Semiconductor Consolidation vs. Specialization

Why despite two years and counting of mega-mergers, there are still many opportunities for growth and profit.

Over the last two years, the semiconductor industry has been swamped by a wave of consolidations. The cumulative transaction value of chip-related mergers during this time topped an eye-popping $200 billion. In terms of dollars spent on deals, the 2015-2016 yearly average is nearly five times that of the previous biggest ever year, 2011 ($23 billion). And of course Mentor Graphics was no stranger to the phenomenon, having been acquired by Siemens AG for $4.5 billion in deal that was finalized this spring.

To many, this consolidation seems natural. After all, the semiconductor industry is now more than 60 years old and the growth rate is slowing, as is common for most maturing industries. Semiconductor sales as a percentage of total electronic equipment sales have been relatively flat for the last twenty years. Consolidation through mergers gradually increases as industries mature so why should the semiconductor industry be any different?

The answer is that eventually the same phenomenon will apply to us — just not quite yet. Our industry was and remains more than a bit different from nearly any other, mostly due to the exponential magic of Moore’s Law. While some consolidation is apparent, my take on the numbers is that specialization is the big trend and opportunity of the moment.

Consolidation Is Not Normal for the Semiconductor Industry

Any consolidation that’s going on is a totally new phenomenon for semiconductors. Because of the rapid innovation that has occurred in the semiconductor industry over many decades, the industry has shown no consolidation through almost all its history. Indeed, even after the flurry of activity in recent years, the combined market shares of the 50 largest companies is actually lower than it was in the early part of this century (Figure 1).

In fact, considering a slightly longer time span, the data reveal a trend toward deconsolidation. In 1966, three companies, TI, Fairchild and Motorola, accounted for approximately 70% of the semiconductor industry revenue. By 1972, that combined market share had dropped to 53%. The largest semiconductor company today, Intel, has approximately 15% market share, just 2% more than the number one, Texas Instruments, had in 1972 and less than Intel had in 2011 (despite the acquisition of Altera) (See Figure 2).

 

After more than 60 years of deconsolidation, the activity of the past two years does register as the first material, though very modest, step toward consolidation, at least when considered from a certain vantage point. Consider that, for example, the combined market share of the top 10 companies is 2.5% higher than the previous high (1984). Each percent of market share represents a little more than $3 billion dollars in revenue, so the amount of the increase is on par with total annual EDA industry review — real money for those of us (like me) who have spent decades in EDA. For many, the agglomeration uptick appears to be the inevitable maturation of an industry in which growth rates have slowed and further increases in return on investment depend upon the efficiencies of consolidation rather than upon growth.

The driving force for consolidation is typically the efficiencies achieved through economies of scale with larger volume. In most industries, manufacturing economies of scale are the most important. This is true for the integrated device manufacturers, or IDMs, of the semiconductor industry as well. However, these IDMs constitute a continuously shrinking share of the semiconductor industry and now consist mostly of memory producers, several very large IDMs like Intel and Samsung, and process-differentiated companies that produce analog, mixed-signal, RF and power semiconductors.

Nearly a third of worldwide IC sales come from fabless companies, up from around 10% at the turn of the millennium. Manufacturing economies of scale generated by the merger of fabless companies are limited (Figure 3). If two large semiconductor companies merge, the volume discounts they receive from the wafer foundries, as well as from the assembly and test subcontractors, are not greatly different. The additional volume achieved through merger likely adds only modest additional discounts.

Do Economies of Scale Apply to the Semiconductor Industry?

Of course, manufacturing costs are only part of the volume-sensitive benefit of mergers. The cost benefits of a merger, often referred to as synergies, include many other types of economies of scale. One would expect to see a correlation of these cost benefits with size of semiconductor companies.

What is remarkable is that such a correlation doesn’t appear to exist (Figure 4).

A look at data from the top 130 public semiconductor companies over a period of five years (Figure 4) shows no relationship between size and profitability. A linear regression coefficient of 0.0544 suggests no correlation at all. This is despite the fact that semiconductor memory companies, whose profitability definitely does correlate with total revenue, are included in this data sample.

Comparison of the most profitable semiconductor companies seems to support this same conclusion (Figure 5). The largest semiconductor company, Intel, had a very respectable operating profit of nearly 26% in 2015 on more than $55 billion of revenue. But Linear Technology, with only $1.4B of revenue achieved a 45% operating profit and Xilinx, with $2.2 billion of revenue, achieved a 30.5% operating profit. You might think that operating profit is dependent upon the sector in which a company operates but the top 20 most profitable companies in 2015 cover most of the range of possibilities — with analog (8), memory (3), microprocessor (2), foundry (3), FPGA (1) and ASIC (4).

So it appears that economies of scale are of limited importance in the motivation for consolidation. As a result, semiconductor companies rarely try to achieve the benefits of volume when a particular sector of the industry matures. We saw this in the wireless industry when most of the market share leaders in wireless baseband chips actually shut down their businesses rather than sell them as the market consolidated. TI was the leader in baseband chips and, like Broadcom, ST Micro, Marvell and more, found no buyers as the market commoditized. Others, like Qualcomm, HiSilicon, Mediatek and Spreadtrum showed a willingness to hang with it as prices declined. Even those who were able to sell what remained of their declining wireless chip businesses, like NXP, were not acquired by a consolidating leader.

Data seem to argue that the current consolidation of the semiconductor industry is not driven primarily by the desire to achieve economies of scale. What other factors, then, are driving semiconductor mergers and acquisitions?

Impact of Liquidity

Another obvious force for semiconductor industry consolidation is liquidity. Cash is available in large quantities at low cost with long term financing and, despite the recent uptick in interest rates, remain at levels that are unprecedented over the history of the semiconductor industry and indeed over a span of several hundred years. Almost any acquisition of a profitable company can be accretive to earnings for the acquirer. Traditional banks have become more conservative because of the new restrictions of Dodd Frank but there are many other sources of borrowing, like private equity, that reflect conditions in the free market.

Figure 6 shows recent cycles of merger and acquisition activity. As central banks have loosened restraints on borrowing, booms have been created. This has resulted in record levels of total acquisitions (not just semiconductor companies) in terms of market value.

The “DotCom” boom of the 1990s peaked at $3.5 trillion of acquisitions in the year 2000 before the crash. In 2007, the number was $4.6 trillion. Current innovations in liquidity creation by central bankers have driven a level of $5 trillion in acquisitions in 2015, a blistering pace that continued in 2016 with mega-deals in a host of industries, including media (Time Warner & AT&T, $85.4 billion), energy (General Electric oil business combining with Baker Hughes to create a $32 billion co-owned entity), tobacco (British American Tobacco & Reynolds American, $47 billion) and agriculture (Bayer & Monsanto, $56 billion). U.S. deal activity in October 2016 was $329 billion, according to Thomson Reuters, a record.

Semiconductor company liquidity has been fed by growth in cash generation and debt (Figures 7 and 8). Some financial analysts would argue that accumulating excess cash beyond the working capital needs of the business, or under-utilizing borrowing power, is an inefficient management of corporate assets. Whether that’s fully accurate or not, semiconductor company executives are paying much more attention to potential uses of cash and debt.

This semiconductor industry liquidity has also been augmented by Chinese government policy. Putting its money where its political policy is, the Chinese government has committed $20 billion of investment in the semiconductor industry with a target of 20% per year growth and a goal to move to semiconductor self-sufficiency. China’s annual spending on ICs is $103.5 billion, accounting for 36% of the total IC market, according to IC Insights.

What’s unusual about this $20 billion investment is that it’s not being done through the normal bureaucratic governmental channels; it’s a direct investment for a minority equity interest in private equity funds like Tsinghua Unigroup, Hua Capital and others that are profit motivated. They will invest in semiconductor manufacturing infrastructure, fabless semiconductor companies and the whole semiconductor ecosystem, motivated by a desire for a good return on their investment. Including the $97 billion in matching funds from the private equity companies, this is a total investment over five years of nearly $120 billion in an industry with annual revenue of only about $350 billion worldwide.

The likely result will be acceleration of an already red-hot semiconductor industry in China where both the number (Figure 9) and size (Figure 10) of IC/fabless firms are on the rise.

All of this acquisition fervor, both from Chinese as well as western companies, is likely to continue, as long as liquidity is freely available. What’s most interesting, however, is what happens to the companies after they are acquired, especially when it comes to the R&D spending that’s the lifeblood of innovation.

Influence of Consolidation on R&D

In almost all years of the semiconductor industry’s history, R&D spending of the public semiconductor companies has increased (Figure 11). Only three years in the last 25 showed a decrease in spending and only two of those decreases were material (near 10%), 2001 and 2009. Keeping up with Moore’s Law is expensive, so semiconductor R&D tends to grow at the same rate as semiconductor revenue growth, unlike industries that reduce R&D spending as a percentage of revenue as they mature. Yet will this trend continue?

Certainly that’s one of the big questions prompted by all the semiconductor merger and acquisition activity during the past five years, particularly since most M&A announcements come with projections of “synergies” that are expected to be achieved through operating expense reductions (Figure 12). These projections have averaged about a net 25% reduction. In most cases, the operating-expense reductions are not broken up into R&D, G&A, marketing, etc., but are expressed as a total amount of cost savings as a percent of revenue.

There are many ways to achieve these synergies. Some obvious ones that are likely to improve operating efficiencies include elimination of duplication of corporate management, e.g. CEO, CFO, etc., or consolidation of functions like corporate marketing, investor relations, public relations and more. Another approach to operating expense reduction is to eliminate the less efficient, or less strategic, businesses via spin-off, thus reducing total corporate operating expense as a percentage of total corporate revenue. A good example of this approach was the acquisition of LSI Logic by AVAGO (now Broadcom PLC) shown in Figure 13.

 

More than $1 billion was generated for AVAGO to invest in higher priority businesses while the businesses that were spun off became opportunities for growth and investment at Intel and Seagate. In this case, total semiconductor R&D spending may have actually increased as a result of the merger since the spin-offs were apparently more highly valued by their new owners than by AVAGO.

The above approaches to operating-expense reduction appear to offer the opportunity to increase efficiency while growing R&D and potentially developing more revenue and products for the semiconductor industry. On the other hand, the promise of a reduction in operating expense frequently leads to investors expecting reduced R&D spending and a consequent reduction in the number of people developing, marketing and supporting new products. At least one major merger in the past year made such a promise. Does this mean that the semiconductor industry has finally reached the level of maturity that requires reduction in R&D spending to continue growing earnings as revenue growth slows? Maybe it does. But there are some other questions to be analyzed.

***

First, what happens to all those R&D engineers who were developing products when a merger eliminated their jobs? Do they enter the unemployment roles for an indefinite period? Figure 14 suggests otherwise.

The current unemployment rate for electronic engineers is a remarkably low 1% and the average over the last fifteen years has been about 3%. “Full employment” would be between 5 and 5.5%. Apparently, when electronic engineers leave a job, they quickly find another.

Acquisitions that plan on improved operating margins from reduced R&D may be playing a short-term game. Since revenue of the acquired company is likely to continue for some time despite reductions in R&D spending, the operating margins of the acquirer are likely to improve. But the void created by these R&D cost reductions offers an opportunity for competitors, or new start-ups, to enter the market to take advantage of the future declining revenue caused by lack of investment. (Figure 15).

What if we really are reaching a point of maturity in semiconductor industry evolution with only limited revenue growth? Despite the cost and challenges of Moore’s Law, perhaps economics requires a reduction in R&D. But by how much? Look back at more than 30 years of the semiconductor industry (Figure 16). R&D spending as a percentage of revenue has been remarkably stable at 13% to 14% of revenue. It would be surprising, but not impossible, if this long term trend abruptly changed as a result of the merger mania transition we’re now experiencing.

The other possibility, one that I believe to be more likely, is that we are in between major waves of growth that are typical of the semiconductor industry. Wireless handsets propelled the most recent wave of growth and we are waiting for something new, like the Internet of Things, to generate a new wave. Historically, new semiconductor growth is ushered in by new applications that become possible when the cost per function, or some other new capability, makes the new application possible (Figure 17). In recent years, the cost per transistor for semiconductors has decreased more than 30% per year, just as it has, on average, for most of the last 60 years. It’s likely that continuation of this trend will, in fact, enable future waves of new semiconductor applications.

In the meantime, the continued low cost of borrowing and other factors like the Chinese government stimulation of their domestic semiconductor industry will provide fuel for some level of increased industry consolidation. But that consolidation is likely to show up as increased dominance by specific companies in focused application segments. The combined market share of the 50 largest semiconductor companies is still less than it was 10 years ago. And the combined market share of the 10 largest semiconductor companies is only 2% higher than the average of the last forty years.

Indeed I think the trend that’s most apparent is specialization and the pursuit of niches, albeit potentially gigantic ones.

Specialization Trend

Consider the automotive market. Qualcomm’s acquisition of NXP makes the San Diego company the world’s biggest automotive semiconductor company, with nearly $4 billion in automotive-related revenue. This is all the more striking considering Qualcomm was No. 41 on the list of automotive semiconductor suppliers in 2014. The same is true when it comes to chips for wireless and enterprise data center applications, a market that’s now dominated by Avago/Broadcom and Intel (after its acquisition of Altera). Yet Intel’s overall market share of the semiconductor industry, as the largest supplier, is about the same at the market share of the number one supplier forty years ago, which was Texas Instruments, where I began my career in the 1970s.

Surely, if the trend of the last two years continues, there will eventually be a significant increase in the market share of the fifty largest semiconductor companies, especially since there are so few new semiconductor startups (and most of those are in China). But there are several reasons why this will take a long time, if it ever happens at all.

First, the smaller companies are growing much faster percentage-wise than the large companies in the semiconductor industry. Naturally, the largest companies have the majority of their revenue in the more mature applications of semiconductors. Second, while there are very few new startups, we are moving into an era where companies engaged in businesses other than semiconductors are becoming “semiconductor companies.” Apple, for example, uses major foundries to directly purchase its application processors, just as fabless semiconductor companies do. Yet you don’t see Apple on any of the Top 20 lists of semiconductor companies. Google provided one of the best keynotes at the 2015 Design Automation Conference highlighting a chip design for medical applications. In addition, the Internet of Things is stimulating interest by all sorts of new entrants into the semiconductor industry who envision packaged solutions for data collection and transmission that will be combined with deep data analysis to form businesses with much more value than the components themselves.

Limited Impact of Merger Mania

So what’s happening to the semiconductor industry?  Is it really like the stagnating consolidation of the steel and automotive industries of the 70ss. Doubtful. The market price per transistor decreased more than 30% last year, just as it has in most years since 1950 (Figure 18). Whenever costs decrease that much, new applications will become possible. Whether these new applications are driven by startups, or by existing non-semiconductor, companies, the semiconductor industry looks like it still has lots of opportunities for growth.

For now, rest assured that the future of the semiconductor industry is likely to be more complex, and quite possibly much better, than the “consolidation and aging” scenario that appears in some of the gloomier recent forecasts. And the profitability of semiconductor companies is likely to continue to improve as individual companies gain dominance through specialization in specific niche applications.