Commentary & analysis of week's chip news, April 16-20
Greetings from Down-East Maine where the weather has been acting like it does in my home state of Nebraska, where the old saw is: "If you don't like the weather, stick around for a few hours." Three days ago, we had a blizzard with six inches of snow and 50 mph winds. It was a mess and the plows worked all night to keep the roads open. Contrast that with today. The snow is long gone, the skies are clear, and the temperature is close to 60F. No budding yet, but I'm ready to start planting a few trees to replace those blown down in this past winter's storms.
News is so bad, market
must be near bottom
One thing about chip industry news these days, it's getting so bad that some observers now believe it's signaling we're close to a market bottom. Take the big foundries, for example.
Taiwan Semiconductor Manufacturing, world's biggest pure-play foundry, now says it expects weak global demand to push down its utilization rate to the lowest level in history. Chairman Morris Chang predicts TSMC's fab utilization rate will fall "very close to 50%"--possibly even a bit lower in the next few months. That would represent a sequential drop from 70% in the first quarter and from the 100% rate in the fourth quarter last year.
"Now it seems we've finally spotted this cycle's bottom," comments Chris Hsieh, senior chip analyst at ING Barings Securities in Taipei.
Chang agrees we're close. "I think the bottom will be reached in the April-May time frame," he predicts. "The recovery may not be a very swift and brisk one, but I do think the third quarter will be better than the second." And next year will be better than 2001, he believes.
Rick Hsu, Nomura Securities analyst, forecasts that 52% of TSMC's capacity would be used in the second quarter, climb to 58% in the third, and hit 76% in the December quarter.
First to pick up would be the PC and consumer electronics markets, analysts believe. Seasonal demand will drive up demand as early as August for the PC and October for consumer products. It will take a lot more time for communications equipment demand to come back, they add.
(See April 18 story.)
The picture is grim
at Chartered Semi
Still not convinced the chip industry has hit bottom because business is so bad? Then check out another one of the Big Three silicon foundries. Chartered Semiconductor came out Friday with first quarter results even worse than people had expected.
The big foundry ran into a brick wall in the first quarter. Foundry shipments declined sequentially by 31.9% to 166,400 8-inch equivalent wafers and 20.8% lower than year-ago numbers. As a result, net profits of the Singapore foundry shot into negative territory, hitting $30.9 million in losses.
CEO Barry Waite seemed dazed by current conditions. "There is growing evidence," he notes, "that the semiconductor industry is facing one of its most challenging years."
Chartered is now forecasting a 25% drop in second-quarter revenues from the first quarter and it expects its capacity utilization to plunge into the mid-40% range from 61% in the first quarter. That would be down from 94% capacity utilization in the fourth quarter and from a 104% rate just a year ago. That's really falling off a cliff! In the last downturn in '98, Chartered's capacity utilization rate fell only to 57%.
(See April 20 story.)
Intel does nearly as badly as
expected, but Street is happy
I still don't understand the stock market. And I've been watching it for a heckuva long time. Take the world's largest chip maker this week. Intel had--for it--a lousy first quarter. Net profits went off the cliff, diving 82% from the year-ago quarter and 78% from the fourth quarter of 2000. But these results were not unexpected by Wall Street, so the chip giant ended up leading a stock market rally on April 18.
First quarter sales fell 16% from a year-ago to $6.7 billion and 23% from the previous quarter. Net income was $1.1 billion, excluding acquisition-related costs, and $485 million when they were included.
But Intel did have some good news. "Our microprocessor business appears to have stabilized and we expect to see normal seasonal patterns going forward from our current business level," declares CEO Craig R. Barrett. Sales for the second quarter should range between $6.2 billion to $6.8 billion, or anywhere from a 2% rise to a 7% drop from a year ago.
(See April 17 story.)
Intel breaks out results
and results aren't pretty
Outside the microprocessor business, Intel ain't all that good. The world's largest chip maker picked a bad time to break out its financial results for the first time. It lost a cool $1 billion in the first quarter in its communications and emerging chip sectors. That compares with a $452 million operating loss a year ago and a $657 million loss in the fourth quarter of 2000.
First quarter sales hit $1.5 billion in these segments, which includes wireless communications and computing, communications products, network communications, and new business. This was up from sales of $1.3 billion in the first quarter a year ago.
On the other hand, Intel's largest business--the Intel Architecture Group--reported an operating profit of $1.7 billion on sales of $5.1 billion in the first quarter. The Intel Architecture Group, which includes microprocessors, motherboards, and chip sets, reported an operating profit of $1.7 billion on sales of $5.1 billion. However, this means operating profits fell by nearly 50% and sales by one-quarter from the preceding and year-ago quarters. That's downright nasty.
(See April 18 story.)
TI accelerates cost cutting as
sales continue to deteriorate
Texas Instruments contributed its share to the mountain of bad news this week. The Dallas firm went to "Defcon Three" and confirmed rumors that it was laying off 6% of its work force--eliminating 2,500 jobs as part of a stepped-up, cost-cutting program.
What triggered all the new belt tightening, of course, was sales that continued to decline rapidly. TI reported a 19% sequential drop in first-quarter chip revenues to $2.18 billion, while operating profits from semiconductor operations dropped in half to $304 million from the first quarter last year.
It doesn't look any better for the second quarter. TI expects revenues to decline sequentially by 20% from the first quarter due to continued weakness across nearly all semiconductor segments. "We are in one of the sharpest decelerations that our industry has experienced, and it requires that we move fast to make hard but prudent business decisions," says CEO Tom Engibous.
Once again, TI is trimming back capital spending. TI now plans to layout $1.8 billion in capital spending, down from $2 billion set in late February. The company originally budgeted $2.3 billion, down from $2.8 billion in 2000. R&D spending is getting trimmed as well. TI will also lower its R&D spending to $1.6 billion in 2001.
(See April 17 story.)
AMD does better than
expected, thanks to MPUs
Jerry Sanders was a happy man this week as he reported Advanced Micro Devices did better than analysts had expected in the first quarter. While the veteran CEO says the second quarter will be the "toughest quarter" this year with sales down as much as 10%, he expects the second half of the year to be better than the first.
AMD did better than most chip makers in the first quarter, thanks to its excellent performance in microprocessors. Sales were up 9% from the fourth quarter due to strong sales of Athlon and Duron processors.
The chip maker was able to improve its market share the microprocessors. According to one estimate, its market share jumped from 17% in the fourth quarter to 21% in the first quarter.
It was record MPU sales in the first quarter than pushed AMD revenues up 1% from the fourth quarter to $1.19 billion and up 9% from the year-ago total.
The company sold more than 7.3 million PC processors, including 6.5 million Athlon and Duron processors in the first quarter, pushing PC revenues 17% higher than fourth quarter and year-ago sales to $661 million. AMD expects PC processor sales to remain at "near-record levels" in the second quarter, which is a seasonally slow period. For the year, AMD is forecasting "high single-digit" growth for the PC unit shipments.
Flash memory sales were a different story, falling 10% in the first quarter from the previous quarter. And it will be flash that could push down AMD sales in the second quarter. "Sales of flash memory products, reflecting a sharp decline in demand from the communications sector of the economy, declined by approximately 10% sequentially to $411 million in the first quarter," Sanders says.
(See April 18 story.)
STMicro: market correction
to continue for rest of year
My feeling is that STMicroelectonics is sailing through this chip downturn just about as well as anyone. The European chip maker checked in this week with better year-to-year profits and revenues for the first quarter, but acknowledges the weak market is pulling down sales.
The firm showed a 12% sequential drop in net revenues to $1.92 billion in the first quarter. Sales in the second quarter are now expected to drop 6-to-14% from first quarter revenues. Net income in the first quarter totaled $340.8 million, up 43% from the same period last year. Revenues were nearly 13% higher than the first quarter in 2000.
"The market correction, which began abruptly with a sharp inventory adjustment in the 2000 fourth quarter, is likely to continue through much of this year," says CEO Pasquale Pistorio. "Its duration is closely tied to macroeconomic conditions, particularly in the U.S. and Japan, as well as to industry-specific issues such as over-capacity and excess inventory levels."
As a result, STM now expects its second quarter revenues to fall below last year's $1.88 billion to a range of $1.65-to-$1.8 billion.
The strongest year-to-year growth in chip shipments in the first quarter came in telecom applications, jumping 36%. Chips for automotive applications were up 12%; for computers up 8%; and for industrial products up 17%. Revenues from digital consumer products were down 11% from a year ago due to excess inventories and a sharp decline in consumer demand.
(See April 19 story.)
still looks like a go
It would surprise me now if the sale of Silicon Valley Group to Holland's ASM Lithography is blocked by the U.S.
Despite the current push to derail the deal by competitors and other business groups, such major factors as Intel and the SIA aren't budging in their support. That's despite a controversial videotape that shows the former president of SVG Lithography, Edward A. Dohring, and David Markle, CTO of Ultratech Stepper, speaking out against the deal because of national security concerns.
Also, the Bush administration late this week was still trying to reach a consensus among U.S. agencies on the proposed purchase, but at the start of the weekend, no official position had been taken on ASML's takeover of SVG. It remains unclear if this could end up delaying the deal by a few more weeks or even a month.
But Intel's position has said all along that the merger will be the best way to keep leading-edge lithography in the U.S. "This is the right thing to do," Jim Jarrett, Intel's vice president of government affairs, tells Semiconductor Business News.
The video tape is being distributed by the U.S. Business and Industry Council, a non-profit business group with 1,500 members. More than 600 copies of the tape have been sent to Congressional leaders and U.S. government officials to try and prevent ASML from buying SVG.
The proposed merger has been under a 45-day government review that expires on Monday, April 23.
(See April 19 story.)
TSMC claims to be first
with 0.10-micron modules
The silicon foundry leader, Taiwan Semiconductor Manufacturing, is making a full court press in a public relations drive to convince the market place that it is--or shortly will be--a global process technology leader. Trying to keep up with the flood of releases can be a trial, but it does look like TSMC is making progress here.
In its latest progress report this week, the Taiwanese chip maker claims it is the first company to complete the development of the basic modules for 0.10-micron CMOS logic processes. It will now concentrate on working with leading-edge customers to push the technology into production by the third quarter of 2002.
TSMC is pushing hard to get its 0.10-micron process into production at about the same time as leading IDTs (integrated device manufacturers)--Intel, IBM, and Texas Instruments.
It also says that it has now achieved "reasonably good yields" in 300-mm pilot production of 0.13-micron 4-megabit SRAM test chips, which were processed with all copper interconnects. TSMC is following up the early April pilot run by testing a few customer designs in the 0.13-micron copper process on 300-mm diameter wafers this month.
TSMC is now building two dedicated 300-mm production facilities--Fab12 in Hsinchu and Fab14 in Tainan. Fab12 should enter early prototype production in the fourth quarter while Fab14 cleanroom construction should be finished by the end of the year.
(See April 18 story.)
These guys aren't giving up
on x-ray for next-gen litho
With the big guys making a lot of noise these days about accelerating the development of extreme ultraviolet (EUV) as the technology for the next-generation lithography system, one little guy laid out plans this week to compete with them.
It looks a bit dicey, but JMAR Technologies plans to acquire Semiconductor Advanced Lithography Inc. (SAL), a supplier of x-ray stepper systems in a strategy to offer next-generation litho systems. The San Diego company says that Vermont-based SAL already has produced X-ray lithography tools that will be able to pattern circuit features below 100 nanometers (0.10 micron). JMAR plans to integrate its own laser plasma x-ray sources with SAL's steppers in an approach it claims will be cheaper and available sooner than EUV-based systems.
JMAR estimates that these x-ray lithography systems will sell for $6-to-$10 million each, which it claims will be much lower than EUV systems currently under development by the U.S. EUV consortium and in Japan by other lithography suppliers.
JMAR says its x-ray lithography system will offer several advantages over other exposure methods. The big one will be the ability to leverage existing proximity mask technology and use relatively inexpensive collimators to intensify and direct the short wavelength x-rays onto mask/wafer targets. This approach avoids the need for the extremely complex and expensive focusing optics that are required for all other types of advanced optical lithography techniques, including deep ultraviolet (DUV) and EUV, JMAR claims.
(See April 17 story.)
Intel-Analog Devices team
makes progress with MSA
Intel demonstrated the first working silicon for its Micro Signal Architecture (MSA) in Tokyo this week.
Its first evaluation board ran at 170 MHz--more than fast enough to handle emerging 2.5G applications--but Intel already has run the core at more than 400 MHz at 1.5 volts in the lab. "We didn't want to stop at 170 MHz because we wanted to create headroom for 3G--cell operators are banking on a lot of data," says Ron Smith, senior vice president.
The design team, which includes engineers from Intel and its codevelopment partner Analog Devices, is concentrating on power consumption at a given performance level since MSA's primary focus will be on the wireless-handset.
But the most important speed feature of the architecture could be the time-to-market edge it promises. Because MSA is optimized for C/C++ programming, Smith claims the architecture could collapse months or years of hand-packing coding into days or weeks. "This means that other DSP technology is pretty ancient," he adds.
Intel is playing coy about when a product will be equipped with the architecture. Smith won't say when Intel will integrate the core with logic or when it will ship DSPs based on the architecture. He would say only that the chip maker will make an announcement on the subject later this year. It's still a development project, Intel says, since the silicon is basically a test chip for the core.
(See April 16 story.)
Infineon, IBM aim to sell
giant magnetic RAM by '04
Infineon Technologies and IBM Microelectronics are partnering in an ambitious attempt to bring out a next-generation memory. They plan to make and market a 256-megabit class magnetic RAM by 2004.
The first products are expected to be made with a 0.13-micron process at their Altis Semiconductor joint venture in Corbeil-Essonnes, France. They are going for this size memory because they need a reasonable size--a demonstrator that could become a lead product to debug the process.
This size MRAM is in marked contrast to Motorola Semiconductor, the other chip maker that has committed to an magnetic RAM introduction. It probably will produce a 4-megabit MRAM with a 0.2-micron process technology and is aiming for samples of the smaller DRAM by 2003.
Infineon also is backing another non-volatile technology, ferroelectric RAM, in its drive to become a major player in next-generation memory markets. It has partnered with Toshiba in a program to bring a 32-megabit FRAM to market by 2003. But the introduction of this FRAM has been complicated by Toshiba's decision to make the parts in a proprietary process at a Yokohama wafer fab. The two companies are now trying to figure out how this technology would transfer to Infineon and whether the German company would buy part of the Yokohama output.
(See April 17 story.)
What kind of nut would
enter foundry business now?
You have to wonder about people who are starting up a silicon foundry in a business climate where the big guys are seeing their capacity utilization figures shrink by nearly two-thirds from year ago numbers.
Then again, turning on a production line or investing in a company when the overall economic picture is as bad as it is now could turn out to be exactly the right thing to do.
South Korea's Dongbu Electronics is one such gambler. Why now? For one thing, the Seoul-based unit of the Dongbu Group wants to become the next major force in the foundry business. It is now processing test wafers in its new 200-mm wafer fab and plans to start volume production in May. It got the 0.25-micron process technology from Japan's Toshiba, which has invested $50 million in the fab.
Dongbu's chip market entry comes after several false starts. In the late '90s, it made noises about entering the gallium-arsenide business, but scrapped those plans. Then it decided to enter the DRAM business and signed a deal with IBM, then dropped that plan after the '98 downturn. Last year, it told the world it wanted to be a big foundry, but has kept a low profile, which led many observers to conclude that the company had another change of heart.
But this time, its market entry seems real. "We plan to make a run at TSMC and UMC (The two largest foundries)," boasts Peter Hillen, who heads up Dongbu's new U.S. subsidiary. Before that happens, though, Dongbu is sighting-in on Charter, the No. 3 foundry. He certainly has plenty of competition. The Korean firm is just one in a wave of untested foundries entering the market at a time when demand has almost disappeared.
(See April 19 story.)
Researchers can't agree
on EDA outlook this year
The big argument this week was over the outlook for the EDA industry. Some observers fear they're seeing the first signs of an EDA downturn and that a protracted semiconductor industry downturn will cut into EDA product purchases as well this year.
As a result, analysts now see EDA industry revenue this year hitting the neighborhood of a 10-to-12% increase, a long way from the 20%-plus growth most analysts were touting at the start of this year.
But Dataquest says they are all wet. Gary Smith, chief EDA analyst at the market researcher, says his firm is sticking with its original 20% EDA growth estimate for 2001. He says that in a group of 30 or 40 "power users," Dataquest found only two that plan to cut EDA expenditures this year.
Cadence, the largest EDA vendor, reported first-quarter revenue of $345 million, a 34% year-to-year rise but down 12% sequentially from the fourth quarter. But revenue from services, including chip and product design support, was up only 6%on a year-to-year basis. Tality, which represents slightly over half of Cadence's overall services revenue, was down 20% from the fourth quarter, with $44 million in revenue.
(See April 19 story.)
Like Avis, the new No. 2
in wafers, is trying harder
What's a fast way to become the world's largest blank wafer supplier? Merge the world's fourth and fifth largest producers to become No. 2 and then try harder. That's what Sumitomo Metal Industries and Mitsubishi Materials are doing.
The two Japanese companies have been negotiating a merger of Sitix, a Sumitomo division, and Mitsubishi Materials Silicon, a wholly-owned subsidiary of Mitsubishi, ever since the two jointly established a 300-mm wafer company named Silicon United Manufacturing (Sumco) in July 1999.
Sumitomo is the fourth-largest wafer supplier with a 12.8% market share and Mitsubishi fifth-largest with a 10.5% share in 1999, according to Dataquest. The combined operations will create the world's second-largest wafer supplier, behind Shin-Etsu Handotai, which held a 24.7% market share, Dataquest figures.
They have agreed to reorganize their domestic and overseas wafer production operations under Sumco. Sumitomo has three factories in Japan, four in the U. S. and one in France-accounting for $920 million in sales last year. Mitsubishi has four plants in Japan, one in the U.S. and one in Indonesia, which pulled in $800 million in sales.
Sumco will be reorganized under a new name on Jan. 1, 2002 and will manage the consolidated operations, as well as the companies' previous R&D and production work on 300-mm wafers. Sumco already has invested $240 million on 300-mm wafer production and is building new lines for ingot production at Mitsubishi Material's factory in Yonezawa and a line for polishing and slicing wafers at Sumitomo's Imari factory.
(See April 19 story.)
If you have any comments, criticisms, or questions, don't hesitate to E-mail us at email@example.com. Have a great weekend!
(Click here for last week's Semiconductor Alert!.)