Happy Holidays! Silicon Strategies has compiled a list of 15 predictions for the IC and chip-equipment industries in 2005 and beyond. The predictions come from crystal balls at Silicon Strategies and various analysts.
1) No growth for ICs in '05
Mark LaPedus, editor of Silicon Strategies, is sticking to his forecast of 22.7 percent growth for the semiconductor industry in 2004. Look for zero percent growth in 2005.
LaPedus also sees 50 percent growth in the IC equipment market in 2004, but a 20 percent decline in 2005. Overall, it's going to be a big step backwards for the IC industry in '05. IC makers are still digesting their fab capacities amid the current lull.
2) Inventory blues
Paul Leming, an analyst with Princeton Tech Research (Princeton Junction, N.J.), also sees zero growth for ICs despite an 8 percent jump in unit shipments for 2005. Leming also believes that the current inventory correction problem will subside in the first quarter of 2005, or early in the second period.
3) Negative territory for CapEx
Mark Bachman, an analyst with Pacific Crest Securities Inc. (Portland, Ore.), also sees a bad year for fab equipment, as capital spending will fall next year. "Just as a point of reference, my CapEx forecast is now down 12 percent year-over-year for 2005," he said.
4) 300-mm or bust
Christian Gregor Dieseldorff, an analyst with Strategic Marketing Associates (Santa Cruz, Calif.), predicts a 64 percent jump in 300-mm fab capacity over the next two years. The analyst also sees 26 new 300-mm fabs up and running over the next two years.
At present, Strategic Marketing Associates lists 37 300-mm fabs that are "on-line." About one-third are for memory chips, less than one-third are for foundry and about one-fourth are geared for logic ICs.
"The 300-mm fabs with the largest capacities are those owned by Powerchip, UMC, Samsung and Infineon," Dieseldorff said. "However, Inotera is ramping up a 300-mm fab with a full capacity of 54,000 (wafers per month), which brings them head-on-head with TSMC."
Indeed, 300-mm is the name of the game in the IC industry. "In the next coming two years, we see an increase of capacity of 64 percent," he said. "We list 26 fabs currently ramping up (constructing or equipping). These fabs will start coming online in 2005. About 10 will reach production phase in 2005 and the rest in 2006."
5) 300-mm in China?
Dieseldorff also sees more 300-mm fab activity in China. "Grace is planning a 300-mm fab in China and SMIC is building a fab right now," he said. "We see more partnering happening and more 300-mm activity in China such as the Hynix-STMicro joint venture in China. Freescale and Philips are also looking for a partnered 300-mm fab."
6) Hynix to upset NAND market
Nam Hyung Kim, a principal analyst with iSuppli Corp. (El Segundo, Calif.), believes that the new NAND flash-memory players will gain significant market share in 2005 -- at the expense of Renesas Technology Inc. and others.
In Q3 of 2004, Samsung was the leader in the NAND market with 54.2 percent share, followed by Toshiba (29.2 percent), Renesas (9 percent), Hynix (5.1 percent), Infineon (0.6 percent) and STMicroelectronics (0.4 percent).
The new players will gain market share, "especially Hynix at the expense of Renesas," Kim said. "Toshiba and Samsung will continue to dominate." Also look for a major thrust by Micron Technology Inc.
7) Samsung to dominate NOR, NAND markets
Intel and Spansion are fighting it out in terms of market share in the NOR flash market. Intel overtook Spansion in the NOR market in terms of share in Q3 of 2004. For 2004, Spansion will most likely remain the leading supplier of NOR parts in terms of overall market share, according to iSuppli's Kim.
However, LaPedus of Silicon Strategies predicts that neither Intel or Spansion will end up the NOR leaders in 2005. He predicts Samsung will end up leading the NOR market, along with NAND, in share. Sorry Intel.
8) MRAM is a non-event
LaPedus of Silicon Strategies also believes that the next-generation nonvolatile memory market will be a bust in 2005. Altis, IBM, Freescale, Infineon, and others have announced MRAMs, but shipments will amount to nearly zero in '05. Other technologies -- including FRAM and OUM -- will not fair much better.
The problem is simple. The newfangled memories are hard to manufacturer. Traditional NAND and NOR flash will remain kings of the hill for some time.
9) DVD chip outlook
From 2004 to 2008, IDC expects total DVD semiconductor player and recorder revenue to have a CAGR of 11 percent and to reach nearly $3.7 billion in 2008, according to IDC.
DVD player semiconductor revenue is expected to experience a CAGR of minus 21 percent and to have revenues of $637.8 million in 2008. DVD recorder semiconductor revenue is expected to achieve a CAGR of 39.1 percent and reach $3 billion in 2008, according to the firm.
10) Foundry marriages
Something has got to give in the foundry market. There are too many players doing essentially the same thing. And there is a huge gulf between the "have" and "have nots." The "haves" own 300-mm fabs; the "have nots" do not.
There are just too many "have nots" in the market. So look for mergers between the "have nots." For example, LaPedus of Silicon Strategies predicts the two Malaysian foundries -- 1st Silicon (Malaysia) Sdn. Bhd. and Silterra Malaysia Sdn. Bhd. -- will merge within the next two years. The two companies have talked off and on for years about merging, but look for more action and less talk.
We could also see a similar merger between two foundry hopefuls in Korea. Within the next two years, MagnaChip Semiconductor Ltd. could also end up buying DongbuAnam Semiconductor Ltd. MagnaChip, the former non-memory operations of Hynix Semiconductor Inc., is not talking to DongbuAnam right now. MagnaChip may be waiting for DongbuAnam's valuation to drop before it starts kicking tires.
However, DongbuAnam's parent company, steel maker Dongbu, may stay in the foundry game. Steel is selling well in China, keeping the foundry business under the radar.
11) Big Blue goes fabless
By the end of this decade, IBM Corp. could sell its Microelectronics Division; it's simply a non-core business and a drain on resources.
IBM may end up selling its IC unit (and fabs) and form a joint venture with chip partner Chartered Semiconductor Manufacturing Pte. Ltd., a Singaporean foundry provider. Maybe Samsung Electronics Co. Ltd. will also become a partner.
12) Stock option blues
Investment banking firm SG Cowen Securities Corp. (New York) sees little benefit in a new accounting rule, which makes companies expense stock options. The rule takes affect 6/15/2005.
In a study, the medium reduction in net income among 17 chip companies would be 36 percent in 2004 -- if the accounting rule was hypothetically enforced. On top of sluggish demand, look for lower earnings for IC makers in '05.
Leming of Princeton Tech Research agrees. "We believe the expensing of stock options is questionable accounting, and prefer to analyze the 'cost' of stock option programs from a cash flow perspective," Leming said. "We have looked at the year-by-year cash flow expenditures necessary to offset the dilution from options in our coverage universe over the last four years," he said.
"The imputed cost of offsetting the dilution from options has been equal to more than 100 percent of Maxim's cash flow from operations -- and 50 percent to 60 percent of cash flow from operations at Linear Technology, Microchip and Analog Devices -- over the last four years . We believe Maxim's continued agressive issuance of stock options is, and will continue to, negatively affect the valuation of its shares."
13) Silicon wafer prices stabilize
With 8-to-9 percent unit growth and price declines of no more than 1-to-2 percent, the silicon wafer industry will grow by 7 percent a year over the next decade, according to Leming.
Prices for silicon wafers will decline at a slower pace than the annual average of 12 percent per year, he said. Current 300-mm silicon pricing is expected to remain around $2.25 per square inch -- or $250 per wafer, he said.
14) Delays for 45-nm node
Immersion lithography may make it in time for the 45-nm node, but the industry will continue to struggle to find a decent high-k solution. ALD and MOCVD will not be ready for prime time, or, for that matter, the production of high-k materials.
Metal gates, FUSI, and other technologies look good on paper, but they will be hard to deploy. Low-k beyond 2.5 will also remain a moving target.
Translation: Don't expect the 45-nm node on time. Some were hoping to deploy the technology in 2007. Silicon Strategies expects the 45-nm node to happen in the latter part of 2008.
By the way, immersion lithography will continue to make strides in 2005, but don't look for scanners based on the technology to move into volume production until 2006, as previously expected, according to Silicon Strategies. But look for more delays and setbacks for EUV lithography, which may struggle to hit the 32-nm node by 2009.
15) Top dogs
Silicon Strategies presents its predications for the top-10 chip makers in 2010, based on worldwide sales (and including the foundries):
1. Samsung -- A juggernaut
2. TSMC-UMC -- Could merge by '10
3. Intel -- Not as important
4. AMD-IBM-Chartered -- Big merger seen
5. Renesas-Toshiba -- More consolidation in Japan
6. ST-Freescale -- Could happen in '05
7. Philips-Infineon-Micron -- Mega-merger
8. SMIC-Grace -- China powerhouse
9. TI-ADI -- Say it ain't so
10. Broadcom-Qualcomm -- Merger creates fabless giant
Here's a prediction for the top-10 chip-equipment makers in 2010:
1. Applied Materials-ASML-ASMI -- Mega-merger
2. TEL -- May seek acquisitions
3. Canon-Nikon -- Big litho merger
4. Novellus-Lam-Varian -- Likely trio
5. KLA-Tencor-Therma-Wave-Rudolph -- Metrology mergers
6. Agilent-Teradyne -- ATE mega-merger
7. Advantest- Yokogawa-LTX -- See # 6
8. Dainippon Screen-Dainippon Printing -- Natural fit
9. Hitachi -- Mystery company
10.Intel -- Will make own EUV litho equipment