Nikkei Electronics Asia -- February 2007
Cover Story
Volume-Cost Strategy Puts Chip Makers Back on Top

Semiconductor manufacturers, floundering until recently, have been able to improve their cost competitiveness by producing in volume, achieving the lowest possible cost, and investing profits back into their business.

"Significant improvements in productivity and yield have already achieved parity in cost competitiveness with DRAM manufacturers in Korea and Taiwan," said Yukio Sakamoto, president & CEO of Elpida Memory Inc of Japan. Masashi Muromachi, corporate executive vice president, president & CEO of the semiconductor company at Toshiba Corp of Japan, said, "We will not lose out in cost competitiveness to anyone in NAND Flash memory."
The key to the good performance enjoyed by Elpida Memory and Toshiba has been a major improvement in cost competitiveness (Fig 1). Both firms achieve high profits by manufacturing product at the lowest possible cost, and are successfully investing profits back to create a healthy cycle. Improved cost competitiveness means that even if product prices drop considerably, they will still produce a profit. Toshiba, for example, didn't start to tremble when the bottom fell out of the NAND Flash memory market. As Fumio Muraoka, representative executive officer of the firm explained, "The per-bit price of NAND Flash memory fell by about 50% in the 6-month period from April to September 2006. We had originally expected a 40% to 50% price drop over the year from April 2006 through March 2007, but the pace was far beyond our expectations." Even so, the firm generated adequate profit.

Volume Cuts Costs
The method these firms used to improve their cost competitiveness is clear: volume production to slash manufacturing costs. At the same time, they adopted new technologies such as finer geometry and multi-value recording in advance of their competitors to further lower costs. In short, their strategy was to leverage technology through volume.
One reason why they are so serious about improving cost competitiveness is that demands from equipment manufacturers for semiconductor vendors to lower their costs have been growing more strident. At the same time, capital requirements for product development, production line equipment and other items have been soaring. The only way that both of these problems can be addressed at once is to sell a lot of low-profit goods (Fig 2).
To achieve better cost competitiveness than their competitors are capable of achieving, they naturally have to aim for the top share of the market. Both firms are well aware of this. "We will hold a 40% share of the NAND Flash memory market in 2008, together with our partner SanDisk," said Toshiba's Muromachi; meanwhile Elpida Memory's Sakamoto predicts, "We will be the number one DRAM maker in the world by 2010."
At present, these strategies are paying off well for both firms, as recent market share trends in memory indicate. From the second quarter (April to June) of 2005 through the second quarter of 2006, Elpida Memory increased its share of the world DRAM market by 3.1 points, reaching 10.1%. In NAND Flash memory, Toshiba's share was 19.2% in the fourth quarter (October to December) of 2005, but jumped 5.5 points to reach 24.7% by the second quarter of 2006.

Other Fields Can Benefit
The strategy of leveraging technology through quantity, however, is effective in more than the semiconductor memory field. It should have the same effectiveness in logic integrated circuits (IC) and other semiconductors, and in equipment such as digital household appliances. In all these fields manufacturers face similar problems in plunging prices and soaring development cost.
The recovery of Toshiba and Elpida Memory indicates that even manufacturers with minimal share have a chance to cover, if they only take the appropriate action. This is the same reason why Advanced Micro Devices Inc (AMD) of the US is doing so well in the personal computer (PC) microprocessor market. In all these cases, the key to a successful effort was technology at the cutting edge of the industry combined with relatively large-scale production.
Unless a certain level of production can be achieved, however, this approach could also mean a firm could be driven from the market entirely. In fact, this well describes what has been happening to a number of Japanese semiconductor manufacturers involved in system-on-chip (SoC) products.

Capital Investment
One strong point that the recovering semiconductor manufacturers share is their ability to assure volume production. Elpida Memory and Toshiba are rapidly increasing investments into memory manufacturing equipment. In fiscal 2006, Toshiba will make a record-breaking investment of about Yen354 billion into semiconductor manufacturing equipment (Fig 3), of which 72% (about Yen255 billion) is earmarked for NAND Flash memory. Capital investment into facilities and equipment for NAND Flash memory four years ago, in 2002, was only a tenth of that or less, at Yen19.8 billion.
Epida Memory is increasing not only its own investments into equipment, but investments by the Si foundries to which it outsources are also increasing. In 2006 the firm will invest about Yen120 billion into equipment and facilities. Added to the investments made by the two silicon foundries the firm outsources to, the total reaches about Yen270 billion, which is roughly six times the total of 2002.
In early 2006 AMD started implementing a number of efforts to boost production capability. From 2005 through 2008 the company plans to double microprocessor production capacity.

Competitive Technology
Merely boosting investment into equipment and facilities, however, is no guarantee of competitiveness. Elpida Memory and Toshiba are enjoying strong results because they have outstanding technology. They were unable to properly leverage that technology in the past because they hadn't invested sufficient capital into equipment. Realizing this, they invested on a level to match that of major non-Japanese memory manufacturers.
Toshiba is about half a year ahead of Samsung Electronics Co Ltd of Korea in the volume production of 8-Gbit NAND Flash memory, the cutting-edge type (Fig 4). Their lead is mostly due to the fact that they were the first to put multi-value technology into commercial use, storing two bits of data per cell. At the end of 2005, Toshiba began volume production of 8-Gbit chips, combining 70nm-generation manufacturing technology with multi-bit design. Samsung Electronics, on the other hand, only began volume production of 8-Gbit chips in the second quarter of 2006, using 63nm-generation technology.
Elpida Memory's excellent performance is also due to technical superiority. A number of DRAM manufacturers are experiencing major drops in yield as they shift to design rules of 90nm or less. This has caused a shortage of DRAM, especially for personal computers, sending prices up and giving Elpida Memory a boost in terms of revenues, profits and more. According to the firm's Sakamoto, the yield of its 90nm memory is "high enough to be proud of anywhere." The Opteron and Athlon 64 microprocessors released by AMD in about 2003, for example, offered lower dissipation per unit operation than the competing Pentium 4 from Intel Corp of the US, helping AMD win considerable user support.

Clear Business Focus
These manufacturers improved their cost competitiveness by identifying clearly which fields to focus on, and concentrating their management resources (people, facilities and equipment, capital) into them.
Menu DRAM manufacturers are also producing NAND Flash memory, but Elpida Memory is sticking to DRAM. Since Toshiba pulled out of general-purpose DRAM, it has been concentrating on NAND Flash memory. AMD, suffering from losses incurred in NOR Flash memory and other devices, spun off Spansion LLC of the US in December 2005, establishing it as a non-consolidated company. By reducing the range of businesses covered, these firms have kept ahead of competitors trying to handle a wide product line-up, in terms of product development capability, fab productivity, and more.
By concentrating on a narrower range, even manufacturers with modest revenues can make equipment investments approaching the scale of the big manufacturers. Elpida Memory, for example, made almost all of its investments into DRAM, and Toshiba did the same for NAND Flash memory. Samsung Electronics, on the other hand, split investment capital between the two product sectors.
Assuming for a minute that Samsung Electronics had invested equal amounts into DRAM and NAND Flash memory in fiscal 2006, it is clear that the investment into either field would not be much different than the amounts invested separately by Elpida Memory and Toshiba.
If volume production is the key to cost competitiveness, then the manufacturer holding the top share enjoys the advantage. At present, though, top-class semiconductor manufacturers such as Intel and Samsung Electronics are making lackluster showings in performance, market share and other key indices. Part of the reason might be because they are trying to do too many things at once. They continue to operate in business sectors that are not too healthy, possibly eroding the cost competitiveness they should be enjoying in the fields they are strong in. Smaller manufacturers, however, can boost their shares by jumping into these sectors.

Major Price Cuts
An important advantage of strong cost competitiveness is that of price control. It is possible to drop prices sharply to kill off competition.
AMD was able to steal the 86-family microprocessor from Intel in exactly this way. In late July 2006, AMD cut prices on its microprocessors by 50% to 60% across the board. Only three days later Intel hurriedly followed suit, reducing its own prices by 50% to 60%, but clearly playing a defensive role to AMD's offense.
The danger of coming in second in a price war is pointed out by Willem P Roelandts, chairman of the board, president and CEO of Xilinx Inc of the US, the largest field-programmable gate array (FPGA) vendor: "I worked at Hewlett-Packard for thirty years, and realized that having products made unsaleable by competitors could spell death for the corporation. That's why here at Xilinx we constantly innovate new technology, releasing new products every two years to make the other guy's products obsolete." Xilinx holds over 50% of the global FPGA market, and enjoyed a high operating margin of about 20%
in the third quarter of 2006 (Fig 5a). When a manufacturer holding the largest share is the first to start slashing prices, in other words, they are able to take and hold the winning position.

SoC Manufacturers
Even in the system-on-chip (SoC) field, where each application can have a range of diverse products, it is crucial to leverage technology through volume. This can be clearly seen in the difference in performance by various SoC manufacturers; there is a significant difference between firms concentrating on volume production of one type of product, and those offering a range of products.
SoC manufacturers which are doing well command large market shares in ICs for use in equipment with massive shipment volume. One representative example is Texas Instruments Inc (TI) of the US, which essentially dominates the core of the mobile phone today. In 2005 the firm controlled about 45% of the global market for digital baseband processors for mobile phones, and about 69% for application processors. The market scale of the former in 2005 was about US$6.6 billion, and of the latter about US$839 million, which means that TI reaped revenues of about US$3.55 billion from this market in ICs for mobile phones. The firm has continued to grow rapidly since then, achieving an operating margin in the third quarter of 2006 of 24.7%.
In contrast with TI's success is NEC Electronics Corp of Japan, which has posted losses for six quarters now, from the second quarter (April to June) of 2005 through the third quarter (July to September) of 2006 (Fig 5b). Of the three major business lines the company is involved in, microcontrollers and discretes are in the black, leaving their SoC business awash in red ink.
SoC revenues for fiscal 2006 are thought to be about Yen260 to Yen270 billion, or between 62% and 65% of TI's income from mobile phone ICs. These numbers represent the total of sales to a wide range of application sectors including game systems, printers, digital videodisc (DVD) driver, digital video recorders, mobile phones and digital cameras. Given this situation, it is clear that the production volume for any particular chip will be lower than that enjoyed by TI. As a result, according to Toshio Nakajima, president and CEO of the firm, "We are unable to recover the massive capital investment we made. Most of our operating losses are from SoCs, and it is unlikely that our SoC business will turn a profit even in a year."

Japanese SoC Makers
NEC Electronics' problem - that there are few SoC products that it can expect to ship in large volume - is shared by other Japanese SoC vendors. As NEC Electronics' Nakajima pointed out, "We aren't satisfied with the way our SoC business is now, and plan some major streamlining after we identify sectors to concentrate on and those to cut." Until they focus better, a recovery for Japanese SoC manufacturers is unlikely.
Outside Japan, SoC manufacturers have already begun rationalizing operations. They are using external capital to pare their business sectors. Companies like Freescale Semiconductor, Inc of the US and NXP Semiconductors of the Netherlands, which was spun off from Koninklijke Philips Electronics NV of the Netherlands, have taken on investment funds as partners, no doubt with plans to separate the wheat from the chaff and concentrate on product sectors with promise. Japanese manufacturers will have a tough time surviving on the world market unless they implement similarly tough programs.
Some non-Japanese SoC manufacturers have already completed restructuring and are beginning to show solid results. Application-specific IC manufacturer LSI Logic Corp of the US, for example, posted a major loss in the third quarter of 2005 (Fig 5c). It sold off its fab in September that year, electing to become a fabless manufacturer. In March 2006 it withdrew from the structured ASIC business to concentrate on application-specific standard products (ASSP) primarily in the storage and digital household appliance sectors. Thanks to these measures, corporate performance has made a sharp recovery.

by Motoyuki Ooishi

Websites:
AMD: www.amd.com
Elpida Memory: www.elpida.com
Freescale: www.freescale.com
Intel: www.intel.com
LSI Logic: www.lsilogic.com
NEC Electronics: www.necel.com
NXP: www.nxp.com
Samsung: www.samsung.com
SanDisk: www.sandisk.com
Spansion: www.spansion.com
Texas Instruments: www.ti.com
Toshiba: www.toshiba.com
Xilinx: www.xilinx.com

MOST POPULAR ARTICLES

24 Hours | 7 Days