Huawei to delay the overseas launch of the Mate 30. However, its market share in China is expected to rise to 50% in 2019 and 55% in 2020 on the back of patriotism marketing. Assuming 2020 smartphone sales in China at 380mn units and Huawei holds a 55% market share, Huawei's domestic shipments will be 210mn units and its export volume will be 10mn units. Because of the absence of Google OS, its export volume should plunge to 10mn from 60mn units, creating a void that Samsung and Apple will gladly fill. The problem is that if Huawei represents 210mn smartphone sales in China, other players such as Apple, Oppo, Vivo and Xiaomi will have to compete over the remaining 170mn market. There is a high possibility that the shipments of Oppo, Vivo, and Xiaomi will fall in 2020. Overseas, Oppo and Vivo have a negligible market share except for Southeast Asia, and Xiaomi's (OTCPK:XIACY) market presence is timid except for India. As such, there is a high probability that Samsung will easily exploit the European and Latin American markets without spilling any blood. Meanwhile, Samsung plans to entrust joint development manufacturing (JDM) to Chinese companies from 2020. The 2020 volume will likely be about 50mn units, all in LCD. Its JDM partner will probably be Wingtech, a current ODM for Huawei. Furthermore, Samsung plans to further strengthen its partnership with HQ. Apple's current EMS providers are three Taiwanese companies (Foxconn (OTCPK:FXCNY), Pegatron, and Wistron); of the three, Foxconn and Wistron have factories in India.

Huawei's smartphone production is 60% in-house and the rest is done by ODM and EMS providers such as Wingtech, Flextronics. Longcheer is believed to be an ODM provider for Xiaomi. Samsung's JDM models are expected to target the mid and low-end of the market, and the Vietnamese plant will mostly be dedicated to the production of premium smartphones. If Huawei's smartphone exports fall by 50mn units and if Samsung takes more than 70% of the market, the impact of Samsung's JDM on Korean component makers could be smaller than expected. Regardless, it is more important than ever for Korean parts suppliers to upgrade their product portfolios with a focus on premium components.

Meanwhile, the Galaxy Fold has been met with enthusiastic response since its debut. In addition to the fact that it folds magically, the Fold has been carefully made to perfection with three or three divided screens for multitasking, cutting edge camera performance and two batteries. Of course, the 4.6-inch external display is too small and thickness and weight need further improvements. That said, there is no doubt about the evolutionary direction of the Galaxy Fold, as Samsung has greatly improved its smartphone quality since the Galaxy S3. Meanwhile, Samsung unveiled the image of a clamshell style foldable smartphone at SDC 2019. It should be easier to mass produce vs. horizontally folding devices, and will be offered at more attractive prices compared with the Galaxy Fold 1. However, in terms of multitasking, horizontally folding smartphones are expected to be superior, and Samsung will likely unveil a new horizontally folding model in 2H20 with an enlarged external display. The Galaxy Fold provides the UX of both smartphones and tablets, and its display is expected to expand to more than 10 inches when unfolded, instead of 7.3 inches. Samsung will eventually fold the 10-inch model not just once but twice, through the development of the folding technology. Meanwhile, out-folding models and twice-folding "Z-Fold" smartphones introduced by Chinese names will find it hard to get past the prototype stage to reach the mass production stage. Huawei's Mate X is believed to have been sold out in China but the quantity is still unknown. As for Xiaomi, the release date has not even been set. Considering all circumstances, it is likely that Samsung will lead the foldable smartphone market by a wide margin for some time to come.

Meanwhile, the Galaxy S11 rear cameras, to be released in 1Q20, are expected to be available in 64MP to 108MP, and the 108MP model will be equipped with SEMCO's folded camera technology. On the other hand, the under display selfie cameras that Chinese players introduced are unlikely to be available in 2020 models due to technological difficulties.

Samsung's smartphone business is expected to enjoy both top- and bottom-line growth in 2020 on the back of: 1) the expansion of 5G services; 2) Huawei's faltering exports; 3) the launch of foldable smartphones; and 4) JDM.

BOE was a primary supplier for Huawei P30 Pro, but it is unclear as to whether it will do the same with Apple since Apple's new OLED models to be released in 2H20 will require the Y-Octa technology. Furthermore, Huawei's Mate 30 Pro, which curves around the edges at almost 90 degrees, are not available due to yield problems. As BOE struggles with the mass production of products that require a high level of technology, its technological gap with Samsung Display and LG Display is continuing to widen. While the P30 Pro requires eight masks, the Mate 30 Pro requires more than 10 masks, and the phone looks almost bezel-less when viewed from the front. Furthermore, the Mate 40 Pro to be launched in 2H20 folds at 120 degrees which is more difficult to make than the 90-degree model. In all, BOEs will be faced with more technological difficulties, not less. Meanwhile, Chinese smartphone names are recently considering using Samsung Display's hole punch display solution after grappling with the camera pop-up module's heavy cost burden and battery problems. As such, most mass-produced Chinese smartphone models in 2020 are expected to have hole punch displays, and BOE will again suffer technologically. In the meantime, Samsung Display will be able to distance competitors further in hole punch displays in addition to foldable displays.

Our visits to Taiwanese MLCC companies confirmed that the MLCC inventory would normalize by the year's end. Accordingly, we believe MLCC prices will stabilize in 2020 after bottoming out in 4Q19. MLCC inventory was extended up to 150 days in 4Q18 but by the year's end, it is expected to come down to 70 days. Meanwhile, Walsin saw a sales decline in 3Q19 as opposed to Yageo (OTC:YAGOY), due to the quality issue of the products that it supplies EMS. In October, Taiwanese MLCC makers saw their sales decline due to reduced number of working days following the Chinese National Day holidays. In November, however, sales should increase MoM and 4Q sales are expected to be flat QoQ. Taiwanese MLCC makers' exposure to China and sales distributors' share of sales have come down to normal levels. Thus, the MLCC price downcycle, triggered by Taiwanese MLCC names, appears to be coming to a close.

Yageo announced the acquisition of Kemet, a US-based company with strength in automotive, industrial, aerospace MLCCs, and tantalum capacitors (the company supplies under the Tokin brand in Korea and Japan). Yageo plans to finalize the deal by 2020 and plans to foray into the automotive and industrial capacitor markets. The automotive MLCC market is currently occupied largely by Murata (OTCPK:MRAAY) and TDK (OTCPK:TTDKY), with SEMCO and Yageo seeking to strengthen their presence. The automotive MLCC market is expected to grow exponentially going forward, given the soaring demand for EVs and ADAS. However, Yageo's acquisition of Kemet, a current supplier, is neutral on the market as product reliability is the most important issue in the market. Meanwhile, Murata's MLCC order backlog is recovering after bottoming in 1Q19. Recently, leading MLCC manufacturers are believed to attempt to raise the prices of ultra-small, high-capacitance MLCCs (mainly against Chinese smartphone makers), driven by increased demand for 5G smartphones.

Top picks Samsung and SEMCO; wait to bargain-hunt Hynix

Ongoing memory price declines and US-China tensions have weighed on earnings but Samsung and Hynix shares have significantly outperformed the broader market vs. at the beginning of the year. We believe this has to do with the market prematurely moving to price in the expectations for a rebound in memory chip prices in 2020 amid production controls and escalating 5G infrastructure investments. Additionally, there is growing optimism that edge computing related to 5G will boost AI demand and create fresh demand for DRAM.

Samsung has realized its "super-gap" strategic vision with the Galaxy Fold, which has helped quench the company's thirst for excess growth in areas other than memory semiconductors. The Galaxy Fold is especially significant in that it will drive the earnings of both the smartphone department and Samsung Display. In addition, while the semiconductor business is currently centered on memory semiconductors, there are growing expectations that the business will see growth in both top- and bottom-lines thanks to growing competence in 5G modem and CIS as well as an expanding client base seeking advanced technologies (7nm or lower EUV manufacturing process) through preemptive investments. In all, we expect Samsung to surge past its 2018 record earnings to report another record in 2022 as the earnings of semiconductors, smartphones and Samsung Display all grow in 2022. Hynix, on the other hand, needs to find fresh momentum in order to facilitate excess growth because shares have already priced in expectations for the DRAM cycle turnaround. In the case of NAND, it needs to enhance the profitability of its solutions. With expectations for a turnaround of memory semiconductors in 2020 already partly reflected into the price performance of both Samsung and Hynix, we believe Samsung is more attractive at this point since it has more potential for excess growth.

As for LGE, we recommend a trading strategy focusing on the seasonality of home appliances considering the continuing losses of the smartphone and VS divisions as well as intensifying competition in the TV set market. It is positive that LG Display's mobile OLED technology is getting better amid the expansion of white OLED capacity. As for SEMCO, we believe its folded camera module technology will gain attention as Samsung moves to increase the portion of smartphones with high-resolution rear cameras (64MP or higher) from next year. That the price declines of its cash cow MLCCs have lost steam is another plus. At LG Innotek, marginal business divisions have gone into restructuring and there is a possibility that the iPhone OLED model to be released in 2H20 will come with a rear ToF camera.

Korean IT shares have been outperforming the broader market. However, we believe the interest in the Korean IT H/W sector will grow even more in 2020 compared with global peers considering the expansion of the 5G services, memory chip price rebounds, Huawei's shrinking market share in the smartphone space, and BOE's struggles with OLED technology. Above all else, Samsung is likely to become a highly attractive investment as its "super gap" product Galaxy Fold is slated to enjoy larger sales.

Accordingly, among Korean IT H/W names, we favor Samsung the most, followed by SEMCO, Hynix, LG Innotek, and LG Electronics.

Meanwhile, the Taiwanese and Chinese server supply chains are also outperforming the market. Their stock momentum will likely remain solid as server demand is expected to recover in earnest from 2020.

Investment highlights

We maintain BUY and our six-month-forward target price of KRW61,000 (1.5x 2020F BPS) on Samsung Electronics. We now believe Samsung's 4Q19 consolidated sales and operating profit will come to KRW60.7tn and KRW6.3tn, falling short of our previous estimates by 0.5% and 6.3% due to a stronger KRW and weaker DRAM prices. In 1Q20, the launch of the Galaxy S11 is expected to shore up IM operating profit but overall operating profit will likely remain flat QoQ as unfavorable seasonality dampens semiconductor and display earnings. However, from 2Q20, server and PC DRAM price hikes should lead to a rebound of DRAM contract prices, and the increased shipments of hole-in-display flexible OLED will likely drive earnings higher. In 3Q20, operating profit should reach the KRW11tn mark first time in seven quarters as DRAM and NAND contract prices both rise. Samsung's DRAM and NAND bit growth in 2020, projected at 17.9% and 37.6%, will likely be faster than the market average. The IM division is expected to enjoy both qualitative and quantitative growth by shipping 314mn smartphones and 5mn Galaxy Folds, taking advantage of Huawei's contracting exports. Meanwhile, Samsung Display should begin supplying foldable display and flexible OLED display based on Y-Octa technology to Apple in 2H20. Samsung Display's earnings are expected to pick up in earnest as it begins to supply its single-hole punch display to Huawei, Oppo and Vivo.

Major issues and earnings outlook

Converting its DRAM Line 13 into an image sensor capacity is a necessary step toward increasing the shipments of high resolution image sensors, and we believe the move will help significantly boost the shipments of CMOS image sensors with high megapixel counts (64MP or higher). As the shipments of 5G modems, PMIC and RFIC also rise, system semiconductor earnings should remain on the solid growth track. We believe 2020 will mark the first year when growth is seen across all four major divisions-memory semiconductors, system semiconductors, smartphones, and displays.

Share price outlook and valuation

We recommend a buy-and-hold strategy as the four major growth engines are starting to pick up steam.

SK Hynix: Hyperscale market is getting bigger and bigger

Investment highlights

We reiterate BUY on SK Hynix and raise our target price from KRW98,000 (1.5x 2019F BPS) to KRW100,000 (1.4x 2020F BPS attributable to controlling interest). We now believe Hynix will report KRW6.68tn in 4Q19 consolidated sales and KRW379.8bn in operating profit, each 2.2% and 46% lower than our previous estimates, amid a stronger KRW and steeper-than-expected drop in DRAM contract prices. However, after hitting the bottom in 4Q19, we believe operating profit will sequentially improve from 2Q20 onwards. We expect operating profit to recover to KRW961bn in 2Q20 on DRAM contract price hikes and the considerable narrowing of NAND losses. In 3Q20, we estimate operating profit will reclaim the KRW2tn level as DRAM and NAND contract prices are both expected to increase more than 7% QoQ. While the dwindling smartphone exports of Huawei, one of Hynix's strategic clients, is a cause for concern, we believe Huawei will be able to offset the decline in exports with robust demand in the domestic market. Additionally, as Chinese telcos move to build the 5G infrastructure, Hynix's server DRAM sales to Huawei should surge in 2020. It is believed that Huawei accounts for more than 80% of servers supplied to Chinese mobile carriers.

Major issues and earnings outlook

Among the three major DRAM manufacturers, Hynix has the largest exposure to server DRAM. Accordingly, it is poised to benefit the most from the robust growth (+10.5% YoY) of hyperscale server demand to be generated by data centers and telcos from 2020 onwards. We believe hyperscale server demand will represent up to 78% of server DRAM demand on the back of the growth of average per device memory content. As for NAND, solutions products are likely to become more competent on client diversification for controller IC. We expect the NAND business to turn to profit in 3Q20 when NAND contract prices gain strong upside momentum.

Share price outlook and valuation

We recommend a buy-and-hold strategy in light of the recovery of hyperscale server demand.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Additional disclosure: Hyundai Motor Company is a passive shareholder in our bank.

See the original post here:
Korea IT H/W: Key Takeaways From Our Q4 2019 Visits To Japan/Taiwan - Seeking Alpha