At the recent GSA Silicon Summit at the Computer History Museum in Mountain View, CA, the keynote was given by Wally Rhines on consolidation in the semiconductor industry, titled "Merger Mania." He started off by pointing out that the number of recent mergers is not unusual, but the size of the mergers is. There were $160B of mergers and acquisitions last year, including Intel-Altera, Avago-Broadcom and NXP-Freescale. Plus, although it hadn't happened by the time of the presentation, ADI/Linear. However, consolidation is not normal in the semiconductor industry. In fact since the 1960s, the industry has been de-consolidating. At the top of the market, the share of the #1 company since 1972 has been flat (TI back in 1972, Intel today). And while the companies have changed over time, the market share of the top 5 and the top 10 companies has grown only very slightly over the period. However, 2015 might be the start of a change. The market share of the top 10 companies has increased by 2.5% over the prior high in 1984. The companies in the top 10 of the semiconductor industry have changed a lot over the years. Over half of the companies that have ever been in the top 10 are no longer in the top 10. In fact if you go all the way back to 1955, only one company, Texas Instruments, is still in the top 10. The primary reason for this is that market and technology discontinuities bring new companies into the market and make them successful. Wally took a stab at explaining why there is suddenly an acceleration in mergers. Generally, mergers happen for one of three reasons: Economies of scale Financial leverage Regulation/government mandate Economies of scale mean an increase in production at lower marginal costs. This seems to still be true in the analog, power, and RF markets, where the many mergers seem to be driven like this. But in the SoC market mainstream digital is mostly fabless, so there are very limited economies of scale since large fabless companies already have as big a discount as anyone can get from the foundries. For example, the merged NXP/Freescale is unlikely to get a bigger discount than NXP and Freescale already got, since they were both already large companies. Looking to the industry as a whole, if the economies of scale story is true, then larger companies should be more profitable. If you plot the data, thought, it is all over the place. For example, Linear has twice the operating margin of Intel despite being a fraction of the size. The reality is that there are all sorts of different semiconductor businesses with different sizes and different margin points. The table below lists the top 20 semiconductor companies by operating margin (2014 numbers). One interesting datapoint is that Altera's margins were higher than Intel's (of course Intel acquired Altera last year so it won't have its own line any more). So if economies of scale don't explain the surge in M&A, what about financial leverage? Interest rates are really low which has two effects. One is that it is cheap to borrow money to finance M&A, and also holding cash (and near cash) is not that attractive. It turns out that the cash held by the top 25 semiconductor companies is growing at a CAGR of over 8% per year (over the last five years). An acquisition is easier with low interest rates. If company A borrows $1B at 3% to acquire company B, the merger will be accretive (and Wall Street will like it) as long as company B is making just $30M/year in profit. There are also some potential tax advantages. The third possible driver for the increase in M&A is regulation and government mandate. The biggest area that this is affecting is China. China has a plan to get increasingly self-sufficient in semiconductor, with a five-year plan to grow at over 20% per year by investing $20B in both the domestic market and through acquisitions. That $20B in government money is augmented by almost $100B more from local and regional governments and private equity. This is a huge sum: $120B to invest in the semiconductor industry that worldwide is $350B. This has resulted in a lot of Chinese M&A activity over the last couple of years. For EDA, the key number is semiconductor R&D. This has been very constant at around 14% of revenue. In fact semiconductor R&D has been between 13-15% of revenue for the last 32 years. The worry is that semiconductor R&D may reduce, but in fact there have only been four years in history when semiconductor R&D has shrunk, it almost always grows even in recessions and downturns. One way to reduce R&D expenses is to sell of parts of the acquired business that are unwanted, as Avago did selling its network processor business acquired from LSI (and others) to Intel. But this doesn't reduce R&D for the industry as a whole, since presumably the buying company wants to invest in the business. Companies can reduce R&D without affecting revenue in the short term, continuing to sell the existing products that have already been designed. But that doesn't work for long. Either the R&D investment has to be made to bring new generations of product to market, or it creates an opportunity for another company to invest in the area and steal market share. Even if engineers do get laid off, the don't stay idle for long. The quickly join (or start) another company. Economists generally regard 5% as "full" employment, with just a few people not working for the odd month between jobs. Right now, engineering unemployment in the semiconductor industry is below 1%. If you listen to any public EDA company's earnings call, everyone points out that the large size of acquisitions might have a long-term negative effect. Of course larger companies can have greater negotiating leverage with EDA companies, but the number of engineers who need tools is not decreasing, despite all the M&A. The big growth in semiconductor tends to come from successive waves of new product categories in the overall electronics market (which is about $2.5T). For decades, new products have had high semiconductor content. In many cases, it is the cost reduction from integrating large parts (or whole) products into silicon that drives the product category. The PC would not happen without microprocessors and DRAM, the smartphone would not happen without application processors and flash. Consolidation tends to be a defensive move when there are not attractive fast growing markets to attack. Wally's conclusions were that the current merger mania is sustained by: Low interest rates Chinese funding of acquisitions Delayed emergence of new semiconductor categories that will grow new markets and create new competitors Herbert Stein is famous for saying "if something cannot go on forever, it will stop." Wally's view was that consolidation may continue for a time but it will stop: Risks of future profitability of the acquired company Historically, large acquisitions have been less successful than anticipated Government veto on some acquisitions (CFIUS, MOFCOM, FTC, EC...) Too many acquisitions drives up the price of target companies R&D reductions are either impossible or reduce revenue growth Long-term interest rates will eventually increase Innovation will enable the next wave of semiconductor applications TL;DR: Current merger mania is limited. Previous: Chipworks Looks at Smartphones![]()
![]()
↧