Applied changes CMP playing field
Competitors fight back by cutting costs and trying to do a better job on surfaces, copper processing, and deep-submicron devices
semibiznews.com
By J. Robert Lineback
The market for CMP gear may have exploded in recent years, but that doesn't mean that business is booming now for all of the 20-or-so companies selling this increasingly popular chip production equipment. In fact, just one company appears to be making it big today in chemical mechanical planarization tools. Moreover, a major shakeout seems to be underway as chip-gear giant Applied Materials Inc. carves out a growing slice of the CMP pie.
"Applied is smoking -- going from $80 million [in 1997] to more than $220 million in CMP last year," points out analyst Risto Puhakka, director of chip-making markets at VLSI Research Inc. in San Jose. That kind of growth has moved Applied from a weak fourth-place position in CMP tools with just 2% of overall sales in 1996 to the leader last year with a 40% market share.
Applied's competitors may be impressed by how fast this 800-pound gorilla has grown, but they certainly aren't giving up on this market. The reason is simple: CMP is becoming pervasive in global wafer fabs as device feature sizes shrink below 0.18-micron and copper metal begins to replace aluminum interconnects on ICs.
"The big debate now is how big CMP will get before it penetrates all major fab applications," comments analyst Ronald Dornseif, who tracks wafer-processing technology at Dataquest in San Jose. "Right now, I'm projecting [the annual market for CMP gear to hit] about $1.5 billion by 2003."
No one could have imagined just how important CMP would become when these tools first began to appear in IBM Corp.'s chip-processing plants in the late 1980s. "People used to shudder at the thought of CMP processes in their fab," Dornseif recalls. "It was nothing but an overgrown sandpaper machine."
"But IBM led the way with the technology and now there is a whole cadre of engineering resources developing CMP." As a result, he adds, "It is becoming less and less a black art and more of a science."
Chemical mechanical polishing, as it was originally called, was used in the beginning to smooth out dielectric surfaces in multi-layer interconnects. Then CMP became a way to precisely scrub metal from wafers for gap-filling and for tungsten plugs. Recently, the "sandpaper machine" has taken off as a critical tool for shallow-trench isolation in deep-submicron devices on logic and DRAM chips.
CMP continues to spread into other applications. It is flattening wafer surfaces before and after process steps to improve results and yields from deep-ultraviolet photolithography. CMP also is aiding in building structures. And it has taken on the essential role of forming copper interconnects in dual damascene processes.
"There isn't a major semiconductor manufacturer today that doesn't have some level of experience with CMP," maintains Rob Davenport, director of product management for Applied's CMP division in Santa Clara, Calif. "Some [experiences] are further along than others, but everyone now accepts CMP as a fact of life in the fab, and there's no way around using it."
Even though Applied may now be forcing a shakeout in this market, the growing need to polish materials off wafer surfaces is still attracting new players who want to grab a piece of the action. Global revenues for CMP last year amounted to $563 million and are expected to grow between 30% and 40% annually through the early 2000s, according to VLSI Research in San Jose.
But it will be tough for some suppliers to keep up. In the past four years, competition in CMP has greatly intensified. Applied's meteoric climb, the emergence of new tool architectures, and a hiccup in CMP's growth last year have all conspired to shuffle the market leadership.
Hardest hit, in terms of market share, was Integrated Process Equipment Corp. (IPEC), which owned a commanding 50% share of the 1995 market. Since then, the Phoenix firm has lost ground to Applied Materials, SpeedFam Corp. in Chandler, Ariz., and Ebara Corp. in Japan.
IPEC shifted much of its CMP effort to an orbital polishing technology, and analysts say that the move caused the company to fall behind competitors in keeping up with overall market growth. Crosstown rival SpeedFam moved fast to fill a need for high-volume, parallel-polishing CMP systems. These tools are capable of running up to 70 wafers an hour - far more than the 35 to 50 handled by other platforms.
But the downturn and Applied's surge has resulted in 1998 revenues falling by more than 50% at both IPEC and SpeedFam. By the end of last year, SpeedFam had agreed to acquire IPEC for about $200 million and is now planning to mount a comeback that will bundle a variety of CMP techniques on a single platform (see "SpeedFam-IPEC will use diversity to try and get back into CMP race" in the current publication).
Other suppliers were also hit hard by the competition from Applied, which only last year had turned into a CMP juggernaut. After a shaky start a few years ago, Applied's Mirra tool has finally taken off due partly to its flexible architecture, which features three platens and four polishing heads. This setup allows chip manufacturers to use different slurry formulations on each platen for multi-step CMP polishing sequences, a feature that is said to be important in the dual-damascene copper processing of interconnects.
Applied uses a proprietary head, called Titan, which employs a flexible membrane and low-pressure polishing to reduce the dishing of surfaces and improve uniformity. Mirra also has an in-situ endpoint-detection system that helps to control against overpolishing.
"To be frank, it took us close to a year to work through some reliability issues," recalls Davenport. "Now we have one of the most reliable systems on the market. Otherwise we couldn't be gaining market share."
It certainly doesn't hurt either to be a $4 billion equipment supplier. Applied is flexing its muscles by bundling tools together for integrated processes in copper interconnects and other key CMP applications. "Applied is the only equipment supplier in the world where the company is the product," says Dataquest's Dornseif. "That's a huge barrier."
More than a dozen niche players and new entries in the CMP market are now running up against that barrier. In Japan, they include Sony, Sumitomo Metals, Tokyo Semitsu, Lapmaster, Okamoto Machine Tool Works, Mitsubishi Materials, Toshiba Machine, and Mitsubishi Diamond Industrial. And in Europe, the lists includes Steag MicroTech and Peter Wolters of Germany. In the U.S., they include two established suppliers - Strasbaugh and Cybeq Nona Technologies-and two Silicon Valley startups - Obsidian Inc. and Aplex Inc.
Now beginning to ship its first CMP tools is etch equipment supplier Lam Research Corp. The product is based on fast-moving belts that planarize wafer surfaces like a sander. The Fremont, Calif., company claims its linear planarization technology has inherent advantages in reducing dishing and unwanted erosion over conventional rotary, or "round-and-round" polishers.
"We were late to market, but we have focused on what we believe are two emerging technologies for CMP applications - copper and STI [shallow-trench isolation," explains Andy Cohen, director of marketing for Lam's new CMP/Clean Products Business Group, which resulted from a merger of its CMP products with cleaning tools offered by its OnTrak Systems subsidiary.
Like Applied, Lam is banking on its size and worldwide support organization. "We are also already seeing a shakeout and that will continue," Cohen predicts. "The barriers are huge [in this market], he says. "You must [not only] have the technology and show the results on the wafer, but after that it comes down to the level of service you can provide worldwide."
Most of the smaller CMP players are banking heavily on technology, niches, and what they believe is a better way to spin polishing wafers. All of them are working hard to attack the huge cost of operating CMP tools by trying to reduce the amount of slurries and polishing pads. Consumables can easily account for half of the total cost of operating a CMP tool, an amount that often far outweighs the cost of tool depreciation, analysts say.
"There are probably 30 or 40 different factors that influence the product you eventually get out of these tools," observes analyst Robert N. Castellano, president of The Information Network in New Tripoli, Pa. They include polishing pads, slurry formulations, amount of pressure, how fast surfaces are moving, and what type of motion is used, he adds.
One new approach to the consumables problem that's been getting a lot of attention recently is slurry-free CMP. It promises to reduce the high cost of consumables in polishing as well as providing more consistent results in high-volume production fabs. Last February a partnership between 3M Co. and Rodel Inc., the biggest supplier of CMP pads, was launched to develop slurryless planarization tools that use 3M's fixed-abrasive pad technology.
Another new approach to CMP comes from startup Obsidian Inc. After more than a year of top-secret development work, the Fremont, Calif., company reportedly has now started sampling prototypes of a fixed-abrasive CMP system.
Nearly every supplier -- including Applied -- is working on slurry-free CMP because it is being touted as a paradigm shift in planarization. "Whether wafers are polished with round-and-round, linear, or orbital motion is not an issue -- but fixed abrasives are different," declares Applied's Davenport. "It will eliminate slurries [and the need to dispose of them] and potentially improve the cost of ownership." But, he adds, "a there is still a lot of work that needs to be done to improve the technology."
That's not keeping Applied from rushing into slurry-less CMP, however. In late May, Applied agreed to acquire Obsidian for $137.5 million. The deal is expected to be finalized by the end of June and could be one more jolt to Applied's struggling competitors.
|
