瑞信-美股-半导体行业-半导体：周期性底部和长期期权性抵消了低迷的复苏-2019.4.14-176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.14 April 2019 Americas/United States Equity Research Semiconductor Devices Semiconductors Research Analysts John W.Pitzer 212 538 4610 john.pitzercredit- Charles Kazarian 212 538 4160 charles.kazariancredit- Ada Menaker 212 325 3472 ada.menakercredit- QUARTERLY Cyclical Bottom and Secular Optionality Offsets Muted Recovery Bottom Line:We expect C1Q mostly IN LINE with a modest downward bias to C2Q/2H.Results will validate a C1Q bottom albeit one dampened by a more modest recovery as macro uncertainty and sluggish demand linger.With the SOX up 30%YTD and at all-time highs,with FTM EPS having modest downside,with revisions at 50%of a normal correction,and with valuations at 5-yr relative highs to SPX,the sentiment is overwhelmingly skewed towards stocks consolidating/retreating thru earnings the Semi Bull is as elusive as the Higgs Boson.We view the“wall-of-worry”as constructive and continue to have a positive bias more than supportive of our CY19 SOX target of 1650 established at the beginning of CY18.Focus on the following:(1)Semis are under-shipping consumption for the first time since early 2017,(2)Inventory is elevated BUT declining especially at end-customers,and(3)Capacity adjustments commenced 3-6 months earlier than usual.More important,structurally Semis still screen as significantly undervalued:Rising barriers to entry,slowing supply growth,and end demand less levered to consumer provides the foundation for structurally improving pricing/profitability and a Semis CAGR that is likely to accelerate from 3-5%to at least 6-8%.In addition,Semis are critical to the Data Driven Economy and as 5G lowers the cost of data transport and AI/ML lowers the cost of data analytics,we see new applications adding 200-400 bps of CAGR upside.Focus too much on the near-term at your own peril.Semis are transitioning from GDP-minus to GDP-plus,and we suspect as the industry demonstrates a track-record of consistently outgrowing GDP,investors will be forced to recognize the transition to a Data Driven Economy opens up new TAMs for Silicon/SW while being at least disruptive if not outright deflationary for traditional industries providing the mechanism for significant value transfer to and multiple expansion for the silicon economy.While investors recognize“SW is eating the world,”they have yet to recognize“Semis have a seat at the dinner table too,”underpinning a$750 bn TAM by 2025 implied 7%CAGR.Tactically based on 3 factors:(1)Investor sentiment,(2)FTM EPS Revisions,and(3)Valuation we prefer the following heading into C1Q/C2Q earnings:(1)SCE stocks LRCX and AMAT no expectation of 2H recovery,solid trough profitability with buybacks,and undervalued cyclically and structurally,(2)5G levered plays XLNX and KEYS too much concern on China inventory builds,upward bias to FTM EPS,albeit not cheap and(3)Diversified Semis NXPI and TXN cheap either relatively and/or absolutely,neutral to negative investor sentiment.INTC should be at least In Line with and MAINTAIN CY19 guide.We remain restricted on NVDA and MLNX and have not focused on off-quarter companies(AVGO,MU,ADI,MRVL)for this report.14 April 2019 Semiconductors 2 Investment Summary Figure 1:Top Picks Source:FactSet,Company data,Credit Suisse estimates Our Secular View Growth to Reaccelerate Still Structural Bulls on Semis:We continue to be Structural Bulls on the Semiconductor Sector our argument centers on Semiconductor Revenue growth that we believe is poised to move from GDP minus to GDP plus against valuations on EV/FCF that are still 20%BELOW the SPX despite faster growth,higher margins and better cash returns.From the early 1970s through the late 1990s,the Semiconductor Industry was a GDP plus grower in large part because ASPs enjoyed a 4%CAGR driven not by Semiconductor companies raising“like for like”pricing as much as by the Software industry developing new applications which demanded faster,more expensive chips every year.Figure 2:Semi Rev as%of Nominal GDP Accelerated Thru 2000 Figure 3:Semi ASPs Increased 100%from 1970-2000 Source:Company data,SIA,Bloomberg Source:SIA By the end of the 1990s,the industry experienced a confluence of events which upset the pricing dynamic specifically:(1)A lowering of barriers to entry with the advent of the Foundry model and a subsequent relative explosion of new entrants/IPOs in the Fabless PerformanceCurrentC1Q19Ticker20182019 YTDCommentary112AMAT-36%31%10%upside to EPS driven by 3D NAND and Display CapEx.Potential for multiple expansion given 20%discount to S&P.213LRCX-26%42%Well positioned to benefit from recovery in Memory CapEx.Compelling valuation.35ADI-4%33%Long duation,stable sockets,highest B2B exposure in Semis.LLTC provides scale advantages,LT FCFPS of$9+.43NXPI-37%37%Discounted valuation despite strong Auto growth prospects and among the best incremental operating leverage515KLAC-15%38%Positioned to benefit from Increasing CapEx/Rev for Logic/Foundry and Gen 5.Potential for 20%+multiple expansion.614ASML-10%30%Growth inflecting higher,EUV de-risked,Litho Share has bottomed,and there is upside to estimates.79TXN-10%24%Street under-modeling Rev growth and OpM expansion.Market share leader in Analog generating sustainably strong FCF.82XLNX26%58%At a multi-year inflection of operating leverage,potential for LT OpM of 35%-incremental confidence in 4-8%Rev growth target.96MCHP-18%33%Consistent execution and strong track record of M&A to augment organic growth.1010AVGO-1%25%20 existing franchises drive steady FCF to support dividend,buybacks,and M&A.114KEYS49%48%Leader in test/measurement and positioned to benefit from 5G.127INTC2%20%Transition from PC to Data-centric well underway,estimates overly conservative-Memory OpM upside,sustainable DCG re-acceleration.138MRVL-25%47%Successfully navigating a turn-around that started with cost cutting and will ultimately end in sustainable growth/OpM leverage.141MU-23%32%Industry supply discipline,strategic value is underappreciated.Free option on 3D NAND and XPoint.1511SNPS-1%40%Fastest growing and best quality EDA company.1616MXIM-3%17%Best in class shareholder returns,execution on cost saving measures,fairly valued.1717CDNS4%49%Solid management team but full valuations.Decelerating revenue growth and higher exposure to Semi M&A.1919TER-25%41%Cyclical risk in 2018,but benefitting from UR and slowing productivity improvements in Test.2020CY-17%26%Rev optionality from USB Type-C and BRCM IoT,however China and Memory remain fundamental concerns.2121BRKS10%20%Benefitting from Growth in WFE and increasing mix of lifesciences.2222RMBS-46%48%Transitioning from an IP licensing company to a diverse product and tech company2323ON-21%38%Typically underperforms a y/y growth peak.Valuation looks cheap on P/E but fair/full on EV/FCF.2424RTEC-14%14%Well poised to outgrow the market due to growth in advanced packaging,but rich valuation.NANANVDA-31%42%Restricted pending NVDA acqusition of MLNX.NANAMLNX43%29%Restricted pending NVDA acqusition of MLNX.NANAQCOM-11%0%Not Rated0.0%0.1%0.2%0.3%0.4%0.5%0.6%0.7%198019841988199219962000Percentage of GDPSemi RevSemi Rev(Ex-Memory)$0.25$0.35$0.45$0.55$0.65$0.75Semi ASPs(3MMA)14 April 2019 Semiconductors 3 Semi space,(2)Accelerating supply growth as Moores Law went from a 3 year cadence to a 2 year cadence and the introduction of 300 mm wafers,and(3)A shift in the demand curve from enterprise to consumer and a software ecosystem which began to innovate at a slower pace driving a“Good Enough”phenomenon.Figure 4:Foundry/Fabless Outgrew Overall Semis Figure 5:300mm Drove 50%Higher Supply Growth Rate Source:Company data,Credit Suisse estimates,SIA Source:Company data,Credit Suisse estimates,SIA As a result,the industry was given the means(cost was coming down faster than normal)and the desire(application-led growth was slowing)to cut pricing in order to try and spur demand i.e.,price elasticity.The data is conclusive.The Semiconductor Industry is NOT price elastic from 2000 thru 2008 ASPs declined on average 2%per year and Semi Rev as a%of Nominal Global GDP declined by 3%based on that measure of value,Semis are no more valuable to the Global economy today than they were in 1994/5.Figure 6:Semi ASPs Declined 40%From 1997-Present Figure 7:Semi Rev as%of GDP Declined 20%2000-2017 Source:Company data,Credit Suisse estimates,SIA Source:Company data,Credit Suisse estimates,Bloomberg Starting around the Financial Crisis thru today,we have seen a decided reversal of many of these dynamics supporting stable to rising ASPs LT and the backdrop for Semi Rev to out-grow Nominal GDP;specifically:(1)Barriers to Entry are Rising:The cost of design has been growing non-linearly driving M&A over IPOs.Over the last 8 years there have been$300bn in M&A transactions,3x more than in the prior 15 years.While foundries mitigated the capital cost for new entrants,R&D has become the new driver for scale.Sub-wavelength processing,multi-core,SoCs and Ecosystem Development have greatly increased the R&D burden.012345671995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005Indexed ValueFoundry,Fabless and Semi Rev Indexed Growth 1997-2005Foundry RevFabless RevSemi Rev0%5%10%15%05001,0001,5002,0002,500199820002002200420062008Wafer Area Growth y/yIC Capacity(WSPW x1000)300mm200mm 200mm98-04 Avg.Y/Y Growth:7.6%04-08 Avg.Y/Y Growth:12.0%$0.25$0.35$0.45$0.55$0.65$0.75Semi ASPs(3MMA)+4%CAGR-2%CAGR0.0%0.1%0.2%0.3%0.4%0.5%0.6%0.7%1981198519891993199720012005200920132017Percentage of GDPSemi Rev as%of Nominal Global GDPSemi RevSemi Rev(Ex-Memory)14 April 2019 Semiconductors 4 Figure 8:Foundry Increased from 3%to 40%of Semi Rev Figure 9:R&D is the New CapEx Source:Company data,Credit Suisse estimates,SIA Source:Company data,Credit Suisse estimates,Bloomberg(2)Supply Growth is Slowing:Moores Law has become more difficult,stretching back to a 3 year cadence,and supply growth has slowed.Wafer size transitions occur 10 years implying 450 mm R&D in 2008.Litho throughput has made 450 mm“uneconomical.”DRAM and Fabless consolidation caused supply growth to decelerate to just 2%from 2010-2017.(3)End-Demand Diversifying:With handsets fully penetrated,end demand has migrated back to non-consumer.Specifically after peaking in 2011 at 2/3 of Semi Rev,Consumer has declined to BELOW 50%of Semi Rev.As Semi Mix moves higher,ASP declines have slowed from a-4%CAGR to flat.We see application driven growth as providing the foundation for stable to rising ASPs longer-term.Figure 10:High Utilization Is Its Own Killer App Figure 11:Moderating Supply Growth Source:Company data,Credit Suisse estimates,SIA Source:Company data,Credit Suisse estimates,Bloomberg SCE Moving From Semi-Minus to Semi-Plus Growth Fab Productivity:Starting in 2000 a number of factors such as improved per tool throughput,automation,scheduling software and foundries helped to meaningfully increase fab productivity i.e.300mm productivity gains were not just a function of pi r-squared.INTC saw increased fab productivity,even against increasing Moores Law complexity,reducing cycle times by 75%.While INTC is an outlier,it is directionally correct.As a result,from 2000-2015 Semi Rev almost doubled on flat WFE and Capital Intensity declined from 30%to 20%.However,the current industry dynamics are significantly different which we believe is under-appreciated by investors as it pertains to our thesis that SCE is moving from Semi-Minus to Semi-Plus Growth.Gains from 300mm have played out,complexity is increasing and will even with EUV creating an upward bias to cycle times following 10 years of declining cycle times.Relative to 450mm,we would note that original expectations were for a 2008 introduction(i.e.we are already 10 years 0%5%10%15%20%25%30%35%40%45%$0$10$20$30$40$50$60$70$80$90$1001993 1996 1999 2002 2005 2008 2011 2014 2017%of Semi RevRev($bn)Foundry as a%of Semi Rev 5%10%15%20%25%30%$0$200$400$600$800$1,000%of RevCost of Deisng($m)Cost of Implementing Complex DesignCost($m)R&D%of RevCapEx%of RevTransistor Utilization vs.ROI on Incremental PerformanceTransistor UtlizationROI on Incremental Performance0%5%10%15%05001,0001,5002,0002,500Wafer Area Growth y/yIC Capacity(WSPW x1000)300mm200mm50%of CapEx,today they are 80%.Even memory saw consolidation DRAM declined from 8 to 3 players despite outsized growth.Importantly,concentration negatively impacted SCE GMs in the 1990s SCE GMs were regularly 50%,today they are in the mid to high 40s.However,the trend of consolidation has begun to change as the negative impact from Foundry is mostly in arrears and without 450mm we are not likely to see additional IDM dropouts.Further,the growing strategic importance of memory has brought new players to the market in particular China.Note we are currently tracking 15 projects in China valued at$60bn.The Data Economy Represents Upside to Our Secular View:In C2Q14 we introduced our Data Paradigm Model for Semis consisting of four key areas:Creation,Storage,Transmission and Analytics.We view this model as central to the evolution towards a truly“Data Driven Economy”enabled by faster than historic silicon content growth in non-traditional markets.While Creation,Storage,and Transmission have all benefited from 0%5%10%15%20%25%30%35%0.00.51.01.52.02.520002001200220032004200520062007200820092010201120122013201420152016201720182019E2020ESemi RevWFECapEx as%of RevCycle Time in Weeks01020304050607065nm40nm28nm20nm-DblPattern14nm-MultPattern14nm-EUV10nm7nm5nm#of Mask Layers05101520Cycle Time$0$250$500$750$1,000$1,250$1,500200520062007200820092010201120122013201420152016201720182019$mn/1%Bit GrowthDRAM CapEx per 1%bit growth(2-yr avg.)$-$200$400$600$80020062007200820092010201120122013201420152016201720182019$mn/1%Bit GrowthNAND Capex per 1%bit growth(2-yr Avg.)14 April 2019 Semiconductors 6 non-linear cost declines(Moores,Butters,Kryders Laws),the cost of analytics has mostly only increased.AI changes that dynamic as the first mechanism to structurally lower the cost of analytics which in turn is likely to drive innovation/new applications which could be disruptive/deflationary to traditional areas of the economy while creating significant new silicon TAM.We see Data as adding 1-300 bps to our LT Semi growth rate of 7%and could underpin a Semi market which approaches$750bn we are on record in 2015 saying$500bn by 2020.Figure 16:Our Data Growth Paradigm Source:Company data,Credit Suisse Research Once Data Analytics becomes ubiquitous,that data becomes inherently valuable,and the acquisition of data beco