1、MMGlobal SemiconductorMemory Hope,but No RecoveryMorgan Stanley does and seeks to do business with companies covered in Morgan Stanley Research.As a result,investors should be aware that the firm may have a conflict of interest that could affect the objectivity of Morgan Stanley Research.Investors s
2、hould consider Morgan Stanley Research as only a single factor in making their investment decision.For analyst certification and other important disclosures,refer to the Disclosure Section,located at the end of this report.+=Analysts employed by non-U.S.affiliates are not registered with FINRA,may n
3、ot be associated persons of the member and may not be subject to NASD/NYSE restrictions on communications with a subject company,public appearances and trading securities held by a research analyst account.A positioning rally that caused memory stocks to perform out of sync with fundamentals is losi
4、ng steam,stock-specific factors are coming into play.Memory conditions have taken a sharp turn for worse,earnings risk is mounting,and our inflection signposts suggest extreme caution from here.March 3,2019 06:21 AM GMTMMMORGAN STANLEY ASIA LIMITEDShawn KimEquity Analyst+8523963-1005Shawn.K Contribu
5、tors MORGAN STANLEY&CO.LLCJoseph MooreEquity Analyst+1212 761-7516Joseph.MMORGAN STANLEY ASIA LIMITEDJonathan F GarnerEquity Strategist+8522848-7288Jonathan.GMORGAN STANLEY TAIWAN LIMITEDCharlie ChanEquity Analyst+8862 2730-1725Charlie.CMORGAN STANLEY MUFG SECURITIES CO.,LTD.Kazuo Yoshikawa,CFAEquit
6、y Analyst+813 6836-8408Kazuo.YMORGAN STANLEY TAIWAN LIMITEDDaniel Yen,CFAEquity Analyst+8862 2730-2863Daniel.YMORGAN STANLEY&CO.INTERNATIONAL PLC,SEOUL BRANCHRyan KimEquity Analyst+822 399-4939Ryan.G.KMORGAN STANLEY ASIA LIMITEDDaisy DaiResearch Associate+8522848-7310Daisy.DMORGAN STANLEY ASIA LIMIT
7、EDAndrew ChungResearch Associate+8522848-6575Andrew.CMORGAN STANLEY ASIA LIMITEDYinan Zhang,Ph.D.Quantitative Analyst+8523963-3507Yinan.ZMORGAN STANLEY&CO.LLCEthan PuritzResearch Associate+1212 761-1744Ethan.PMMGlobal SemiconductorMemory Hope,but No RecoveryMore vulnerable after the rally.Due to a s
8、harper deterioration in memory market conditions,with significant oversupply ahead,we reduce our 2019-20e EPS estimates by 21-47%across global memory coverage and downgrade Nanya Tech to UW.We remain UW on stocks SK Hynix and specialty memory due to one of the most challenging cycles in history.We a
9、re now 20-40%below consensus and see 25-30%more downside to all key memory plays in this sce-nario.We see stocks moving back to trough multiples after the YTD positioning rally,and driven by earnings expectations falling sharply.Top-down,we see better value in other cyclical industries in the early
10、phases of Chinas economic upturn cycle.One of the most challenging memory cycles in history.We esti-mate DRAM will stay fundamentally oversupplied until 3Q20 due to a supply mismatch,inventory build throughout the year and despite demand for cloud spending,PCs and phones improving in 2H19.Supply and
11、 inventory dynamics parallel the 1996 downturn rather than 2016,and thus do not appear well understood by the market.Pricing should mean revert back to the long-term price/Gb trend line,and for DRAM prices this means a peak/trough DRAM ASP of-70%(or-40%from here back to 2016 lows),and NAND prices be
12、low gross margin and potentially near cash costs.This is beginning to play out but is clearly not in share prices.WHATS CHANGEDSK HynixFromToPrice Target63000.0060000.00Inflection signposts sell signals accumulate.We would be pre-pared to become more constructive under the right conditions,but ultim
13、ately it is about inflections in the cycle and trajectory of earn-ings estimate revision breadth the former relates to pre-trading a recovery post the YTD re-rating and latter to negative earnings risk mounting.Our belief is that(i)this cyclical downturn is headed into a deeper recession;(ii)recent
14、growth indicators,such as cloud improving are masked by mounting inventory;and(iii)earnings esti-mate achievability for memory stocks still appears distant.(iv)Valuation is less compelling after the YTD rally,and(v)tactically buying on weak data releases no longer makes sense;we expect better entry
15、points in future.Conclusion.While we are negative on the memory sector,we do see opportunities to generate alpha with carefully chosen stocks and recommend companies leveraged to an early cloud demand recovery:Aspeed and Intel(see Cloud capex semi plays bottoming out).We stay UW on stocks exposed to
16、 commodity-linked memory(SK Hynix,Nanya Tech,Advantest);EW Micron,WDC and Samsung;and UW specialty memory stocks Giga Device,Winbond and Macronix.A positioning rally that caused memory stocks to perform out of sync with fundamentals is losing steam,stock-specific factors are coming into play.Memory
17、conditions have taken a sharp turn for worse,earnings risk is mounting,and our inflection signposts suggest extreme caution from here.Industry ViewS.Korea TechnologyCautiousGreater China Technology HardwareIn-LineM4M Contents 5Memory The Sell Signals Accumulate101)Cycle Watch Too Early for a Second
18、Derivative Inflection142)Growth Indicators Not Supported by Data163)Earnings Estimate Revision Indicator Deeper Sector Earnings Recession194)Valuation Less Compelling but Still Attractive225)Tactical Indicators How Long Will the Sentiment Rally Continue?26Why We Think a Recovery in 2H for DRAM is Hi
19、ghly Unlikely28Implications for US Semi Stocks31Implications for Asia Semi Stocks39Implications for Japanese Semi Cap Equipment Stocks40 Risk-Reward:Samsung Electronics(005930.KS,W47,350,EW,PT W40,000)41Financial Summary:Samsung Electronics42Risk-Reward:SK Hynix(000660.KS,W75,400,UW,PT W60,000)43Fin
20、ancial Summary:SK HynixMMORGAN STANLEY RESEARCH5MOne of the most challenging memory cycles in history.Playing the second derivative inflection in DRAM(meas-ured in terms of the monthly rate of change in contract pricing illustrated in Exhibit 1)could still see contract pricing fall a long way as cou
21、ld share prices,and we fore-cast material downside for memory prices and earnings.Memory valuations are low,but earnings expectations are not.Both the cycle and demand present high downside risk,and structural growth in cloud continues to undershoot the industry long-term perceived expectations for
22、memory demand,which has led to the current supply driven boom and bust cycle.Capex and production cuts(NAND)may provide some relief,but the need to reduce costs through technology migration continues to drive production(well above demand),which hits pricing,hence,we think it is too early for a cycli
23、cal upturn trade.Exhibit 1:The Memory Cycle Disconnect Multiples Pre-Trading a Turn in the Cycle Too Soon-58%-80%-40%0%40%80%120%0.50.91.31.72.12.52005200620072008200920102011201220132014201520162017201820192020DRAM EV/SalesDRAM Contract Price YoY(RHS)Source:DRAMExchange,Morgan Stanley Researchcompo
24、unded by a major leg down in earnings.We reduce our 2019-20e EPS estimates by 21-47%,now 20-40%below consensus and could see 25-30%more downside to all key memory stocks in this scenario.We downgrade Nanya Tech from EW to UW with 24%downside to PT.We struggle to see any fundamental improvement this
25、year,yet consensus for the sector remains relatively optimistic,despite major cuts since November 2018.We see most companies outlooks and guidance as likely to move sharply lower during CY1Q19,in our view,and put into question the narrative on a per-ceived second half demand recovery.Memory The Sell
26、 Signals AccumulateExhibit 2:One of the Most Challenging Memory Cycles in History Tracking the 1996(PC)and 2000(Y2K)Downturns.0.00.40.81.21.6M-6MM+6M+12M+18M+24M+3096 Cycle00 Cycle10 CycleRebased to M-6 Server DRAM at$100*Server DRAM at$120*Assuming constant server DRAM sales volume based on current
27、 server DRAM asp of$200 Source:WSTS,Morgan Stanley ResearchExhibit 3:But Disconnect with Share Price Performance Pricing a Cyclical Recovery0.40.81.296 Cycle*00 Cycle15 Cycle18 Cycle*Excluded SK Hynix as the Company was private until Dec-96 Rebased to M M M+6 M+12 M+18 M+24 50%to Peak Source:Thomson
28、 Reuters,Morgan Stanley ResearchOur key signposts for a trough in memory are distant.We acknowledge that a lot of bad news was priced in December,how-ever,the YTD rally has reduced the attractiveness of memory valua-tion too early.While we would not exclude the positioning rally extending further,we
29、 think fundamentals matter more than tech-nicals,and we see clear incremental risk that the earnings recession ahead for our sector is even deeper and longer than we currently forecast.We expect consensus earnings estimate revisions are likely to be some of the most negative we have seen in the last
30、 three dec-ades of DRAM history,thus,investors need to pay attention to these M6Mchanges since earnings achievability will likely determine whether stocks can continue to rally,and which ones remain vulnerable.We have analyzed the signposts to watch for DRAM and NAND stabi-lization across 5 factors,
31、which we believe are necessary to call a fundamental trough for stocks all point to caution on valuation,with some exception:1.Cycle watch too early for a second derivative inflection.We have compared the current cyclical downturn against prior cycles across five peak-to-trough absolute and relative
32、 factors.Capital spending has been too high in the last two years with back-to-back years of big increases in DRAM cap-ital spending,up 81%in 2017 and another 40%in 2018.This is causing an extended oversupply that we think will take much longer than previous normal cycles to reach equilib-rium.Aside
33、 from cloud,core end markets continue to deterio-rate and inventory is unlikely to clear all year,suggesting a steeper path for price erosion.2.Growth indicators false sense of confidence in a 2H demand recovery.Memory cycles are supply driven and the memory revenue cycle is often independent of the
34、 economic cycle,but an economic downturn(in-line with our global GDP estimate)can exacerbate a memory downturn.As such,we believe the magnitude of an economic recovery would be not relevant for memory in 2H,and a fully supportive Fed policy and China stimulus would not change the cycle.Nevertheless,
35、we expect demand for PCs and phones to improve seasonally on new product cycles,and we are encouraged that the overall trajectory of cloud spending,which still appears relatively robust.3.Earnings still too optimistic and this is not well known.We struggle to find any fundamental EPS improvements in
36、 2019-20 as the weakness is just beginning.We see significant downside risks to DRAM pricing to mean revert back to pre-vious trough levels(-40%from here),with the revision breadth likely to turn significantly worse(than the 2016 downturn)before finding a bottom.NAND pricing likely moves below gross
37、 margins for most producers(ex-Sam-sung)and near cash costs.We cut 2019-20e EPS forecasts by 21-47%are now 20-46%below consensus numbers.Memory stocks relative earnings estimate revisions breadth has declined significantly since late last year,and while such extreme readings are typically good buyin
38、g signals,we think relative performance is now ahead of the turn in relative neg-ative revisions.4.Valuation appears reasonable,although less compelling after rally.Stocks are at-1SD below mid-cycle valuations and 16%and 54%from trough P/B and EV/Sales multiples,respectively.P/E is a contrarian indi
39、cator and currently stands above+1SD mid-cycle multiples,but this should be looked at relative to trough EPS which is likely to decline sharply.If fur-ther evidence emerges of a deeper downturn in 2019,or if we get no signs of recovery in 2H demand,then we expect stocks to move closer to trough mult
40、iples,as in past downturns.The real problem on the upside is that we see fundamentally no driver for continued sector demand and/or EPS growth.This will limit upside to valuations,in our view.5.Is the stock(not fundamental)risk/reward better now?Tactically,the market appears complacent,per our strat
41、e-gists(FX Analytics:Top Indicators We Are Watching),with the sentiment index(Global Risk Demand Index STGRDI*)and a recent survey showing long EM as the most crowded trade for asset managers.But the risk outlook is far from rosy increased confidence,including de-escalating trade ten-sions,became a
42、key driver of unprecedented funds flows into risk assets in EM and semis,which we believe has created a distorted perception of the memory cycle.From a tactical perspective,we think stocks correct sharply from here to reflect underlying fundamentals after the YTD positioning rally and recommend inve
43、stors stay cautious.Top down,our EM strategist,Jonathan Garner,acknowledges much improved absolute and to a lesser extent relative valuations versus the MSCI EM benchmark,but see better value in other cyclical industries(most notably Materials and Industrials in the early phases of Chinas economic c
44、ycle turning up).Earnings revisions breadth has yet to turn and we think other industries will inflect upwards earlier.MMORGAN STANLEY RESEARCH7MExhibit 4:South Korea Semi Export vs.SK Hynix Stock Price Retracing Fundamentals200040006000800010000120001400010000200003000040000500006000070000800009000
45、01000002010201120122013201420152016201720182019SK Hynix Stock PriceSouth Korea Exports-Semiconductor(RHS)Source:Bloomberg,Morgan Stanley ResearchExhibit 5:Funds Flow Weakening Cumulative Foreign Inflow Into Samsung+SK Hynix,Pace Slowing-5-4-3-2-101234Jan-18 Feb-18Mar-18 Apr-18 May-18 Jun-18 Jul-18 A
46、ug-18 Sep-18 Oct-18 Nov-18 Dec-18 Jan-19 Feb-19Billions Saumsung+HynixRest of Kospi2Source:Bloomberg,Morgan Stanley ResearchExhibit 6:Samsung Electronics Shares Appears Tactically Overbought.2000025000300003500040000450005000055000600002016201720182019Last PriceSMAVG(50)SMAVG(100)SMAVG(200)Source:Bl
47、oomberg,Morgan Stanley ResearchExhibit 7:.as well as SK Hynix Shares20000300004000050000600007000080000900001000002016201720182019Last PriceSMAVG(50)SMAVG(100)SMAVG(200)Source:loomberg,Morgan Stanley ResearchM8MStock implications we are warming to pure cloud exposure but not to memory.We stay UW tho
48、se exposed to commodity-linked memory stocks(e.g.,SK Hynix),and companies that have rallied YTD where we think them likely to miss consensus by a wide mark.We would take profits across all memory and specialty stocks.Pricing is tracking historical mean reversion on a price per gigabit basis,and this
49、 should create a large incremental headwind to forecasts in 2019 and 2020.We see 11-14%downside to our base case price targets,and our bear case revolves around a longer down cycle similar to 1996,and politics and trade tensions affecting markets,where an alignment of technicals and fundamentals dri
50、ve multiples closer to GFC crisis levels or 29-40%lower from here.Key event catalysts for share prices near term include:(i)pricing negotiations which have become ad hoc(no longer quarterly)and rate of change in inventory both leading indicators of the cycle;(ii)expected pre-announcements and manage
