ratings.
- Key Trends Could Be Positive
Several key trends are progressing in the chip industry that could have positive effects on ratings for some companies in coming years (see table below). These include:
+ The shift of the industry towards consumer electronics, from enterprise-based computing applications;
+ Technology differences between computing and consumer semiconductors;
+ A deceleration of overall growth rates as the enterprise computing sector matures;
+ The ever-increasing cost of leading edge factories;
+ Electronics companies' requirements for global support; and
+ A transition towards asset-light or fabless models for many chip companies.
As noted below, most semiconductors used for computing applications likely will continue to be manufactured in large, expensive factories, owned by the chip design company, while most chips used for consumer applications could well be designed by fabless "design houses," but manufactured on their
behalf by major foundries in coming years. Fabless companies' free cash flows tend to be more stable than manufacturing-centric chip companies, and the growing trend towards fabless and fab-lite companies could contribute to selectively higher ratings over time.
At the same time, leveraged buyout (LBO) companies have shown increased willingness to undertake large acquisitions in the chip industry in the past year. That same superior free cash flow stability that fabless companies enjoy could lead to heightened LBO interest in the sector, and trigger credit
deterioration for the targets, whether through defensive measures or following completion of an LBO.
Credit implications for specific issuers will be assessed as any deals are announced.
A chip-industry shift toward consumer electronics, rather than enterprise-based
computing applications Most enterprise desktops and industrial benchtops have been PC-equipped for many years. As a result, those markets are no longer in strong growth modes, and are largely driven by multi-year scheduled replacement cycles. Replacement cycles can be stretched when companies feel financial pressures, as they "make do" with existing equipment until business conditions recover (unlike conditions during the industry's growth phase, when computerizing a business was a "must do"). Because enterprise PCs principally are used as portals to corporate databases (rather than for their standalone computing capability), performance enhancement rarely is the reason for business PC replacement. Enterprises may stretch their PC cycles to take advantage of new releases of Microsoft Windows or Office, significantly
because of enhancements to security and collaboration capabilities; but once the software is available, their natural conservatism means they prefer not to be early adopters of new technology.By contrast, the diversity and pace of new consumer product introductions clearly has accelerated: They now are the major revenue growth driver for the electronics industry. The proliferation of consumer electronic devices includes flat panel televisions (now that the price point has become acceptable to consumers), phones, game consoles, music players, PDAs, portable DVD players, digital cameras, navigation systems, advanced set-top boxes and smart appliances. Many of these did not exist 10 years ago, and are still in fairly early stages of market penetration. Because many consumer products are
priced below $300, they are often discretionary purchases, which helps fuel a rapid upgrade cycle. These factors are compounded by the rapid penetration of consumer products in emerging markets such as China and India.
Unlike enterprises, consumers do purchase new PCs when new releases of operating system software are introduced, because the successive releases of Windows have included major improvements in media processing, which shows no sign of abating. Maturity in the consumer electronics market is many
years in the future.
Consumer electronics and computing require different classes of semiconductors
As far as the semiconductor markets are concerned, there is a technology gulf between computing and consumer applications. Consumer electronics devices perform very different functions than PCs, and largely use a different class of semiconductors with different technological requirements. Consumer PCs,
in particular, tend to have higher quality audio and graphics capabilities than enterprise PCs.
While enterprise PCs are used for data-base access, document creation, or calculations, consumer PC applications generally involve the capture, storage, transfer, and presentation of large sound or image files, with little alteration or adjustment of the data. As such, consumer applications generally do not create content, and are not "computing" applications. Other than PCs, most consumer electronics chips fall in the classes of power management (portable devices), analog (amplifiers), flash memory (portable storage), and radio-frequency (communication) chips.
The traditional chip companies grew up 30 or more years ago in a computing-centered market. They were compelled to use inhouse chip manufacturing capacity because there were no other options at the time.
Leading computing-based companies include microprocessor suppliers Intel Corp. and Advanced Micro Devices Inc. (AMD), as well as dynamic memory (DRAM) suppliers Samsung Electronics Co. Ltd., Micron Technology Inc., Qimonda AG, NEC Electronics Corp., Hynix Semiconductor Inc., Renesas Technology Corp., and Toshiba Corp. Notably, these companies continue to manufacture their own products in house, and are highly asset-intensive. Ratings in this sector will continue to recognize high asset intensity, potential volatility, and a relatively narrow business base.
However, many consumer-oriented semiconductor companies originated more recently, and could use excess capacity belonging to the well-established companies; the newcomers had the freedom to become fabless design houses from inception. Most consumer semiconductors, other than flash memory, are not
technology-intensive, and do not require advanced manufacturing resources. Further, because most consumer-focused chips are much smaller than microprocessors, few consumer electronics semiconductors are made in high volumes, measured in terms of wafers per year: It is increasingly hard
to justify in-house manufacturing under this scenario. The major exceptions have been the Japanese integrated electronics companies such as Sony Corp., NEC Corp., Matsushita Electric Industrial Co. Ltd., or Fujitsu Ltd., many of which make chips (largely for internal consumption), electronic products and
appliances or electrical goods, but even here, changes are occurring.
Consumer-focused companies include most suppliers of analog chips, such as Analog Devices Inc. or National Semiconductor Corp.; power-management suppliers, such as International Rectifier Corp.; radio-frequency specialists, such as RF Micro Devices Inc. or Conexant Systems Inc.; and cell-phone
specialists such as Texas Instruments Inc. (TI), Qualcomm Inc., and Freescale Semiconductor Corp.
Ratings in this sector continue to recognize that analog-based processes do not require leading edge equipment, are moderately asset intensive, and are centered on unique process requirements, rather than the speed- and size-based criteria common for many computing applications.Semiconductor growth is decelerating as the industry shifts towards consumer electronics, but the slowdown is in computing, not consumer applications
The majority of semiconductors used in PCs (on a dollar basis) are either microprocessors or DRAM chips. Because microprocessors now are $50-$100 items--one-tenth their price 10 years ago--the total dollar semiconductor content of a PC has declined over the years, even as chips make up a rising percentage of PC content. Although PCs will continue to penetrate underserved areas in the developing world, and replacement cycles will continue in the developed economies, the steady price decline of key PC components likely will lead to sustained PC-related semiconductor sales growth in only the mid-single
digits percentages over the longer term.
While the overall semiconductor industry's revenues grew 21% from 2000 to 2006, key PC supplier Intel Corp. grew 5%, and the global DRAM industry (again, largely a PC market), also grew only 5% over the same period. The Semiconductor Industry Association expects growth of 9% for 2006-2009, while In-Stat forecasts a compound annual growth rate of 6.2% for 2005-2010, and other forecasters suggest similar growth ranges; most of the growth is in non-PC areas. Although PCs will remain a large portion of total electronics production, and PCs will continue to be major consumers of semiconductors for many years, the semiconductor industry's growth clearly will be centered on consumer applications. This shift has stimulated some companies to redefine their target markets in order to continue good growth rates.
As examples, AMD, Micron, and Samsung have recognized the shift in the industry's center by their moves into consumer markets including graphics chips and NAND flash memory, while Intel has entered markets such as communications with an eye towards diversification, although their success in this area has been limited. While their business models will continue to evolve, these companies' business models likely will be dominated by their computing focus, and their asset-intensive business models, for the longer term.
Leading-edge factories are beyond the reach of all but a handful of companies
It is uneconomical to manufacture chips having ever-shrinking feature sizes and ever-increasing numbers of layers, unless those chips are produced in high volumes. Resulting factory capital costs significantly exceed $1 billion. Only a handful of very large companies--with revenues well above $5 billion--can
manage investments of this level without subsidies. The list is short and centered on thecomputing-centered companies, plus the major foundries. (Microprocessors require small feature sizes to attain high clock speeds, while the growing number of transistors in successive chip generations must be
offset by smaller features to keep the overall chip size within bounds. Memories require small feature sizes to maximize the number of chips produced per wafer, because of severe price competition for these commodity devices. The huge volumes produced by the major foundries also justify large factories.) The only asset-intensive consumer chip market is the NAND flash business, for portable music players, cameras and phones; the key participants in this market are DRAM makers seeking diversification and new growth opportunities; the manufacturing equipment and processes are highly fungible between the two markets. Most other semiconductor companies cannot afford to build new factories, although they may update older facilities, frequently with second-hand machinery; this equipment has been well-maintained but no longer meets the needs of leading-edge companies, and is perfectly suited to
mainstream companies.
The largest companies have the scale to be economically efficient even with a large manufacturing asset base and will likely remain largely vertically integrated. While the asset base provides a very high barrier to entry, it brings the requirement for companies to maintain the assets through continued capital expenditures, even if operating cash flows are depressed. This makes the companies subject to substantial free cash flow volatility, which can largely be offset by ample liquidity. For example, Intel Corp.'s cash balances were $10 billion at Dec. 31, 2006.
Mid-scale companies likely will shift their production increasingly towards foundries as their inhouse equipment ages; this changed business model can offer lower total operating costs despite higher marginal costs as inefficient internal operations are eliminated.
In addition to the rising cost of manufacturing equipment, process development costs continue to skyrocket. In June 2006, Toshiba, Fujitsu, NEC Electronics, and Renesas Technology announced they would jointly develop process technologies for 45 nanometer chips, and were considering shared investment in a 45-nanometer factory. At the same time, Toshiba, Renesas Technology, and Hitachi scrapped a plan to build a 65-nanometer Japanese chip foundry. Japanese semiconductor makers' manufacturing strategies remain in transition.
In January 2007, TI, an industry founder and technology leader, cancelled its 32-and-beyond nanometer process development because it duplicated work being done by its foundry partners, which already make 50% of its logic chips. TI still intends to use a mix of in-house and foundry manufacturing resources for its logic chips, while maintaining in-house capacity for analog products. Similarly, Cypress Semiconductor Corp. is selling its development business to a private equity consortium, and will purchase services as needed from the unit.
Mid-scale companies will have ample choices as they move towards an asset-light strategy. The foundries have ample mainstream capacity available, as they continually upgrade to meet leading edge demands of their key customers. The foundries offer a very high degree of process expertise and scale
economies to mid-scale chip companies, which largely offset the foundries' high per-wafer processing charge. Fabless semiconductor suppliers were about 20% of industry sales in 2006, and have more than doubled their share since 2000; similar growth numbers apply to the foundries. Manufacturing efficiencies
will continue to drive the paired growth of the fabless and foundry sectors over the long term, and only the largest vertically integrated chip companies, and smaller companies who rely on specialty processes, will rely principally on in-house capacity.
A global customer base requires a stronger supplier network, which can lead to consolidation
The electronics industry relies on its supplier network for worldwide support, while all participants seek to minimize inventory levels and strengthen customer-supplier relationships. These factors place pressures on suppliers to consolidate; well-executed consolidations tend to reduce total inventory, eliminate redundant R&D projects and customer-support sites, and create scale economies for other operating expenses. Consolidation usually leads to broader addressed markets and geographies, leading to diversification and new technical capabilities. Mainstream semiconductor companies likely will continue
the trend toward mergers, such as the pending LSI Logic Corp.-Agere Systems Inc. transaction. AMD's acquisition of ATI Technologies Inc. in late 2006 strengthened AMD's supply chain and should help the company to migrate towards the faster-growing consumer electronics market. Fairchild Semiconductor
International Inc.'s planned $180 million acquisition of Taiwan-based System General Corp. should strengthen Fairchild's position in the key Taiwan market. Similarly, International Rectifier Corp.'s (IR) sale of a broad group of lower-margin products to commodity specialist Vishay Intertechnology (also late 2006)
provided Vishay with additional scale, which should contribute to its operating efficiency. Following a leveraged transaction, Avago sold its storage semiconductors to PMC-Sierra Inc. and its printer ASICs to Marvell Technology Group Ltd. to further concentrate on its core business, and pay down debt.
The ratings implications of these deals have varied, depending on the effects on business profile and the financing involved. The announcement of the stock-funded LSI/Agere transaction led to a CreditWatch positive placement for LSI, while the debt-funded AMD/ATI agreement led to a negative outlook on AMD.
The IR/Vishay asset sale did not affect our rating or outlook for either company; the Avago transactions were contemplated in our initial rating for the company, and our rating for Fairchild had long anticipated acquisitions similar to the System General transaction.
Free cash flow volatility depends on a company's asset intensity
The semiconductor industry overall remains extremely capital-intensive. Global industry capital expenditures in 2006 were about $41 billion, or about 16% of chip-industry sales. However, expenditures vary widely from company to company, with fabless companies spending far less than asset-based companies.
For example, over the past five years, fabless Cirrus Logic Inc.'s capital expenditures were 2.5% of sales; Conexant spent 3%; and Nvidia 4.6%. Asset-intensive companies spent far more: Intel Corp.'s five-year average was 16%; AMD's was 23%; Taiwan Semiconductor Manufacturing Corp. was 28%; and Micron Technologies Inc. spent 25% of sales. (These expenditures include wafer fabrication equipment, assembly and test gear, capitalized software, buildings, and ancillary equipment; wafer fabrication equipment is roughly half the total.)
Because capital expenditures are deferrable over the short term, chip companies can reduce expenditures to preserve cash in a downcycle. However, the onset of a true downturn (as opposed to a brief fluctuation) is hard to gauge in an atmosphere of competitive pricing actions, product introduction
cycles, and seasonal factors. Ultimately, a chipmaker's decision to defer capital expenditures tends to lag the onset of a downturn by several quarters. Similarly, companies would want to expand or update capacity in advance of a recovery, but the timing of the recovery is equally murky. Thus, asset-intensive companies' capital expenditures can vary widely from year to year, and may not synchronize with operating cash flows, leading to potentially volatile free cash flows. As examples, Micron's annual free cash flows have varied from $650 million to negative $700 million in the past five years, and AMD's annual free cash flows have varied from negative $760 million to positive $50 million over the same time.
This volatility is a major reason for vertically integrated chip companies to retain ample cash balances. The picture is quite different for fabless companies. While even the fabless company must eventually upgrade its computers and similar equipment, the expenditure levels remain far lower than for asset-based companies. This reduces downside negative free cash flows, allowing them to carry lower cash balances without inordinate liquidity risk.A possible future: Lower cash flow volatility can lead to higher ratings
Our investment-grade rating on Xilinx reflects its highly liquid balance sheet and good free cash flows, which in turn result from high margins and its fabless business model. In the past five years, Xilinx' free cash flows varied between $155 million and $415 million. Our outlook on Agere was revised to stable from negative in late 2003, when it consolidated its manufacturing in one site and offered that site for sale. Similarly, our rating outlook for LSI was revised to stable from negative following its decision to become fabless in early 2006, and revised to CreditWatch positive following the announcement of the Agere merger. While the majority of fabless companies are debt-free and not rated, we believe a fabless business model would generally be favorable in our rating assessment of a company.
Another possible future: Lower cash flow volatility can lead to additional balance sheet leverage
The trend toward lower free cash flow volatility for fabless companies (and to a lesser degree, for fab-lite companies) eases the way for such companies to be leveraged. Freescale Semiconductor Inc., with post-LBO cash balances of about $600 million, compared with $3.5 billion before the LBO, is one such
fab-lite case. Freescale benefits from the large capital expenditures made several years earlier when it was a Motorola division; the company's resulting ample manufacturing capacity has allowed it to trim capital expenditures since its separation from Motorola, continuing into its new life as a privately held
company. Freescale's depreciation expenses have exceeded its capital expenditures despite revenue growth, contributing to free cash flows, important to this highly leveraged company. Freescale outsources 20%-25% of its wafer production. Furthermore, the company's role as an embedded supplier for its key customers' platforms helps to reduce near- to intermediate-term revenue risk, compared with companies that rely primarily on merchant market sales.
NXP B.V., also substantially a platform supplier, has a long history of working with foundries: Its former parent company, N.V. Philips, was the founding member of TSMC, the Taiwanese foundry. NXP expects to increase outsourcing to about 40% over the next five years. Like Freescale, NXP benefits from significant capital expenditures made under Philips in recent years, and should be successful with only modest capital expenditures over the medium term. To further cut its capital expenditure commitments, NXP has exercised its rights to exit the Crolles 2 technology-development joint venture, and has strengthened its technology relationship with TSMC; Freescale is also exiting Crolles to enter a technology partnership with IBM Corp. The third Crolles senior member, STMicroelectronics NV, has indicated it will look elsewhere for 32 nanometer technology.
While the industry's asset-light trend is favorable for ratings, the longer-term ratings effect of this evolving business model will depend on whether companies adopt more aggressive financial policies to offset their reduced business risk, and whether the sector LBO trend continues. We will monitor the industry's overall risk profile as it evolves, and reflect the changes in our ratings.
¬ Haymarket Media Limited. All rights reserved.
