By Vladimir Romanchenko (firstname.lastname@example.org)
Today's forecasting in the semiconductor industry is much more modest and cautious: independent analysts and experts at companies speak about implementation of some or other technologies much more carefully. The economic slump of 2001 - 2002 had a great effect on forecasting. As a result, starting at least from the end of 2002 most analytic researches have been much weightier than before. However, by the end of 2003 such materials got some optimistic signs.
Right before the two-year depression, by the end of 2000 the analysts poured out generous figures; at the end of 2001 we got rollicking reports about the end of the slump and beginning of the upturn. In the middle of 2002 all such reports were junked when the number of redundant workers exceeded all possible limits and it was clear that the financial data of most companies would be far different from the forecasts made in Q1. If you look through the roadmaps of the semiconductor leaders drawn two-three years ago you'll be surprised how optimistic they were. In 2000, the year of rapid economic growth, everything was simple and clear. Who could imagine that the industry development in 2001 would significantly slow down and the year 2002 would be spent for recovery?
That is why in spite of the sign of the economic upturn at the beginning of 2003 most analytic companies made modest forecasts. Well, it was justified: the second Iraqi military campaign and the threat of SARS (Severe Acute Respiratory Syndrome) froze the success made early in 2003.
The last year ended well, but there was one more peak of the Near-Eastern crisis that poured into the military actions in Iraq. The semiconductor industry as a part of the world economy was also influenced by those events. It wasn't only the drop of stock-exchange indices or weakened interest toward this industry: in some cases the companies had to put off the terms of release of new chips and even change the traditional routes of deliveries of finished products.
By the middle of the year the situation was bettering except the strange correlation between the top world currencies. Moreover, the Q3 results in the semiconductor industry and in most other world industries let the analysts speak about the serious upturn and even realistic prospects of further growth. The factories in the semicnductor sector were loaded by 88.7% in Q3 and even exceeded 92% in October, which was a good sign of the economic recovery and of the need to build new production facilities. Moreover, the 85% point was exceeded for the first time since 1994 and it was a clear record since Q3 2001.
Last year we were mostly focused on the transition to 300mm silicon wafers and development of the new fab processes. The shift from 0.18 micron technology to 0.13 micron was completed by the most leaders yet in the first six months; moreover, by the year-end the 0.13 micron technology was spoken about as a thing of the past, and the 90nm technology was at the forefront. However, taking into account the problems with implementation of the 0.13 micron process the analysts are not such optimistic about the 90nm norms because the gap between the sample production and beginning of the mass production of the chips increased. It's easier for those who invested a lot into the technology development and didn't give up even in the heavy periods. Just look at Intel: the company spent a great deal of money for R&D regarding the new fab processes, and in 2003 they even cut down the expenses.
The transition to the mass production of chips based on 300mm silicon wafers from 200mm ones looks more complicated. It's very profitable but taking into account even the most optimistic profits of $150 for one squared inch of the wafer, one 300mm silicon wafer would cost about $20,000. So, the company which decides to start production of 300mm wafers with the monthly output of 25,000 pcs will have to sell semiconductors for at least $4.5 billion to justify erection of such factory. Another condition for payback is the full monthly load of all the production lines. At the same time, it's so expensive to build such lines that some companies have to give up this idea. Other manufacturers (especially Japanese) are thinking about collaboration with their competitors to build a shared factory for 300mm chip production. Also, it's more beneficial for some to place orders for such chips with others. Finally, to provide the 100% load of production lines some firms offer their fab processes for third-party orders.
So, the transition from the 0.18 micron norms to 0.13 wasn't that ruinous as some of the companies affirmed. At the same time, the number of factories producing 300mm wafers didn't increase much in 2003. The decreased sales of semiconductors in 2001 - 2002 and the reduced expenses for the fundamental reequipment, in particular, for purchasing modern lithographic tools, made the companies cut down the expenses for modernization and R&D in some cases.
According to Strategic Marketing Associates (SMA), at the end of 2003 there
were only 25 factories able to produce 300mm wafers with the total
output of 400,000 wafers per month. The largest supplier of 300mm
wafers is UMC; it's followed by Intel, Samsung, Powerchip and Texas
Instruments. Taiwan and USA control up to 55% of 300mm silicon wafers
production, and more than half of 300mm lines is loaded by DRAM production.
It's not that easy to combine in one review the most characteristic last-year trends, you can always miss something important. But even if you look at this material as a selection of some interesting industrial events you will get a structure that forms the future of the semiconductor industry.
The final statistical data for 2003 are still being prepared for publication, but we can say that the traditional spheres of semiconductor consumption are going to have the most considerable growth: consumer electronics, mobile communication, notebook PCs, desktop PCs, wireless LAN; the manufacturers of wired communication devices and blade servers are less lucky. According to the preliminary data, the world semiconductor market grew up by 14%. It's interesting that three European companies got into Top Ten 2003 - but it happened not because of the growth in delivery by because of the growth of Euro relative to Dollar, by 19%.
The first position is still taken by Intel with the expected sales of $27 billion ($24 billion last year.). Samsung and then Renesas (Hitachi- Mitsubishi) follow it.
Texas Instruments and Toshiba take the fourth and fifth positions - their sales increased by 23% in 2003; STMicroelectronics is the sixth - its sales growth was 12%. Infineon got the greatest growth - 28% benefiting from the growth of prices for DRAM and rate of Euro but the company takes only the seventh position.
NEC with $6.2 billion sales goes next. TSMC with $5.9 billion takes the second places in sales growth and 9th place in the summary table. The next one is Philips with $5.6 billion sales and the 28% growth (like that of Infineon).
Motorola that was on the 9th place in 2002 will probably drop out of the top ten judging by the statistics of 2003.
Speaking about the last-year trends on the market of finished computer systems, I should say that the delivery of notebook PCs was the most significant since more and more business and home users buy mobile devices with PC, communication and entertainment functions. The demand for smartphones also grew up by the same reasons, but still, most people tend to get a notebook for this purpose.
According to the analytical calculations based on the data received from the Taiwanese manufacturers the deliveries of 15" notebooks exceeded 50% of the total portable PC sales long before the end of 2003. One of the reasons was the growth of popularity of portable PCs in the consumer goods sector (in contrast to the corporate sector). So, the correlation in sales of 14.1" and 15" notebooks in the consumer sector is predicted for 2004 as 2:8.
At the same time, the prices for many models of Intel Centrino based notebook PCs significantly dropped down in Q4. Earlier, in Q1 the supplies of notebooks in USA exceeded the sales volumes of desktop PCs from the financial standpoint. That is why we can admit now that the PC market is rapidly shifting to the mobile footing.
The latest reports from the CPU market didn't deliver precise data on the volumes of CPUs supplied from different manufacturers. Well, Intel remains the unbeatable leader, but AMD managed to prove its power to compete in all market sectors. However, it should have its own competitive mainstream processor vs Pentium 4, and continue using the carefully developed system of CPU marking.
Unfortunately, Transmeta and VIA Technologies will probably remain niche players. In 2002 VIA sold 2 million processors, and the preliminary forecast for 2003 is not more than 6 million chips. Nevertheless, the combined marketshare of VIA and Transmeta in 2003 was only 1.6%.
It's expected that during 2003 - 2004 VIA's marketshare won't change, and Transmeta's one remains uncertain because of the transition from TM5800 line to TM8000 (Efficeon) - and it will depend on the rate of CPU production. The key problem for VIA in the future will be the problem of FSB in 2005; the key difficulty for Transmeta will be implementation of the socket for TM8000 chips.
By the way, in August we lost one player of the x86 CPU market: National Semiconductor sold its Information Appliance department to AMD, which developed Geode chips for inexpensive network terminals, desktop consoles and other similar devices. With Information Appliance AMD got such National Semiconductor's clients as Samsung and NEC.
On the memory market we could witness the shrinking gap between the prices for DDR333 and DDR400 chips and shortage of NAND flash. As a result, the price wars among two leaders of NAND flash production - Toshiba and Samsung Electronics - were put an end to. Besides, the volumes of supplies dropped down because of the started modernization of the production lines and simultaneous transition of three leading manufacturers - Samsung, Toshiba and Renesas - to higher-density chips; and it also affected shortage of the chips.
The world market of graphics cards shrank in the second half of 2003 mostly because of the growing popularity of integrated solutions. NVIDIA remains the leader in sales of discrete graphics processors, ATI dominates in the notebook graphics sector. The profits for 2004 are very vague, especially because of the transition to the new PCI Express 16x bus and because of such new-comers as S3 Graphics and XGi Tech.
However, it's better to discuss trends of certain markets in the respective sections. In closing let me give you some overall figures. According to the preliminary data (iSuppli), the world semiconductor market ran into $178.4 billion in 2003 which exceeds the results of 2003 by 13.9%. At the same time, the results of the second half-year exceeded those of the first six months by 18% thanks to the absence of wars and stock-exchange fevers!
This year, the industrial sales are expected at the level of $208.8
billion, which will be greater by 17%. In 2005 the market is expected to
grow up by 12.9% up to $235.7 billion. And only in 2006 the industry might
face a new collapse when the supplies will increase only by 1.1% to make
$238.4 billion. The most drastic results in 2004 will be obtained by manufacturers
of PCs, mobile devices, flash memory and DRAM.
Write a comment below. No registration needed!