Technology

Productivity growth is experiencing a cyclical rebound, but remains structurally weak. The end of the deepening of globalization, statistical hurdles, and the possibility that today’s technological advances may not be as revolutionary as past ones all hamper productivity. On the back of rising market power and concentration, companies are increasing markups instead of production. This is depressing productivity and lowering the neutral rate of interest. For now, investors can generate alpha by focusing on consolidating industries. Growing market power cannot last forever and will meet a political wall. Structurally, this will hurt asset prices.   “We don’t have a free market; don’t kid yourself. (…) Businesspeople are enemies of free markets, not friends (…) businesspeople are all in favor of freedom for everybody else (…) but when it comes to their own business, they want to go to Washington to protect their businesses.” Milton Friedman, January 1991. Despite the explosion of applications of growing computing power, U.S. productivity growth has been lacking this cycle. This incapacity to do more with less has weighed on trend growth and on the neutral rate of interest, and has been a powerful force behind the low level of yields at home and abroad. In this report, we look at the different factors and theories advanced to explain the structural decline in productivity. Among them, a steady increase in corporate market power not only goes a long way in explaining the lack of productivity in the U.S., but also the high level of profit margins along with the depressed level of investment and real neutral rates. A Simple Cyclical Explanation The decline in productivity growth is both a structural and cyclical story. Historically, productivity growth has followed economic activity. When demand is strong, businesses can generate more revenue and therefore produce more. The historical correlation between U.S. nonfarm business productivity and the ISM manufacturing index illustrates this relationship (Chart II-1). Chart II-1The Cyclical Behavior Of Productivity The Cyclical Behavior Of Productivity The Cyclical Behavior Of Productivity Chart II-2Deleveraging Hurts Productivity Deleveraging Hurts Productivity Deleveraging Hurts Productivity Since 2008, as households worked off their previous over-indebtedness, the U.S. private sector has experienced its longest deleveraging period since the Great Depression. This frugality has depressed demand and contributed to lower growth this cycle. Since productivity is measured as output generated by unit of input, weak demand growth has depressed productivity statistics. On this dimension, the brief deleveraging experience of the early 1990s is instructive: productivity picked up only after 1993, once the private sector began to accumulate debt faster than the pace of GDP growth (Chart II-2). The recent pick-up in productivity reflects these debt dynamics. Since 2009, the U.S. non-financial private sector has stopped deleveraging, removing one anchor on demand, allowing productivity to blossom. Moreover, the pick-up in capex from 2017 to present is also helping productivity by raising the capital-to-workers ratio. While this is a positive development for the U.S. economy, the decline in productivity nonetheless seems structural, as the five-year moving average of labor productivity growth remains near its early 1980s nadir (Chart II-3). Something else is at play. Chart II-3 The Usual Suspects Three major forces are often used to explain why observed productivity growth is currently in decline: A slowdown in global trade penetration, the fact that statisticians do not have a good grasp on productivity growth in a service-based economy, and innovation that simply isn’t what it used to be. Slowdown In Global Trade Penetration Two hundred years ago, David Ricardo argued that due to competitive advantages, countries should always engage in trade to increase their economic welfare. This insight has laid the foundation of the argument that exchanges between nations maximizes the utilization of resources domestically and around the world. The collapse in new business formation in the U.S. is another fascinating development. Rarely was this argument more relevant than over the past 40 years. On the heels of the supply-side revolution of the early 1980s and the fall of the Berlin Wall, globalization took off. The share of the world's population participating in the global capitalist system rose from 30% in 1985 to nearly 100% today. Generating elevated productivity gains is simpler when a country’s capital stock is underdeveloped: each unit of investment grows the capital-to-labor ratio by a greater proportion. As a result, productivity – which reflects the capital-to-worker ratio – can grow quickly. As more poor countries have joined the global economy and benefitted from FDI and other capital inflows, their productivity has flourished. Consequently, even if productivity growth has been poor in advanced economies over the past 10 years, global productivity has remained high and has tracked the share of exports in global GDP (Chart II-4). Chart II-4The Apex Of Globalization Represented The Summit Of Global Productivity Growth The Apex Of Globalization Represented The Summit Of Global Productivity Growth The Apex Of Globalization Represented The Summit Of Global Productivity Growth This globalization tailwind to global productivity growth is dissipating. First, following an investment boom where poor decisions were made, EM productivity growth has been declining. Second, with nearly 100% of the world’s labor supply already participating in the global economy, it is increasingly difficult to expand the share of global trade in global GDP and increase the benefit of cross-border specialization. Finally, the popular backlash in advanced economies against globalization could force global trade into reverse. As economic nationalism takes hold, cross-border investments could decline, moving the world economy further away from an optimal allocation of capital. These forces may explain why global productivity peaked earlier this decade. Productivity Is Mismeasured Recently deceased luminary Martin Feldstein argued that the structural decline in productivity is an illusion. As the argument goes, productivity is not weak; it is only underestimated. A parallel with the introduction of electricity in the late 19th century often comes to mind. Back then, U.S. statistical agencies found it difficult to disentangle price changes from quantity changes in the quickly growing revenues of electrical utilities. As a result, the Bureau Of Labor Statistics overestimated price changes in the early 20th century, which depressed the estimated output growth of utilities by a similar factor. Since productivity is measured as output per unit of labor, this also understated actual productivity growth – not just for utilities but for the economy as a whole. Ultimately, overall productivity growth was revised upward. Chart II-5Plenty Of Room To Mismeasure Real Output Growth Plenty Of Room To Mismeasure Real Output Growth Plenty Of Room To Mismeasure Real Output Growth In today’s economy, this could be a larger problem, as 70% of output is generated in the service sector. Estimating productivity growth is much harder in the service sector than in the manufacturing sector, as there is no actual countable output to measure. Thus, distinguishing price increases from quantity or quality improvements is challenging. Adding to this difficulty, the service sector is one of the main beneficiaries of the increase in computational power currently disrupting industries around the world. The growing share of components of the consumer price index subject to hedonic adjustments highlight this challenge (Chart II-5). Estimating quality changes is hard and may bias the increase in prices in the economy. If prices are unreliably measured, so will output and productivity. Pushing The Production Frontier Is Increasingly Hard Chart II-6A Multifaceted Decline In Productivity A Multifaceted Decline In Productivity A Multifaceted Decline In Productivity Another school of thought simply accepts that productivity growth has declined in a structural fashion. It is far from clear that the current technological revolution is much more productivity-enhancing than the introduction of electricity 140 years ago, the development of the internal combustion engine in the late 19th century, the adoption of indoor plumbing, or the discovery of penicillin in 1928. It is easy to overestimate the economic impact of new technologies. At first, like their predecessors, the microprocessor and the internet created entirely new industries. But this is not the case anymore. For all its virtues, e-commerce is only a new method of selling goods and services. Cloud computing is mainly a way to outsource hardware spending. Social media’s main economic value has been to gather more information on consumers, allowing sellers to reach potential buyers in a more targeted way. Without creating entirely new industries, spending on new technologies often ends up cannibalizing spending on older technologies. For example, while Google captures 32.4% of global ad revenues, similar revenues for the print industry have fallen by 70% since their apex in 2000. If new technologies are not as accretive to production as the introduction of previous ones were, productivity growth remains constrained by the same old economic forces of capex, human capital growth and resource utilization. And as Chart II-6 shows, labor input, the utilization of capital and multifactor productivity have all weakened. Some key drivers help understand why productivity growth has downshifted structurally. Let’s look at human capital. It is much easier to grow human capital when very few people have a high-school diploma: just make a larger share of your population finish high school, or even better, complete a university degree. But once the share of university-educated citizens has risen, building human capital further becomes increasingly difficult. Chart II-7 illustrates this problem. Growth in educational achievement has been slowing since 1995 in both advanced and developing economies. This means that the growth of human capital is slowing. This is without even wading into whether or not the quality of education has remained constant. This is pure market power, and it helps explain the gap between wages and productivity. Human capital is also negatively impacted by demographic trends. Workers in their forties tend to be at the peak of their careers, with the highest accumulated job know-how. Problematically, these workers represent a shrinking share of the labor force, which is hurting productivity trends (Chart II-8). Chart II-7 Chart II-8Demographics Are Hurting Productivity Demographics Are Hurting Productivity Demographics Are Hurting Productivity   The capital stock too is experiencing its own headwinds. While Moore’s Law seems more or less intact, the decline in the cost of storing information is clearly decelerating (Chart II-9). Today, quality adjusted IT prices are contracting at a pace of 2.3% per annum, compared to annual declines of 14% at the turn of the millennium. Thus, even if nominal spending in IT investment had remained constant, real investment growth would have sharply decelerated (Chart II-10). But since nominal spending has decelerated greatly from its late 1990s pace, real investment in IT has fallen substantially. The growth of the capital stock is therefore lagging its previous pace, which is hurting productivity growth. Chart II-9 Chart II-10The Impact Of Slowing IT Deflation The Impact Of Slowing IT Deflation The Impact Of Slowing IT Deflation Chart II-11A Dearth Of New Businesses A Dearth Of New Businesses A Dearth Of New Businesses   The collapse in new business formation in the U.S. is another fascinating development (Chart II-11). New businesses are a large source of productivity gains. Ultimately, 20% of productivity gains have come from small businesses becoming large ones. Think Apple in 1977 versus Apple today. A large decline in the pace of new business formation suggests that fewer seeds have been planted over the past 20 years to generate those enormous productivity explosions than was the case in the previous 50 years.   The X Factor: Growing Market Concentration The three aforementioned explanations for the decline in productivity are all appealing, but they generally leave investors looking for more. Why are companies investing less, especially when profit margins are near record highs? Why is inflation low? Why has the pace of new business formation collapsed? These are all somewhat paradoxical. Chart II-12Wide Profit Margins: A Testament To The Weakness Of Labor Wide Profit Margins: A Testament To The Weakness Of Labor Wide Profit Margins: A Testament To The Weakness Of Labor This is where a growing body of works comes in. Our economy is moving away from the Adam Smith idea of perfect competition. Industry concentration has progressively risen, and few companies dominate their line of business and control both their selling prices and input costs. They behave as monopolies and monopsonies, all at once.1 This helps explain why selling prices have been able to rise relative to unit labor costs, raising margins in the process (Chart II-12). Let’s start by looking at the concept of market concentration. According to Grullon, Larkin and Michaely, sales of the median publicly traded firms, expressed in constant dollars, have nearly tripled since the mid-1990s, while real GDP has only increased 70% (Chart II-13).2 The escalation in market concentration is also vividly demonstrated in Chart II-14. The top panel shows that since 1997, most U.S. industries have experienced sharp increases in their Herfindahl-Hirshman Index (HHI),3 a measure of concentration. In fact, more than half of U.S. industries have experienced concentration increases of more than 40%, and as a corollary, more than 75% of industries have seen the number of firms decline by more than 40%. The last panel of the chart also highlights that this increase in concentration has been top-heavy, with a third of industries seeing the market share of their four biggest players rise by more than 40%. Rising market concentration is therefore a broad phenomenon – not one unique to the tech sector. Chart II-13 Chart II-14     This rising market concentration has also happened on the employment front. In 1995, less than 24% of U.S. private sector employees worked for firms with 10,000 or more employees, versus nearly 28% today. This does not seem particularly dramatic. However, at the local level, the number of regions where employment is concentrated with one or two large employers has risen. Azar, Marinescu and Steinbaum developed Map II-1, which shows that 75% of non-metropolitan areas now have high or extreme levels of employment concentration.4 Chart II- Chart II-15The Owners Of Capital Are Keeping The Proceeds Of The Meagre Productivity Gains The Owners Of Capital Are Keeping The Proceeds Of The Meagre Productivity Gains The Owners Of Capital Are Keeping The Proceeds Of The Meagre Productivity Gains This growing market power of companies on employment can have a large impact on wages. Chart II-15 shows that real wages have lagged productivity since the turn of the millennium. Meanwhile, Chart II-16 plots real wages on the y-axis versus the HHI of applications (top panel) and vacancies (bottom panel). This chart shows that for any given industry, if applicants in a geographical area do not have many options where to apply – i.e. a few dominant employers provide most of the jobs in the region – real wages lag the national average. The more concentrated vacancies as well as applications are with one employer, the greater the discount to national wages in that industry.5 This is pure market power, and it helps explain the gap between wages and productivity as well as the widening gap between metropolitan and non-metropolitan household incomes. Chart II-16 Growing market power and concentration do not only compress labor costs, they also result in higher prices for consumers. This seems paradoxical in a world of low inflation. But inflation could have been even lower if market concentration had remained at pre-2000s levels. In 2009, Matthew Weinberg showed that over the previous 22 years, horizontal mergers within an industry resulted in higher prices.6 In a 2014 meta-study conducted by Weinberg along with Orley Ashenfelter and Daniel Hosken, the authors showed that across 49 studies ranging across 21 industries, 36 showed that horizontal mergers resulted in higher prices for consumers.7 While today’s technology may be enhancing the productive potential of our economies, this is not benefiting output and measured productivity. Instead, it is boosting profit margins. In a low-inflation environment, the only way for companies to garner pricing power is to decrease competition, and M&As are the quickest way to achieve this goal. After examining nearly 50 merger and antitrust studies spanning more than 3,000 merger cases, John Kwoka found that, following mergers that augmented an industry’s concentration, prices increased in 95% of cases, and on average by 4.5%.8 In no industry is this effect more vividly demonstrated than in the healthcare field, an industry that has undergone a massive wave of consolidation – from hospitals, to pharmacies to drug manufacturers. As Chart II-17 illustrates, between 1980 and 2016, healthcare costs have increased at a much faster pace in the U.S. than in the rest of the world. However, life expectancy increased much less than in other advanced economies. Chart II-17 In this context of growing market concentration, it is easy to see why, as De Loecker and Eeckhout have argued, markups have been rising steadily since the 1980s (Chart II-18, top panel) and have tracked M&A activity (Chart II-18, bottom panel).9 In essence, mergers and acquisitions have been the main tool used by firms to increase their concentration. Another tool at their disposal has been the increase in patents. The top panel of Chart II-19 shows that the total number of patent applications in the U.S. has increased by 3.6-fold since the 1980s, but most interestingly, the share of patents coming from large, dominant players within each industry has risen by 10% over the same timeframe (Chart II-19, bottom panel). To use Warren Buffet’s terminology, M&A and patents have been how firms build large “moats” to limit competition and protect their businesses. Chart II-18Markups Rise Along With Growing M&A Activity Markups Rise Along With Growing M&A Activity Markups Rise Along With Growing M&A Activity Chart II-19How To Build A Moat? How To Build A Moat? How To Build A Moat?   Why is this rise in market concentration affecting productivity? First, from an empirical perspective, rising markups and concentration tend to lead to lower levels of capex. A recent IMF study shows that the more concentrated industries become, the higher the corporate savings rate goes (Chart II-20, top panel).10 These elevated savings reflect wider markups, but also firms with markups in the top decile of the distribution display significantly lower investment rates (Chart II-20, bottom panel). If more of the U.S. output is generated by larger, more concentrated firms, this leads to a lower pace of increase in the capital stock, which hurts productivity. Second, downward pressure on real wages is also linked to a drag on productivity. Monopolies and oligopolies are not incentivized to maximize output. In fact, for any market, a monopoly should lead to lower production than perfect competition would. Diagram II-I from De Loecker and Eeckhout shows that moving from perfect competition to a monopoly results in a steeper labor demand curve as the monopolist produces less. As a result, real wages move downward and the labor participation force declines. Does this sound familiar? Chart II-20 Chart II-   The rise of market power might mean that in some way Martin Feldstein was right about productivity being mismeasured – just not the way he anticipated. In a June 2017 Bank Credit Analyst Special Report, Peter Berezin showed that labor-saving technologies like AI and robotics, which are increasingly being deployed today, could lead to lower wages (Chart II-21).11 For a given level of technology in the economy, productivity is positively linked to real wages but inversely linked to markups – especially if the technology is of the labor-saving kind. So, if markups rise on the back of firms’ growing market power, the ensuing labor savings will not be used to increase actual input. Rather, corporate savings will rise. Thus, while today’s technology may be enhancing the productive potential of our economies, this is not benefiting output and measured productivity. Instead, it is boosting profit margins.12 Unsurprisingly, return on assets and market concentration are positively correlated (Chart II-22). Chart II-21 Chart II-22     Finally, market power and concentration weighing on capex, wages and productivity are fully consistent with higher returns of cash to shareholders and lower interest rates. The higher profits and lower capex liberate cash flows available to be redistributed to shareholders. Moreover, lower capex also depresses demand for savings in the economy, while weak wages depress middle-class incomes, which hurts aggregate demand. Additionally, higher corporate savings increases the wealth of the richest households, who have a high marginal propensity to save. This results in higher savings for the economy. With a greater supply of savings and lower demand for those savings, the neutral rate of interest has been depressed. Investment Implications First, in an environment of low inflation, investors should continue to favor businesses that can generate higher markups via pricing power. Equity investors should therefore continue to prefer industries where horizontal mergers are still increasing market concentration. Second, so long as the status quo continues, wages will have a natural cap, and so will the neutral rate of interest. This does not mean that wage growth cannot increase further on a cyclical basis, but it means that wages are unlikely to blossom as they did in the late 1960s, even within a very tight labor market. Without too-severe an inflation push from wages, the business cycle could remain intact even longer, keeping a window open for risk assets to rise further on a cyclical basis. Third, long-term investors need to keep a keen eye on the political sphere. A much more laissez-faire approach to regulation, a push toward self-regulation, and a much laxer enforcement of antitrust laws and merger rules were behind the rise in market power and concentration.13 The particularly sharp ascent of populism in Anglo-Saxon economies, where market power increased by the greatest extent, is not surprising. So far, populists have not blamed the corporate sector, but if the recent antitrust noise toward the Silicon Valley behemoths is any indication, the clock is ticking. On a structural basis, this could be very negative for asset prices. An end to this rise in market power would force profit margins to mean-revert toward their long-term trend, which is 4.7 percentage-points below current levels. This will require discounting much lower cash flows in the future. Additionally, by raising wages and capex, more competition would increase aggregate demand and lift real interest rates. Higher wages and aggregate demand could also structurally lift inflation. Thus, not only will investors need to discount lower cash flows, they will have to do so at higher discount rates. As a result, this cycle will likely witness both a generational peak in equity valuations as well as structural lows in bond yields. As we mentioned, these changes are political in nature. We will look forward to studying the political angle of this thesis to get a better handle on when these turning points will likely emerge. Mathieu Savary Vice President The Bank Credit Analyst   Footnotes 1       A monopsony is a firm that controls the price of its input because it is the dominant, if not unique, buyer of said input. 2       G. Grullon, Y. Larkin and R. Michaely, “Are Us Industries Becoming More Concentrated?,” April 2017. 3       The Herfindahl-Hirschman Index (HHI) is calculated by taking the market share of each firm in the industry, squaring them, and summing the result. Consider a hypothetical industry with four total firm where firm1, firm2, firm3 and firm4 has 40%, 30%, 15% and 15% of market share, respectively. Then HHI is 402+302+152+152 = 2,950. 4       J. Azar, I. Marinescu, M. Steinbaum, “Labor Market Concentration,” December 2017. 5     J. Azar, I. Marinescu, M. Steinbaum, “Labor Market Concentration,” December 2017. 6     M. Weinberg, “The Price Effects Of Horizontal Mergers”, Journal of Competition Law & Economics, Volume 4, Issue 2, June 2008, Pages 433–447. 7     O. Ashenfelter, D. Hosken, M. Weinberg, "Did Robert Bork Understate the Competitive Impact of Mergers? Evidence from Consummated Mergers," Journal of Law and Economics, University of Chicago Press, vol. 57(S3), pages S67 - S100. 8    J. Kwoka, “Mergers, Merger Control, and Remedies: A Retrospective Analysis of U.S. Policy,” MIT Press, 2015. 9     J. De Loecker, J. Eeckhout, G. Unger, "The Rise Of Market Power And The Macroeconomic Implications," Mimeo 2018. 10     “Chapter 2: The Rise of Corporate Market Power and Its Macroeconomic Effects,” World Economic Outlook, April 2019. 11     Please see The Bank Credit Analyst Special Report "Is Slow Productivity Growth Good Or Bad For Bonds?"dated May 31, 2017, available at bca.bcaresearch.com. 12     Productivity can be written as: Image 13     J. Tepper, D. Hearn, “The Myth of Capitalism: Monopolies and the Death of Competition,” Wiley, November 2018.

Stick With Resilient Software Stocks Stick With Resilient Software Stocks ​​​​​​​Overweight, High-Conviction On June 10th we tightened our stops on the overweight call in the S&P software index, as a risk management measure in the context of our cautious broad equity market stance. Our bullish software thesis has not changed, and we reiterate that the only way to monetize gains in these highflying stocks is via tightening stops. Yesterday’s ultra-dovish Fed meeting boosted the appeal of high growth stocks, including software, as the Fed is seriously considering a cut in the late-July meeting. Moreover, software investment is the last pillar keeping overall U.S. capital outlays in positive territory. Not only is software investment rising, but it is also garnering a larger slice of the overall capex pie (middle & bottom panels). Another source of support is that software is a service-based industry and, at the margin, mostly insulated from the U.S./China trade dispute, so investors have been finding refuge in these equities. Adobe’s and Oracle’s recent healthy earnings reports and upbeat guidance confirm that software profits will remain upbeat and will likely continue to outpace the broad market (bottom panel). Bottom Line: We remain cyclically overweight the S&P software index (it is also a high-conviction overweight), but we will obey our stops in case a riot point materializes in the broad equity market. The ticker symbols for the stocks in this index are: BLBG: S5SOFT – MSFT, ORCL, ADBE, CRM, INTU, ADSK, RHT, CDNS, SNPS, ANSS, SYMC, CTXS, FTNT.  

(Part II) Global Semiconductors: The Key Driver (Part II) Global Semiconductors: The Key Driver Global smartphone sales, which drive 29% of global semiconductor revenues, are currently contracting. According to the International Data Corporation (IDC), in Q1/2019 global smartphone shipments declined 6.6% year-on-year (yoy) in volume terms. The slowdown is also picking up pace, as last year’s drop was 4.4% (see chart). We also expect smartphone shipments to continue contracting in the second half of this year. Major markets such as mainland China and advanced economies have entered the saturation phase of mobile-phone demand. For example, U.S. shipments were down 15% yoy in Q1 due to near-full market penetration. Investors are also mistakenly betting on 5G technology. Although Samsung, Huawei, OnePlus, Xiaomi, Motorola, LG, and ZTE have either released or will release their 5G phones this year, the sales growth from 5G phones will not be able to offset the loss in 2G, 3G and 4G phone sales, at least not in 2019. IDC estimated that 5G phones would only reach 0.5% of the global mobile-phone market share this year. 5G phones will likely only begin boosting overall semiconductor demand next year, when they will garner a larger slice of the global smartphone market. Bottom Line: Global semiconductor stocks are still facing considerable downside. Our Emerging Markets Strategy service remains negative on Asian semiconductor share prices in absolute terms. A continued contraction in global semiconductor sales will further squeeze their profits. For additional details, please see this past Monday’s Special Report authored by Ellen JingYuan He, Associate Vice President of Emerging Markets Strategy. ​​​​​​​​​​​​​​

(Part I) Global Semiconductors: Demand (Part I) Global Semiconductors: Demand The odds of a cyclical upturn in global semiconductors over the next three-to-six months are low as global demand remains feeble and is contracting 15%/annum (top panel). Drilling deeper into global demand reveals that the slowdown is structural, affecting a number of geographical areas (bottom five panels). Chip company revenues have so far contracted by 24% since the October 2018 peak, which is disproportionally more than the decline in share prices. The global semiconductor equity index is only 14% below its March 2018 high. It appears as though the market is expecting a quick recovery in semi sales. However, as we highlighted in our most recent Special Report authored by Ellen JingYuan He, Associate Vice President of Emerging Markets Strategy, there are structural problems in each of the key segments that drive global semiconductor sales, warning that the odds of an upturn are low. Please see the next Insight where we discuss the major demand driver that accounts for 29% of the world’s total semiconductor sales. ​​​​​​​

Highlights The odds of a cyclical upturn in the global semiconductor sector over the next three to six months are low. Global semiconductor demand will continue to decline due to contracting demand for smartphones, automobiles, personal computers (PCs), and servers. Global semiconductor stocks are still facing considerable downside in absolute terms. We recommend going long Asian semiconductor stocks versus the U.S. S&P 500 semiconductor index. Dedicated EM equity portfolios should stay neutral on the Taiwanese bourse and Korean technology sector relative to the overall EM benchmark. Feature Chart 1 shows share prices of the global semiconductor sector and global semiconductor sales in the past two decades. Chart 1Global Semiconductor Market: Sales & Share Prices Global Semiconductor Market: Sales & Share Prices Global Semiconductor Market: Sales & Share Prices Was last December’s trough in global semiconductor equity prices the ultimate bottom in this cycle? The odds are in favor of a continued contraction in global semiconductor sales and further downside in semiconductor share prices over the next three to six months. Cycle-On-Cycle Analysis Semiconductor sales experienced five recessions over the past 20 years. Table 1 illustrates the peak-to-bottom percentage decline in nominal global semiconductor sales and the magnitude of the drop in global semiconductor share prices in U.S. dollar terms during these five cyclical downturns in this industry. It also indicates the duration of each downturn and the number of months that semiconductor stocks led the bottom in global semiconductor sales. Table 1Key Statistics Of Five Cyclical Downturns In Global Semiconductor Market The Global Semiconductor Sector: Is A Cyclical Upturn Imminent? The Global Semiconductor Sector: Is A Cyclical Upturn Imminent? The current shrinkage of semiconductor sales is worse than the 2011-12 and 2015 downturns. Yet, it is still smaller than the magnitude during the 2008 Great Financial Crisis and the 2001 tech bubble bust. The revenue of semiconductor companies has so far contracted by 24%, which is disproportionally more than the decline in share prices of these companies. The global semiconductor equity index is only 14% below its March 2018 high. It appears as though the market is expecting a quick recovery in semiconductor sales. As per Table 1, in the downturns of 2008, 2011 and 2015, global semiconductor stocks all bottomed before the bottom of global semiconductor sales. Only in the 2001 episode, stock prices bottomed eight months after the bottom in sales. In the current cyclical downturn, global semiconductor sales have so far had only four months of growth contraction,1 far less than the 13-16 months experienced in all the past four cycles. All in all, we would lean against the market’s expectation of an imminent recovery in the semiconductor cycle. The demand downturn will last another three to six months and share prices are facing major headwinds. Global Semiconductor Demand Semiconductor sales are in contraction across countries and regions (Chart 2). In April – before President Trump’s tweet on imposing new import tariff on China, global semiconductor sales growth sank to a negative 15% year-on-year. The short-term (three-to-six month) outlook for global semiconductor demand remains dismal. Chart 3 shows global semiconductor revenue breakdown in terms of end usage. Mobile phones account for the largest share (29%) of the market, followed by PCs (12%), miscellaneous consumer products (12%), and servers (11%). All of these major demand sources are under downward pressure: Smartphone Sales Global smartphone sales are shrinking (Chart 4). According to the International Data Corporation (IDC), global smartphone shipments declined 6.6% year-on-year in volume terms in the first quarter of this year, worse than last year’s 4.4% drop. Chart 2Semiconductor Sales Are In Contraction Across Countries Semiconductor Sales Are In Contraction Across Countries Semiconductor Sales Are In Contraction Across Countries  In the current cyclical downturn, global semiconductor sales have so far had only four months of growth contraction, far less than the 13-16 months experienced in all the past four cycles. Chart 3 Chart 4Global Smartphone Sales: Contracting Global Smartphone Sales: Contracting Global Smartphone Sales: Contracting We expect smartphone shipments to continue contracting in the second half of this year. Major markets such as mainland China and advanced economies have entered the saturation phase of mobile-phone demand. For example, U.S. shipments were down 15% year-on-year in the first quarter due to near-full market penetration. In China, smartphone sales have shown signs of stabilization (Chart 5). However, this is probably temporary and has been driven by the boom in Huawei smartphone sales in China since early this year. The incredible 50% year-on-year growth of Huawei smartphone sales in the first quarter is not sustainable. While global sales of Huawei smartphones increased by 20 million units, total global smartphone sales of all brands fell by 22 million units (Chart 6). The U.S. punitive actions towards China and Huawei have also instigated nationalism in China. This has triggered a Chinese buying-spree of the Huawei smartphone. Chart 5Chinese Smartphone Sales: Temporary Stabilization Chinese Smartphone Sales: Temporary Stabilization Chinese Smartphone Sales: Temporary Stabilization Chart 6 Yet, this has probably reduced the number of potential Chinese smartphones buyers in the near future. After all, many buyers likely made the purchase earlier than otherwise planned in the absence of a trade war. Although Samsung, Huawei, OnePlus, Xiaomi, Motorola, LG, and ZTE have either released or will release their 5G phones this year, the sales growth from 5G phones will not be able to offset the loss in 2G, 3G and 4G phone sales, at least not in 2019. The IDC estimated that 5G phones would only account for about 0.5% of the market share this year. 5G will likely only begin affecting overall semiconductor demand next year, when they account for a larger share of smartphone sales.   Huawei is the market leader in 5G technology. The U.S. boycott of Huawei will likely continue. This will only slow the pace of 5G phone adoption and the development of 5G networks worldwide. On balance, global smartphone demand may only recover next year. Server Demand Global server shipments also experienced a 5% contraction in volume terms in the first quarter of this year, according to IDC (Chart 7). The outlook for the rest of 2019 does not look promising. Global server demand will likely remain in contraction in the second half of this year. Many hyperscale data centers have already purchased considerable amounts of servers in advance of the trade war to avoid tariffs.2 Meanwhile, the escalation in the U.S.-China confrontation has increased economic uncertainties. This may delay potential datacenter investments. Decelerating 5G network development worldwide due to the U.S. ban on Huawei will also tend to discourage new datacenter and cloud services projects. This is because the 5G technology enables datacenter and cloud services to experience a huge improvement in terms of data transfer speeds, latency, connectivity, capacity, reliability and mobility. Chart 7Global Server Shipment: Are In Contraction Global Server Shipment: Are In Contraction Global Server Shipment: Are In Contraction Personal Computers (PC) PCs sales are also in contraction (Chart 8). PC demand has entered into the deep-maturation phase while facing strong competition from tablets and laptops. Auto Sales Global auto sales also sank by 5% in April from a year ago, registering the worst contraction since 2009 (Chart 9). Chart 8Global PCs Sales: Deeply Saturated Global PCs Sales: Deeply Saturated Global PCs Sales: Deeply Saturated Chart 9Global Auto Sales Are In Contraction As Well Global Auto Sales Are In Contraction As Well Global Auto Sales Are In Contraction As Well Regarding auto demand, the Chinese government may continue to implement more supportive policies to stimulate car sales in China. However, we believe the recovery will be delayed. The government has already implemented a number of policies to lift domestic car sales since late January, including providing subsidies to encourage new energy vehicle sales, to promote auto sales in rural areas, and to increase auto replacement. The central government recently loosened auto sales restrictions in the first tier cities of Guangzhou and Shenzhen that have restrictive auto sales policies. However, all of these policies have failed to lift Chinese domestic car sales out of deep contraction. The key reason has been a diminishing willingness to spend among Chinese consumers, as suggested by falling households’ marginal propensity to consume (Chart 10). Bottom Line: Global semiconductor demand growth will likely remain weak and will fail to recover in the second half of this year. The basis is that its major upstream markets (smartphone, servers, PCs and automobiles) are all facing cyclically declining demand. Chart 10Chinese Consumers: Diminishing Willingness To Consume Chinese Consumers: Diminishing Willingness To Consume Chinese Consumers: Diminishing Willingness To Consume Inventories And Prices Chart 11 shows the semiconductor supply chain illustrating the process of manufacturing semiconductors starting with silicon wafers and up to final electronic products. Chart 11 Box 1 explains the role of key segments and players along the supply chain. Box 1 A Brief Explanation Of The Key Segments/Players Of The Supply Chain Both integrated device manufacturers (IDM) and foundries are at the center of the supply chain, responsible for chip manufacturing. In terms of semiconductor sales revenue, Samsung, Intel and SK Hynix are the world’s top three IDM companies and TSMC, Global Foundries and United Microelectronics Corp (UMC) are the world’s top three foundries. While IDMs cover most of the process from IC design, chip fabrication, assembly, testing and packaging, IDM companies still have to purchase raw materials and equipment for the chip-making process. Foundry companies receive orders from IC designing companies like Qualcomm, Nvidia, and Huawei HiSilicon, then purchase needed raw materials and equipment to proceed in the chip-manufacturing process. Both IDMs and foundries can either outsource the tasks of semiconductor assembly and testing or perform them on their own. The final semiconductor products will be used in electronics products, such as smartphones, computers, home appliances, automobiles, etc. Global semiconductor demand growth will likely remain weak and will fail to recover in the second half of this year. The basis is that its major upstream markets (smartphone, servers, PCs and automobiles) are all facing cyclically declining demand. In a typical business cycle, a cyclical downturn begins with a slump in demand for final electronic products (upstream demand). This leads to falling semiconductor sales. As a result, inventory buildup will occur across most of the segments along the semiconductor supply chain. Chipmakers: Producers’ semiconductor inventory in both Taiwan and Korea have reached either a record high or a near-record high (Chart 12). The installed wafer capacities at these two countries are the world’s largest, together accounting for 43% of total global wafer capacity. In addition, the inventory of some major electronic parts and components have also increased considerably in Taiwan (Chart 13). This also implies weaker demand for semiconductor raw materials. Chart 12Chipmakers: A Rapid Buildup In Inventory Chipmakers: A Rapid Buildup In Inventory Chipmakers: A Rapid Buildup In Inventory Chart 13Rising Inventory Of Some Major Electronic Parts And Components Rising Inventory Of Some Major Electronic Parts And Components Rising Inventory Of Some Major Electronic Parts And Components Raw material suppliers: Silicon wafer is the indispensable raw material required in the chip manufacturing process. Japanese companies account for over half of global silicon wafer supply.Chart 14 shows that silicon wafer inventory in Japan has had a significant buildup in volume terms since late last year. Importantly, it is not declining yet. Chart 14Silicon Wafer Inventory: A Significant Buildup As Well Silicon Wafer Inventory: A Significant Buildup As Well Silicon Wafer Inventory: A Significant Buildup As Well Outsourced semiconductor assembly and test (OSAT) providers: Both Singapore and Thailand are OSAT providers while they also manufacture, assemble and export electronic products. Both countries are closer to the downstream side of the semiconductor supply chain. Semiconductor inventory at these two countries has also jumped to a record high (Chart 15).   Chart 15Singapore and Thailand: Record High Semiconductor inventory Singapore and Thailand: Record High Semiconductor inventory Singapore and Thailand: Record High Semiconductor inventory   Importantly, a marginal improvement in demand will tend to support spot prices. For example, in the memory chip market, falling prices denote weak demand relative to excess supply. When prices of DRAM and NAND start to form a bottom or decisively move up, this may indicate the arrival of a cyclical upturn. So far, both DRAM and NAND prices are continuing to fall (Chart 16). In addition, the prices of silicon wafer – the most important raw material used in the chip-making process – have declined in the first half of this year.3  Chart 16Still Falling Memory Chip Prices Still Falling Memory Chip Prices Still Falling Memory Chip Prices Chart 17Deflating DRAM Prices Suggest Downside Risks To Korean Tech Stocks Deflating DRAM Prices Suggest Downside Risks To Korean Tech Stocks Deflating DRAM Prices Suggest Downside Risks To Korean Tech Stocks In short, prices are the most important variable to monitor. Chart 17 exhibits the high correlation between DRAM prices and the Korean technology sector stock prices. Falling DRAM prices suggest downside risks to technology stocks in Korea. Samsung accounts for about 65% of Korea’s tech index and 27% of the overall Korean equity index. Memory chips accounted for 68% of Samsung’s operating profits in the first quarter of this year.   Bottom Line: There has been involuntary inventory accumulation along the entire supply chain of semiconductors. This and ongoing price deflation among various types of semiconductors foreshadow a downbeat near-term outlook. The Interpretation Of Some Positive Developments There have been some positive developments in the past several months. Taiwanese PMI new orders diffusion index in the electronics sector jumped out of deep contraction to reach zero, and Chinese semiconductor imports halted their decline in both volume and value terms (Chart 18). The improvement in the aforementioned data was probably mainly due to large semiconductor purchases by China to hedge the rising risk of U.S. blocking China’s technological development (Chart 19). Chart 18Some Positive Development Some Positive Development Some Positive Development Chart 19China: More Semiconductors Purchases Before The Tariff And U.S. Huawei Ban? China: More Semiconductors Purchases Before The Tariff And U.S. Huawei Ban? China: More Semiconductors Purchases Before The Tariff And U.S. Huawei Ban? Besides, Huawei smartphone sales have been booming, which we deliberated on page 5, could have been responsible for the improvement in these data. This one-off surge will likely dwindle going forward. Investment Conclusions We remain negative on Asian semiconductor share prices in absolute terms. A continued contraction in global semiconductor sales will further squeeze their profits. In relative terms, we are neutral on the Asian semiconductor sector: we continue recommending market-weight allocation to Taiwan’s overall market and the Korean technology sector within the EM equity benchmark. As a new trade, we recommend going long Asian semiconductor stocks and short the S&P 500 semiconductor index over the next three to six months (Chart 20). The Bloomberg Asia Pacific semiconductor index has nine stocks. Samsung and TSMC account for 42% and 38% of the index, respectively. There has been involuntary inventory accumulation along the entire supply chain of semiconductors. This and ongoing price deflation among various types of semiconductors foreshadow the downbeat near-term outlook. Samsung will likely benefit from the U.S. ban on Huawei in the smartphone sector outside of China. In addition, Samsung will win some market share from Apple as the latter does not have a 5G phone to release this year. These positive factors may partially offset the negative impact from falling memory prices and demand on Samsung. The S&P 500 semiconductor index has 13 stocks. Intel, Broadcom, Texas Instruments and Qualcomm are the top five constituents, together accounting for nearly 70% of the index. Most of these companies are IC designing companies, which will likely suffer as Chinese demand for their products shrink due to the U.S. administration's ban on Huawei. This position will also benefit from U.S. dollar appreciation. A firm dollar will hurt profits of U.S. semiconductor stocks. In turn, currency depreciation in Korea and Taiwan will on the margin benefit Asian semiconductor stocks. Chart 20Recommend Long Asia Pacific Semiconductor Stock Vs. S&P 500 Semiconductor Index Recommend Long Asia Pacific Semiconductor Stock Vs. S&P 500 Semiconductor Index Recommend Long Asia Pacific Semiconductor Stock Vs. S&P 500 Semiconductor Index Chart 21The 2015 Experience The 2015 Experience The 2015 Experience Chart 21 shows that global foundry companies outperformed global IC designing companies during the final phase of the 2015 cyclical downturn. Odds are that these dynamics will play out in this downturn as well. Finally, the relative performance of Asian semiconductor stocks versus U.S. ones is oversold and might stage some sort of mean reversion (Chart 20). Ellen JingYuan He, Associate Vice President ellenj@bcaresearch.com Footnotes 1      Please note that here the calculation for “the number of months of the growth contraction” is different from the one for the “peak-to-bottom duration” in Table 1. “The number of months of the growth contraction” equals the number of months when semiconductor sales year-on-year growth is negative. 2      https://marketrealist.com/2019/05/nvidias-data-center-revenue-inference-rendering-and-edge/ 3      http://www.sohu.com/a/300386061_132567, http://news.moore.ren/industry/104958.htm

Downgrade Tech Hardware Storage & Peripherals To Neutral Downgrade Tech Hardware Storage & Peripherals To Neutral Neutral In the context of further de-risking the portfolio, we downgraded the S&P tech hardware storage & peripherals index (THS&P) to neutral in our most recent Weekly Report. Four reasons underpin our downgrade of this index that comprises almost 1/5 of the S&P tech market cap. First, index heavyweight Apple has 20% foreign sales exposure to the Greater China region. While we doubt the Chinese will directly retaliate to the U.S. restriction on Huawei by directly targeting Apple, it is still a risk. Moreover, recent news of the FTC and the DOJ targeting GOOGL and FB pose a risk to Apple, especially given its App Store dominance. Any negative news on either front would take a bite out of the sector’s profits. Second, the S&P THS&P index’s internationally sourced revenues are near the 60% mark, and computer exports are also flirting with the zero line. Worryingly, deflating EM Asian currencies are sapping consumer purchasing power and are weighing on industry exports (bottom panel). For the other two reasons that compelled us to downgrade the S&P THS&P index, please refer to our most recent Weekly Report. Bottom Line: Downgrade the S&P THS&P index to neutral for a modest relative loss of 1.0% since inception. The ticker symbols for the stocks in this index are: BLBG: S5CMPE – AAPL, HPQ, HPE, NTAP, STX, WDC, XRX.  

Highlights The odds of a cyclical upturn in the global semiconductor sector over the next three to six months are low. Global semiconductor demand will continue to decline due to contracting demand for smartphones, automobiles, personal computers (PCs), and servers. Global semiconductor stocks are still facing considerable downside in absolute terms. We recommend going long Asian semiconductor stocks versus the U.S. S&P 500 semiconductor index. Dedicated EM equity portfolios should stay neutral on the Taiwanese bourse and Korean technology sector relative to the overall EM benchmark. Feature Chart 1 shows share prices of the global semiconductor sector and global semiconductor sales in the past two decades. Chart 1Global Semiconductor Market: Sales & Share Prices Global Semiconductor Market: Sales & Share Prices Global Semiconductor Market: Sales & Share Prices Was last December’s trough in global semiconductor equity prices the ultimate bottom in this cycle? The odds are in favor of a continued contraction in global semiconductor sales and further downside in semiconductor share prices over the next three to six months. Cycle-On-Cycle Analysis Semiconductor sales experienced five recessions over the past 20 years. Table 1 illustrates the peak-to-bottom percentage decline in nominal global semiconductor sales and the magnitude of the drop in global semiconductor share prices in U.S. dollar terms during these five cyclical downturns in this industry. It also indicates the duration of each downturn and the number of months that semiconductor stocks led the bottom in global semiconductor sales. Table 1Key Statistics Of Five Cyclical Downturns In Global Semiconductor Market The Global Semiconductor Sector: Is A Cyclical Upturn Imminent? The Global Semiconductor Sector: Is A Cyclical Upturn Imminent? The current shrinkage of semiconductor sales is worse than the 2011-12 and 2015 downturns. Yet, it is still smaller than the magnitude during the 2008 Great Financial Crisis and the 2001 tech bubble bust. The revenue of semiconductor companies has so far contracted by 24%, which is disproportionally more than the decline in share prices of these companies. The global semiconductor equity index is only 14% below its March 2018 high. It appears as though the market is expecting a quick recovery in semiconductor sales. As per Table 1, in the downturns of 2008, 2011 and 2015, global semiconductor stocks all bottomed before the bottom of global semiconductor sales. Only in the 2001 episode, stock prices bottomed eight months after the bottom in sales. In the current cyclical downturn, global semiconductor sales have so far had only four months of growth contraction,1 far less than the 13-16 months experienced in all the past four cycles. All in all, we would lean against the market’s expectation of an imminent recovery in the semiconductor cycle. The demand downturn will last another three to six months and share prices are facing major headwinds. Global Semiconductor Demand Semiconductor sales are in contraction across countries and regions (Chart 2). In April – before President Trump’s tweet on imposing new import tariff on China, global semiconductor sales growth sank to a negative 15% year-on-year. The short-term (three-to-six month) outlook for global semiconductor demand remains dismal. Chart 3 shows global semiconductor revenue breakdown in terms of end usage. Mobile phones account for the largest share (29%) of the market, followed by PCs (12%), miscellaneous consumer products (12%), and servers (11%). All of these major demand sources are under downward pressure: Smartphone Sales Global smartphone sales are shrinking (Chart 4). According to the International Data Corporation (IDC), global smartphone shipments declined 6.6% year-on-year in volume terms in the first quarter of this year, worse than last year’s 4.4% drop. Chart 2Semiconductor Sales Are In Contraction Across Countries Semiconductor Sales Are In Contraction Across Countries Semiconductor Sales Are In Contraction Across Countries  In the current cyclical downturn, global semiconductor sales have so far had only four months of growth contraction, far less than the 13-16 months experienced in all the past four cycles. Chart 3 Chart 4Global Smartphone Sales: Contracting Global Smartphone Sales: Contracting Global Smartphone Sales: Contracting We expect smartphone shipments to continue contracting in the second half of this year. Major markets such as mainland China and advanced economies have entered the saturation phase of mobile-phone demand. For example, U.S. shipments were down 15% year-on-year in the first quarter due to near-full market penetration. In China, smartphone sales have shown signs of stabilization (Chart 5). However, this is probably temporary and has been driven by the boom in Huawei smartphone sales in China since early this year. The incredible 50% year-on-year growth of Huawei smartphone sales in the first quarter is not sustainable. While global sales of Huawei smartphones increased by 20 million units, total global smartphone sales of all brands fell by 22 million units (Chart 6). The U.S. punitive actions towards China and Huawei have also instigated nationalism in China. This has triggered a Chinese buying-spree of the Huawei smartphone. Chart 5Chinese Smartphone Sales: Temporary Stabilization Chinese Smartphone Sales: Temporary Stabilization Chinese Smartphone Sales: Temporary Stabilization Chart 6 Yet, this has probably reduced the number of potential Chinese smartphones buyers in the near future. After all, many buyers likely made the purchase earlier than otherwise planned in the absence of a trade war. Although Samsung, Huawei, OnePlus, Xiaomi, Motorola, LG, and ZTE have either released or will release their 5G phones this year, the sales growth from 5G phones will not be able to offset the loss in 2G, 3G and 4G phone sales, at least not in 2019. The IDC estimated that 5G phones would only account for about 0.5% of the market share this year. 5G will likely only begin affecting overall semiconductor demand next year, when they account for a larger share of smartphone sales.   Huawei is the market leader in 5G technology. The U.S. boycott of Huawei will likely continue. This will only slow the pace of 5G phone adoption and the development of 5G networks worldwide. On balance, global smartphone demand may only recover next year. Server Demand Global server shipments also experienced a 5% contraction in volume terms in the first quarter of this year, according to IDC (Chart 7). The outlook for the rest of 2019 does not look promising. Global server demand will likely remain in contraction in the second half of this year. Many hyperscale data centers have already purchased considerable amounts of severs in advance of the trade war to avoid tariffs.2 Meanwhile, the escalation in the U.S.-China confrontation has increased economic uncertainties. This may delay potential datacenter investments. Decelerating 5G network development worldwide due to the U.S. ban on Huawei will also tend to discourage new datacenter and cloud services projects. This is because the 5G technology enables datacenter and cloud services to experience a huge improvement in terms of data transfer speeds, latency, connectivity, capacity, reliability and mobility. Chart 7Global Server Shipment: Are In Contraction Global Server Shipment: Are In Contraction Global Server Shipment: Are In Contraction Personal Computers (PC) PCs sales are also in contraction (Chart 8). PC demand has entered into the deep-maturation phase while facing strong competition from tablets and laptops. Auto Sales Global auto sales also sank by 5% in April from a year ago, registering the worst contraction since 2009 (Chart 9). Chart 8Global PCs Sales: Deeply Saturated Global PCs Sales: Deeply Saturated Global PCs Sales: Deeply Saturated Chart 9Global Auto Sales Are In Contraction As Well Global Auto Sales Are In Contraction As Well Global Auto Sales Are In Contraction As Well Regarding auto demand, the Chinese government may continue to implement more supportive policies to stimulate car sales in China. However, we believe the recovery will be delayed. The government has already implemented a number of policies to lift domestic car sales since late January, including providing subsidies to encourage new energy vehicle sales, to promote auto sales in rural areas, and to increase auto replacement. The central government recently loosened auto sales restrictions in the first tier cities of Guangzhou and Shenzhen that have restrictive auto sales policies. However, all of these policies have failed to lift Chinese domestic car sales out of deep contraction. The key reason has been a diminishing willingness to spend among Chinese consumers, as suggested by falling households’ marginal propensity to consume (Chart 10). Bottom Line: Global semiconductor demand growth will likely remain weak and will fail to recover in the second half of this year. The basis is that its major upstream markets (smartphone, servers, PCs and automobiles) are all facing cyclically declining demand. Chart 10Chinese Consumers: Diminishing Willingness To Consume Chinese Consumers: Diminishing Willingness To Consume Chinese Consumers: Diminishing Willingness To Consume Inventories And Prices Chart 11 shows the semiconductor supply chain illustrating the process of manufacturing semiconductors starting with silicon wafers and up to final electronic products. Chart 11 Box 1 explains the role of key segments and players along the supply chain. Box 1 A Brief Explanation Of The Key Segments/Players Of The Supply Chain Both integrated device manufacturers (IDM) and foundries are at the center of the supply chain, responsible for chip manufacturing. In terms of semiconductor sales revenue, Samsung, Intel and SK Hynix are the world’s top three IDM companies and TSMC, Global Foundries and United Microelectronics Corp (UMC) are the world’s top three foundries. While IDMs cover most of the process from IC design, chip fabrication, assembly, testing and packaging, IDM companies still have to purchase raw materials and equipment for the chip-making process. Foundry companies receive orders from IC designing companies like Qualcomm, Nvidia, and Huawei HiSilicon, then purchase needed raw materials and equipment to proceed in the chip-manufacturing process. Both IDMs and foundries can either outsource the tasks of semiconductor assembly and testing or perform them on their own. The final semiconductor products will be used in electronics products, such as smartphones, computers, home appliances, automobiles, etc. Global semiconductor demand growth will likely remain weak and will fail to recover in the second half of this year. The basis is that its major upstream markets (smartphone, servers, PCs and automobiles) are all facing cyclically declining demand. In a typical business cycle, a cyclical downturn begins with a slump in demand for final electronic products (upstream demand). This leads to falling semiconductor sales. As a result, inventory buildup will occur across most of the segments along the semiconductor supply chain. Chipmakers: Producers’ semiconductor inventory in both Taiwan and Korea have reached either a record high or a near-record high (Chart 12). The installed wafer capacities at these two countries are the world’s largest, together accounting for 43% of total global wafer capacity. In addition, the inventory of some major electronic parts and components have also increased considerably in Taiwan (Chart 13). This also implies weaker demand for semiconductor raw materials. Chart 12Chipmakers: A Rapid Buildup In Inventory Chipmakers: A Rapid Buildup In Inventory Chipmakers: A Rapid Buildup In Inventory Chart 13Rising Inventory Of Some Major Electronic Parts And Components Rising Inventory Of Some Major Electronic Parts And Components Rising Inventory Of Some Major Electronic Parts And Components Raw material suppliers: Silicon wafer is the indispensable raw material required in the chip manufacturing process. Japanese companies account for over half of global silicon wafer supply.Chart 14 shows that silicon wafer inventory in Japan has had a significant buildup in volume terms since late last year. Importantly, it is not declining yet. Chart 14Silicon Wafer Inventory: A Significant Buildup As Well Silicon Wafer Inventory: A Significant Buildup As Well Silicon Wafer Inventory: A Significant Buildup As Well Outsourced semiconductor assembly and test (OSAT) providers: Both Singapore and Thailand are OSAT providers while they also manufacture, assemble and export electronic products. Both countries are closer to the downstream side of the semiconductor supply chain. Semiconductor inventory at these two countries has also jumped to a record high (Chart 15).   Chart 15Singapore and Thailand: Record High Semiconductor inventory Singapore and Thailand: Record High Semiconductor inventory Singapore and Thailand: Record High Semiconductor inventory   Importantly, a marginal improvement in demand will tend to support spot prices. For example, in the memory chip market, falling prices denote weak demand relative to excess supply. When prices of DRAM and NAND start to form a bottom or decisively move up, this may indicate the arrival of a cyclical upturn. So far, both DRAM and NAND prices are continuing to fall (Chart 16). In addition, the prices of silicon wafer – the most important raw material used in the chip-making process – have declined in the first half of this year.3  Chart 16Still Falling Memory Chip Prices Still Falling Memory Chip Prices Still Falling Memory Chip Prices Chart 17Deflating DRAM Prices Suggest Downside Risks To Korean Tech Stocks Deflating DRAM Prices Suggest Downside Risks To Korean Tech Stocks Deflating DRAM Prices Suggest Downside Risks To Korean Tech Stocks In short, prices are the most important variable to monitor. Chart 17 exhibits the high correlation between DRAM prices and the Korean technology sector stock prices. Falling DRAM prices suggest downside risks to technology stocks in Korea. Samsung accounts for about 65% of Korea’s tech index and 27% of the overall Korean equity index. Memory chips accounted for 68% of Samsung’s operating profits in the first quarter of this year.   Bottom Line: There has been involuntary inventory accumulation along the entire supply chain of semiconductors. This and ongoing price deflation among various types of semiconductors foreshadow a downbeat near-term outlook. The Interpretation Of Some Positive Developments There have been some positive developments in the past several months. Taiwanese PMI new orders diffusion index in the electronics sector jumped out of deep contraction to reach zero, and Chinese semiconductor imports halted their decline in both volume and value terms (Chart 18). The improvement in the aforementioned data was probably mainly due to large semiconductor purchases by China to hedge the rising risk of U.S. blocking China’s technological development (Chart 19). Chart 18Some Positive Development Some Positive Development Some Positive Development Chart 19China: More Semiconductors Purchases Before The Tariff And U.S. Huawei Ban? China: More Semiconductors Purchases Before The Tariff And U.S. Huawei Ban? China: More Semiconductors Purchases Before The Tariff And U.S. Huawei Ban? Besides, Huawei smartphone sales have been booming, which we deliberated on page 5, could have been responsible for the improvement in these data. This one-off surge will likely dwindle going forward. Investment Conclusions We remain negative on Asian semiconductor share prices in absolute terms. A continued contraction in global semiconductor sales will further squeeze their profits. In relative terms, we are neutral on the Asian semiconductor sector: we continue recommending market-weight allocation to Taiwan’s overall market and the Korean technology sector within the EM equity benchmark. As a new trade, we recommend going long Asian semiconductor stocks and short the S&P 500 semiconductor index over the next three to six months (Chart 20). The Bloomberg Asia Pacific semiconductor index has nine stocks. Samsung and TSMC account for 42% and 38% of the index, respectively. There has been involuntary inventory accumulation along the entire supply chain of semiconductors. This and ongoing price deflation among various types of semiconductors foreshadow the downbeat near-term outlook. Samsung will likely benefit from the U.S. ban on Huawei in the smartphone sector outside of China. In addition, Samsung will win some market share from Apple as the latter does not have a 5G phone to release this year. These positive factors may partially offset the negative impact from falling memory prices and demand on Samsung. The S&P 500 semiconductor index has 13 stocks. Intel, Broadcom, Texas Instruments and Qualcomm are the top five constituents, together accounting for nearly 70% of the index. Most of these companies are IC designing companies, which will likely suffer as Chinese demand for their products shrink due to the U.S. administration's ban on Huawei. This position will also benefit from U.S. dollar appreciation. A firm dollar will hurt profits of U.S. semiconductor stocks. In turn, currency depreciation in Korea and Taiwan will on the margin benefit Asian semiconductor stocks. Chart 20Recommend Long Asia Pacific Semiconductor Stock Vs. S&P 500 Semiconductor Index Recommend Long Asia Pacific Semiconductor Stock Vs. S&P 500 Semiconductor Index Recommend Long Asia Pacific Semiconductor Stock Vs. S&P 500 Semiconductor Index Chart 21The 2015 Experience The 2015 Experience The 2015 Experience Chart 21 shows that global foundry companies outperformed global IC designing companies during the final phase of the 2015 cyclical downturn. Odds are that these dynamics will play out in this downturn as well. Finally, the relative performance of Asian semiconductor stocks versus U.S. ones is oversold and might stage some sort of mean reversion (Chart 20). Ellen JingYuan He, Associate Vice President ellenj@bcaresearch.com Footnotes 1      Please note that here the calculation for “the number of months of the growth contraction” is different from the one for the “peak-to-bottom duration” in Table 1. “The number of months of the growth contraction” equals the number of months when semiconductor sales year-on-year growth is negative. 2      https://marketrealist.com/2019/05/nvidias-data-center-revenue-inference-rendering-and-edge/ 3      http://www.sohu.com/a/300386061_132567, http://news.moore.ren/industry/104958.htm   Equity Recommendations Fixed-Income, Credit And Currency Recommendations

Put Tech on Downgrade Alert Put Tech on Downgrade Alert We are compelled to put the S&P tech sector on our downgrade watch list as President Trump’s hawkish trade talk and actions since May 5 warn that tech revenues (60% export exposure) and profits will likely remain under intense downward pressure. Our tech EPS model is also flashing red on the back of sinking capex and an appreciating U.S. dollar (bottom panel). We will be downgrading the tech sector to underweight via the S&P software index, the tech sector’s largest industry group on a market cap basis. A downgrade to neutral in the S&P software index would push our S&P tech sector weight to a below benchmark allocation. Thus, we are initiating a stop near the 10% relative return mark on the S&P software high-conviction overweight call since the December 3, 2018 inception. We also lift the stop to 27% from 17% relative return on the cyclical overweight we have on the S&P software index since the November 27, 2017 inception. Bottom Line: We are compelled to put the tech sector on our downgrade watch list. We will execute the S&P tech sector downgrade to underweight when the S&P software index’s stops are triggered. This would push the S&P software index to neutral from currently overweight.