瑞信-美股-半导体行业-美国半导体2019年展望：就像糖果店里的孩子-2019.1.13-73页.pdf

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.13 January 2019 Americas/United States Equity Research Semiconductor Devices Semiconductors Research Analysts John W.Pitzer 212 538 4610 john.pitzercredit- Ada Menaker 212 325 3472 ada.menakercredit- Charles Kazarian 212 538 4160 charles.kazariancredit- Jeff Pullen 212 325 3837 jeff.pullencredit- SECTOR FORECAST 2019 Outlook:Like a Kid in a Candy Store Reset 1H,Rebound 2H.We entered 2018 cyclical cautious,hopeful secular drivers would re-assert in 2H18.China/Macro and Handset/Crypto have lengthened/deepened the correction albeit models still support a modest reset.With 3%Global GDP,CY19 Semi(ex-Memory)growth will be 0.5%vs.CY18 of 7.6%.Trough in C2Q at-2.0%y/y;rebound to+4.1%y/y by C4Q.Stocks Discounting Much.After increasing 7 of 9 yrs and outperforming 4 of 5,the SOX declined 8%in 2018,underperforming the SPX by 200 bps.Peak-trough stocks declined 25%vs.prior 12 cycles of 26%,relative underperformance was 1,480 bps vs.prior 12 cycles of 1,150 bps.but FTM EPS Is Not.SOX FTM EPS is still too high.SOX FTM EPS peaked 10/5/18 at$100 but has only declined by 5%to-date the average peak to trough revisions over the last 12 cycles(ex 2009 financial crisis)has been 30%w/the mildest correction in 2014 at 16%,and the severest in 2002 at 65%.Cycle Metrics Mostly Benign.The CY18 upturn was well-behaved,with 8 of 10 cyclical metrics below prior peak median.The duration/magnitude of Semis over-shipping end-demand has been 86%/23%of normal.In C4Q/C1Q Semis are under-shipping end demand for the first time since 1H16 an historically bullish tactical indicator.Valuations Compelling If Reset Is Modest.Stocks are currently trading at EV/FCF of 14x,a 32%discount to SPX well below 5-yr average of 24%.Assuming average reduction in FTM EPS,stocks would be trading at a 3%discount to SPX,assuming mildest/severest reduction,stocks would be trading at 19%discount/94%premium.Secular Drivers Are Intact.Semis/SCE moving from GDP/Semi minus growth to GDP/Semi plus growth.Semis are more integral to the Global Economy with fewer producers driving increased value.SOX EPS of$100 is 110%higher than any other peak and while cyclically“scary”,growth was M&A,buybacks,and improved profitability,NOT cyclical excess.Data Is Unifying Theme.We view Semis as the cheapest,most levered play on the Data Economy.AI/ML lowers the cost of analytics for the first time,driving elasticity of use-cases underpinning a Silicon TAM of$750 bn by 2025(6.8%CAGR).A 1%value capture of Global COGS,OpEx and CapEx would increase Silicon TAM by$420 vs.CY18 TAM of$474 bb.2019 Focus Areas:(1)Cyclical bottom/Valuation,(2)Infrastructure/5G,and(3)Analytics/DataCenter.Avoid Handsets until 2020.Our top 3 ideas are MU,XLNX,and NXPI and by theme:(1)NXPI,MCHP,MRVL(2)XLNX,ADI,KEYS and(3)MU,NVDA,INTC.In SCE/EDA our top ideas include:LRCX,AMAT and SNPS.Be tactical into C4Q Earnings Season;more aggressive after.13 January 2019 Semiconductors 2 Executive Summary Semi Industry Growth Forecast:We expect Semi Rev to decline 3.0%in CY19,the first y/y decline since 2015(-0.5%y/y)and a significant deceleration from the 14.3%growth in CY18 and the 3/5 year CAGR of 12.1%and 8.9%.Ex-Memory,we are modeling Semi Rev+0.5%y/y,the slowest since 2015(+0.3%y/y),vs.7.6%in CY18 and below Street estimates embedding 3.6%y/y growth.We model C1Q Rev 100 bps below seasonal vs Street 50 bps above seasonal.After peaking in C2Q18 at up 11.9%y/y,we see ex-memory Rev turning negative in C1Q at down 0.1%y/y,troughing in C2Q at down 2.0%y/y with C3Q declines moderating to down 0.2%y/y,and returning to growth in C4Q at up 4.1%y/y.The industry should begin to show accelerating growth exiting CY19 as compares start to moderate with C4Q18 up 1.0%y/y versus C1Q-C3Q18 up 10 y/y.Our preliminary 2020 growth estimate is+5.0%,+5.5%ex-Memory._Semi WFE Forecast:After growing 35%/8%y/y in CY17/18,we expect CY19 WFE to decline 17%y/y to$42b,lower than current bottom-up Street estimates which embed a 6%decline.CS Japan forecast is$40b,down 20%y/y.We expect C1Q19 to be an absolute WFE bottom($9.5bn,down 35%y/y),recovering to 11%y/y by C4Q19;we see upward bias to WFE in 2H.We are expecting CapEx to Rev of 19.1%,below the 5-yr avg of 19.3%but above the 10-year avg of 17.9%,with Memory CapEx to Rev of 27.5%,below the 5/10 year avg of 32%/29.2%,and logic CapEx/Rev at 15.2%,above the 5/10 year avg of 14.7%/14.3%.We are expecting NAND/DRAM/Foundry spending to decline 28%/23%/4%,and Logic spending to remain roughly flat.Relative to linearity,we expect C2H to be stronger representing 55%of overall WFE,vs.5-year average of 50%.We are expecting a likely resumption of WFE growth in 2020 to up 14%y/y to$48B which should provide the basis for Semi Cap companies to see growth of 14-15%y/y given 15%y/y growth in Services.CYCLICALSECULAR 13 January 2019 Semiconductors 3 Our Cyclical View:While less exposure to PC/Handsets with new applications across Industrial,Auto,and Infrastructure is a clear positive,the start to CY19 is a stark reminder that Semis are and always will be cyclical an underappreciated view for most of CY18.PC/Handsets are still 40%of Rev dwarfing any other application.Supply is still fungible across end-markets,and lengthening lead-times and upward ASP still creates incentives for end-customers to build inventory.While Semis will not be as cyclical as the 1990s,it is likely to be MORE cyclical than the last 7-9 years especially if our structural thesis of rising ASPs and slowing supply growth continues.While increased cyclicality tends to dampen multiples,we continue to argue that improving structural growth and returns should provide the means for a re-rating of the Semiconductor sector over time especially as the Data Economy accelerates._ The Cycles Impact on 2019:Our analysis supports a mostly well-behaved upturn with 8 out of 10 cyclical metrics below historical medians and most importantly Semi over-shipping which while 86%of normal duration was only 23%of normal magnitude.We expect any excesses to be worked thru by C1H19.While Semi Rev(ex-Memory)was above seasonal from C3Q16 to C2Q18(aggregate of 1,700 bps)C1Q15 to C2Q16 was below seasonal(aggregate of 1,580 bps).Seasonal growth from C4Q14 implies C4Q18 Semi Rev only 4 ppts above trend and C1Q19 only 1 ppts above trend.While Nov IC Units(ex-Memory)are currently 6-7%above trend and C3Q Supply Chain Inventory of 43 days is at the high-end of its post crisis range of 38-44 we expect C3Q18 thru C1Q19 Semi Rev(ex-Memory)8-9 ppts below seasonal.Importantly,after Semis over-shipped end-demand from C3Q16-C3Q18 we expect Semis to under-ship end-demand in C4Q18/C1Q19/C2Q19 which has historically proved to be a bullish tactical indicator._ Stock Discounting Most of a“Normal”Correction.After increasing 7 of 9 yrs and outperforming 4 of 5 yrs,the SOX declined 8%in 2018 underperforming the SPX by 200 bps and by 1,290 bps since 3/12/18 the worst underperformance since 2012.Peak-trough underperformance for the SOX vs the SPX was 1,480 bps in CY18,with a bottom on 12/12/18.P-T underperformance over prior 12 cycles averaged 1,730 bps with a range of 640 bps(2016)to 3,350 bps(2005).The SOX is down 20%off the peaks vs median/average declines over the past 12 cycles of 26%/30%.A normal correction would imply 6-10%further downside.However,4 periods saw 30-35%declines implying 10-15%downside.The SOX typically troughs 5-6 months after it peaks,with a high of 8 months.Using 8/27 peak levels(when the SOX re-approached the 3/5 peak)would imply a bottom in late Jan thru late Feb,or late April at the high-end.13 January 2019 Semiconductors 4 FTM EPS Is NOT Discounting Enough.SOX FTM EPS peaked on 10/5 at$100 and has declined by only 5%to-date.While stocks reflect negative sentiment with the SOX down 20%from peaks,FTM EPS is still too high and needs to come down in 1H19.The average peak to trough EPS revisions over the last 12 cycles(ex 2009 financial crisis)has been 30%with the mildest correction in 2014 at 16%and the severest correction in 2002 at 65%;2009 was 140%.The SOX EPS typically troughs 6-7 months after it peaks,with a high of 12 months(2007).Using the 10/5 peak implies a EPS trough in early April to early May,with the high-end implying EPS doesnt trough until early October.While we expect most to meet C4Q guidance(aggregate Rev down 4.0%q/q vs seasonal up 0.5%q/q),C1Q Rev of down 4.0%q/q vs seasonal of down 4.6%q/q likely has 3-5 ppts of downside more for handsets,less for infrastructure._ Semis and the Macro:CS Macro is estimating 2019 GDP of 3.1%y/y,down from 3.8%in CY18.Recent China data points and China/US Trade tensions have skewed expectation to the downside.2015 was the only year since 2000 that Semi Rev declined with GDP growth of at least 2.5%.Semi Rev growth has a 0.78 correlation to ISMs Global Manufacturing Index with higher elasticity to accelerations than decelerations(2x-3x vs 2.5%average 2.3x(1x 3.7x).,Semi Rev has been up every year since 2000 when real GDP grows at least 2.5%._ Our View on China Trade:While limited domestic Chinese production insulates US Semis from direct tariffs by China,US Semis are impacted directly/indirectly by China/US Trade tensions as China represents 50%of ship-in Rev and 25%of consumption.China is 19%of Global GDP but approximately 25%of PCs,24%of Handset,30%of Auto,and 22%of Infrastructure CapEx,per SIA.Macro weakness in China,whether trade related or not,will impact Semi demand.China/US tensions are more than trade deficits.IP protection is key;which is particularly germane to Semis.While NT volatility is likely(incl.potential for SCE embargos worst case),ultimately Semis win in an environment where IP is respected especially Memory.Given MOFCOMs(now decade-long)history of using regulatory approval to extract concessions,we expect delays and blocks to continue especially as CIFUS becomes increasingly concerned by said concession.M&A will happen,but the process will be longer.13 January 2019 Semiconductors 5 Our Secular View Growth to Reaccelerate:Semi Rev represented 0.50%of nominal GDP,well below peak of 0.61%in 2000 and the same level as 1994/95,despite Semis being a more indispensable commodity today than 23 years ago.Relative declining chip value over the last 2+decades was caused by ASP pressures after 30+years of upward bias.Rising barriers to entry,consolidation,and slowing supply growth provide the foundation for better pricing.Semis are poised to transition from GDP-minus to GDP-plus growth.In addition,Semi Cap should move from Semi-minus to Semi-plus growth after 15 years of 300 mm productivity gains which now appears to be in arrears.We see LT growth in Semis increasing from 3-5%to 7-9%and for Semi Cap from 2-4%to 8-10%.Neither scenario incorporates our view on“Data”as a demand driver which could add an additional 1-300 bps to LT growth rates._ SOX EPS Growth More Structural than Cyclical:While EPS peaked at$100 in Oct well above prior peak of$48 in Oct.2014 and well above the implied trend line growth of$60 EPS looks more structural than cyclical:(1)Content Growth,(2)New Applications/End Markets,(3)Improved profitability,and(4)M&A/Buybacks.Over the past 12 cycles median T-P EPS growth has been 24%with the largest growth post financial crisis of 120%in 2014($22 to$48)and overall at 777%in 2004($3 to$24).Median P-T EPS decline has been 27%with the largest decline post crisis of 38%in 2011($40 to$25)and overall at 65%in 2002($8 to$3).After EPS troughed in 2/16 at$39,it grew 155%to$100 over 32 months.While this upturn has been the most pronounced since 2010 it is broadly similar to the 11/12-10/14 upturn when EPS grew 120%over 22 months.Further,that EPS upturn was only followed by a 16%EPS decline versus the 5%we have seen to-date.EPS growth above trend(i.e.above$60)was 40-50%driven by M&A/Buyback,15-20%by improved profitability,and 15-20%by new applications/end markets.Overshipping cycle to date has still been significantly more modest than typical._ Our View on Memory Cyclical and Structural:While we are modeling the first y/y decline in Memory(-10%y/y)since 2016,we still see more evidence to support structural improvements:(1)Consolidation,(2)Structurally slowing supply growth,(3)CapEx reductions ahead of peak profitability,and(4)New Demand drivers in Data Center/infrastructure.We expect CY19 DRAM/NAND Rev to decline-9%/-12%y/y versus 2018 of+47%/+29%.We model CY19 DRAM/NAND bit growth of 16%/35%.We model DRAM/NAND ASPs to decline 25%/40%.We expect ASP stability in DRAM/NAND by 3Q19 and 4Q19 respectively.We model DRAM/NAND CapEx down 23%/25%y/y,yielding DRAM CapEx/Rev of 18%below the 10 year average of 23%and NAND CapEx to Rev of 39%,below the 10 year average of 45%.DRAM/NAND CapEx peaked in 1Q18.13 January 2019 Semiconductors 6 High exposure to Smartphones is a headwind(DRAM 33%,NAND 38%).Accelerating Compute TAM,5G and SSDs are tailwinds.MUs F2Q19 Rev decline of 25%,the worse since 2009 implies materially under-shipping demand.The Data Economy Represents Upside:In C2Q14 we introduced our“Data Paradigm”consisting of 4 key areas:Creation,Storage,Transmission and Analytics.We view Semis are key enablers to and the most levered,cheapest investments on Data Driven Economy.Meaningful cost declines in Create,Store,Transmit have to date been matched by cost increases in Analytics its why we only analyze 1%of the data we create.AI structurally lowers the cost of analytics leading to elasticity of use-cases.AI is a silicon-centric solution,which should allow Semis to capture value in the global economy after 20+years of losing value.A 1%value capture by AI would expand Semi TAM by$420 bb against a CY18 TAM of$470 bn.We see Data adding 1-300 bps to our LT Semi growth rate of 7-9%and could underpin a Semi TAM which approaches$750 bb by 2025 we are on record in 2015 saying$500 bn by 2020._ China Inc.Continues to Invest:Government-aided entities continue to invest in Semis to bridge the gap between domestic Semi production(7%)and consumption(25%);we are currently tracking 15 active fab projects,not to mention continued ecosystem investment in design and back-end.Efforts to buy IP via acquisitions have largely been halted by regulators.We are modeling China domestic CapEx for CY18/CY19 at$7.25bn/$8.3bn up 76%and 14%y/y respectively,with a bias towards brick and mortar,and estimate domestic China WFE for 2018/2019 at$3.2bn/$3.6bn up 60%and 13%y/y respectively Given that China is still in the learning phase,we see no material impact to global supply in NAND/DRAM unti