Market moves toward balance
By James Carbone -- Purchasing, 5/20/1999
Nineteen-ninety-nine is shaping up to be a year of recovery for the electronics industry. After three years of falling prices due to oversupply, supply and demand are becoming more in balance, and the effects already are being felt by electronics buyers.Prices for many semiconductors in the first quarter were stabilizing. In fact, tags for memory chips rose, although that was expected to be short term. Production cutbacks, the exiting of some chip companies from the memory business, and cutbacks in capital investments mean that supply and demand are equalizing.
At the same time, demand continues to increase. Shipments of computers and communication equipment will see double-digit growth this year. Communication equipment, including networking equipment and cellular phones, is forecast to grow 20%. The PC industry should see around an 11% overall increase in unit shipments in 1999. Tony Massimini, chief of technology with Semico Research, says that desktop PC shipments will fall below 10% per year. But notebook shipments will be up 15%, and workstations and high-end servers will be in the 20% range.
But if the economies of Asia and Japan recover in 1999, there is potential for stellar growth. Some chip makers say Y2K concerns could have a big impact on sales in the second and third quarter. They say many companies that have ignored the Y2K issue will buy new compliant systems later in the year, and there is potential for a huge upside in demand for components that would mark the beginning of a new boom in semiconductor demand.
Shortages ahead?
If all these factors come into play, there is potential for shortages in the third and fourth quarters because semiconductor manufacturers have seriously cut back capital investments. "There's a possibility of inventory builds with the Y2K issue," says Bill McClean, president of IC Insights, a market researcher. Parts that would be vulnerable to shortages are leading-edge parts built on .18 micron process technology and higher performing 64-Mb drams in different packages and organization.
"The market could turn rapidly," says McClean. "In 1995 the market turned rapidly from undersupply to oversupply. The same thing can happen because everything is moving to .18 micron. Many companies need parts made on .18 micron."
However, McClean and other analysts don't see a boom in semiconductor sales growth in 1999, but they forecast a healthy double-digit increase. The semiconductor market is forecast to grow 11% to $140 billion after an 8% decline in 1998, according to market researcher IC Insights. While that's good news for chip makers, it's still below the $144-billion mark the industry achieved in 1995. Passives sales will grow a modest 4.5%, and worldwide connector demand is forecast to grow a paltry 1.4%.
Expect downward price pressure
But while supply and demand may become more in balance, buyers can still expect modest downward pressure on many components for the rest of the year. However, price declines won't be as steep as in previous years. For passives and electromechanical devices there will be less price pressure than last year, but connectors should see more price erosion than in 1998.
Driving electronic component demand are the computer and communication industries. Those two segments consume 70% of all semiconductors produced and probably 90% of leading-edge components such as the latest and greatest microprocessors and memory chips. The good news for component manufacturers is that these two segments will have double-digit growth in 1999. Computer shipments are forecast to grow 11% and communication equipment will rise 15% in 1999, according to market researcher Semico Research.
Bring on the recovery
If there is a segment of the electronics component industry that is ready for a recovery, it is memory chips. Despite three years of 75%-80% annual increases in bit demand, the dram industry has been hammered. Consider: In 1995, the dram market climbed to $41 billion, capping a three-year period of revenue gains. Because of that success, semiconductor producers built too much capacity, resulting in plummeting prices despite healthy bit demand. The market fell to $14 billion, just about one-third of what it was in 1995. While the semiconductor business is always cyclical, such a decline was unprecedented, according to dram manufacturers.
But semiconductor buyers point out that while the market fell in the 1996 to 1998 time period, the memory market had a remarkable run from 1993 to 1995, during which dram prices increased. dram tags historically decline 30% per year as a certain density part matures. The price starts out high and drops as production volumes build. From 1993 to 1995 that decline didn't happen. Prices actually increased. So while the huge drop in prices was unprecedented, so was the increase in prices from 1993 to 1995. The three-year drop in prices was just a cyclical response to the three-year run up in 1993-1995.
For that reason buyers and analysts say that the dram market bottomed out last year, and 1999 will see some growth. The question is how much? Bill McClean, president of IC Insights, says the dram market will grow 25% to $17.5 billion in 1999. In 2000 and 2001 the market will grow 55% and 67% respectively, and hit $45 billion in 2001, he says. McClean says his forecast may be conservative, at least for 1999.
Hyundai, a leading memory chip maker, expects the dram market to double this year. The economies in Asian countries are recovering and demand for electronic equipment is expected to pick up. "Hyundai has pricing staying firm throughout the year at around $9.00 for 64-Mb parts," says McClean. "If that happens, the market will double. It wouldn't shock me, but we are not expecting that at this point."
However 64-Mb dram tags did rise in the first quarter. And the dram market grew 8% in January compared to January 1998, and 9%-10% in February and March, says McClean. And prices are higher than in 1998. In December, 64-Mb tags averaged $9.00, up from $7.80 last July. Tags climbed to $9.25 in January, but by March were down to $9.00 again.
"Right now pricing has firmed and dram pricing is still better than it was in the middle of 1998," says McClean. "But we are looking at slowly declining prices for the rest of the year." He says prices should drop 15%-20% this year, but with bit demand in the 65% range, the market should grow 25%.
Sixteen-megabit dram tags averaged around $2.95 in the first quarter, and should move down slowly. Bit shipments for 16-Mb parts are declining because most PC companies have transitioned to 64-Mb parts.
256-Mb parts on the horizon
While 64-Mb drams will be the dominant memory chip this year, dram buyers may purchase small volumes of 256-Mb drams this year for high-end workstations and servers. Samsung will ship several million, but many manufacturers are pushing the 128-Mb dram, a kind of "half-density" part between the 64-Mb and 256-Mb parts. The 256-Mb part will ship, but only in large volumes until next year when it moves to .18 micron process technology.
Later in the year buyers can expect to see some Rambus drams ship, but in very small volumes. The Rambus rollout has been delayed because Intel has delayed production of chip sets needed for Rambus until September. "Rambus would have only been about 10%-15% of the dram market in 1999," says McClean. With the delay it will be 5%-10%.
Rambus is a different architecture than synchronous. Most drams shipped today are synchronous, but as microprocessors get faster they will benefit from the Rambus architecture.
McClean estimates that Rambus will get 5% of the market this year, but that figure will grow to 25% in 2000, 50% the following year, and then become the dominate dram architecture, just as synchronous replaced extended-data-out drams. He says it appears that Rambus is the best architecture to work with high-end processors, especially at the 800-MHz level.
"There is writing on the wall and it spells Rambus," says Avo Kanadjian, vice president of memory marketing for Samsung Semiconductor. "Today, you can argue that what Rambus does for a system is overkill. However, [to get] where we are going in terms of megahertz, we will clearly need that 800-MHz Rambus chip because it has two-channel performance, which can be doubled to four channels. So without redesigning anything, you can double the performance of your memory."
Higher-performing memory chips are needed because of high-performing and faster microprocessors. Buyers can expect Intel to ship an 800-MHz version of Pentium IIs by the end of the year, as well as speedier versions of lower-end Celerons.
Over the past 18 months, Intel has aggressively gone after all parts of the PC market with processors for low-cost home computers, mid-range desktops, and portable PCs as well as high-end workstations and servers. AMD has been trying to carve out a niche at the lower end of the market; the competition has resulted in lower processor tags.
Gone are the days when Intel would cut tags once per quarter. Now it cuts prices as it introduces new speedier versions of Pentium IIs and Celerons. But AMD is not rolling over and playing dead. For instance, soon after Intel announced a 433-MHz version of its Celeron MPU, AMD announced a 475-MHz version of its K62 chip, a speedier equivalent. The chip will be used in IBM's Aptiva consumer PC. The 9.3-million-transistor processor is manufactured on AMD's 0.25-micron, five-layer-metal process technology and costs $213.
Although AMD has struggled, buyers can expect the company to continue to offer alternatives to Intel processors, at least in the short term. "It comes down to whether they can manufacture parts," says Massimini, the analyst at Semico Research. "Every time it looks like they have it together, they stumble." AMD has had yield and manufacturing problems.
"AMD tries to leap with both feet at the same time," he says. "Intel moves one step at a time. They will do a process improvement using a well-known processor design. Then they will use a new MPU design using a well-known process. AMD tries to do both at once."
AMD's next processor--the K7--will be an important product for them. "If they can produce it in high volumes and with high enough performance, they could move up higher in the price range, bringing up their overall average selling prices, which should improve their profit margins," says Massimini. That will be important for AMD's long-term success because processors represent about half of the company's overall revenue.
Buyers can expect samples of Intel's Merced processor to ship by the end of the year, and volume shipments to begin in 2000. Merced is a new 64-bit processor different from the x86 series that includes Pentium, which is a 32-bit chip. Merced will be able to run x86 programs through a hardware translation mechanism. First version of Merced will run between 600 Mhz and 1 gigahertz.
Slow growth for passives
Nineteen-ninety-nine will be a year of slow growth for passives and connectors. The connector industry will grow a paltry 1.4% to $23.8 billion in 1999, according to Ron Bishop, an industry analyst. He says manufacturers will ship more units to achieve a very small amount of growth. "There's still a lot of price erosion out there," he says. In fact, connector tags should drop 6% this year, about twice what they declined last year. The contract manufacturing industry and overall supplier-base reduction is helping drive prices down. Contract manufacturers have huge volumes, and they demand price concessions. Large OEMs such as IBM, Lucent, and Compaq also demand price drops every quarter.
Rob Rix, director of the networking communication segment for AMP, a division of Tyco, says the price erosion means that AMP and other connector makers need to analyze costs carefully.
"What has happened is you have to know in detail your cost structure and have aggressive plans to reduce it," says Rix. "We have had major migration from automated manufacturing to hand assembly in low-labor-cost countries. The acceleration of that trend is here," he says.
Besides lowering price, a key challenge in 1999 is meeting time-to-market and time-to-volume requirements of customers. "People are trying to introduce new equipment at a faster and faster rate," says Rix. "Often they need a new connector. If they can design the equipment in six months, we have to design the connector, build it, and get up to volume in that time frame. It's a challenge."
AMP and other connector makers also are aligning toward industries. "We align so we can focus and have a better product offering to that particular segment," says Rix. "The generalist view of the world is not germane anymore. It's not just products. Logistics have to match the needs of the customer base, too."
Buyers also can expect weaker prices for most passives in 1999, although overall the passives market in North America should grow 4.5% to $24.1 billion, says Ralph Anavy, president of Electronic Outlook Corp. He says one reason revenue there will be modest is falling prices.
Prices should fall about 3% on average because there is still a lot of capacity. As with connector companies, passives manufacturers are moving more production to low-labor-cost countries. But he says unit demand for capacitors and resistors will increase in 1999. Prices may drop but not as steeply as last year.
Talkback
Related Content
Related Content
Sponsored Links
View All Blogs
Now Playing:
