Foundries Eye 300mm Analog Fabs

Whether they’re better than 200mm is often a matter of opinion and market segment, but competition is definitely heating up.

popularity

By Mark LaPedus
In 2009, Texas Instruments changed the semiconductor landscape when it opened the industry’s first 300mm fab for analog chips.

Until then, analog chip production was conducted in fabs at 200mm wafer sizes and below. With a 300mm fab, TI potentially could gain a die-size and cost advantage over its analog rivals. On paper, a 300mm wafer provides 2.5 times more chips than a 200mm substrate, thereby enabling TI to lower its overall manufacturing costs.

Over the last year, Infineon and STMicroelectronics have begun ramping up select analog chips within their respective 300mm fabs. And seeking to fill the void for fabless and fab-lite customers, GlobalFoundries is one of the first foundry vendors to offer 300mm capacity for analog/mixed-signal technologies. “TI has these technologies and also 300mm,” said Michael Noonen, executive vice present of global sales, marketing, quality and design at GlobalFoundries. “How do (foundry customers) compete against the likes of TI? We want to be the answer to that question.”

And in what will continue to change the landscape, Powerchip, TowerJazz, TSMC and UMC have begun offering—or may provide—300mm analog foundry capacity. But most specialty foundries don’t have 300mm fabs and say they are not at a disadvantage. Analog, they argue, is process-oriented and not tied to the wafer size.

It’s unclear if analog chipmakers can absorb the new 300mm foundry capacity. Some analog IDMs keep their manufacturing in-house, while others are already using foundries. And there are other prevailing questions. Does a 300mm analog fab give chipmakers a competitive edge? And will the 300mm fab vendors drive out the analog chipmakers that process chips in smaller plants?

“Fabs running 200mm wafers will continue to be profitable for many more years and will be used to fabricate numerous types of ICs, such as specialty memories, image sensors, display drivers, microcontrollers, analog products, and MEMS-based devices,” said Trevor Yancey, an analyst for IC Insights. “For the most part, 300mm fabs are, and will continue to be, limited to production of high-volume, commodity-type devices like DRAMs and flash memories, and very recently image sensors and power management devices.”

Indeed, in large-volume analog markets such as interface chips and power management, a 300mm fab gives a chipmaker a die-size advantage, said Susie Inouye, research director at Databeans. “There are many analog products that are not cost-efficient to run in a 300mm fab, such as op amps and others,’’ Inouye said. “I would say there is an advantage in other products. In power management, for example, TI has an advantage with a 300mm fab.”

On the other hand, 300mm analog fabs are potentially difficult to fill and could lead to unwanted overcapacity in slower periods. In fact, TI itself is engulfed with its own, internal fab capacity glut right now. For the industry in general, there is overcapacity for older processes at 0.35-micron and above, but mainstream 0.18-micron technologies are tight.

Analog: Still the darling?
For years, thanks to stable and steady growth, analog has been the darling of Wall Street. Last year the analog IC market took a hit, falling 7% in 2012 over 2011, according to Databeans. The overall IC market fell by 3% in 2012, according to the firm. But the analog IC market is expected to rebound and grow 6% in 2013, Inouye said. Smartphones and tablets continue to drive analog, but the more traditional markets remain sluggish, she said.

On the manufacturing front, there is also some uncertainty, if not a sea of change. The dynamics changed overnight in 2009, when TI opened RFAB, the industry’s first 300mm analog fab. Located in Richardson, Texas, RFAB is in production and running LBC7, a linear BiCMOS process, and C05, a 0.18-micron analog technology.

When TI opened the fab, there were fears among some that the company could drive out the competition and bomb the prices. As it turned out, TI did not drive out its rivals; nor did it engage in a price war. In fact, the company wants to maintain its margins.

But RFAB gives TI an edge on two fronts. “40% of all analog being manufactured on 300mm is power management. 300mm allows them the ability to put more die on the wafer,” said Databean’s Inouye. Secondly, with 300mm capacity, TI also can provide customers with a guaranteed source of supply. This is a key for large customers such as Foxconn and Samsung, which require enormous volumes during their product ramps, she added.

Foundries eye 300mm analog
TI, Infineon and STMicroelectronics are among the elite few with 300mm analog/mixed-signal fabs. The other analog IDMs do not have 300mm fabs, meaning they must reach out to the foundries if they need larger wafer sizes.

To one degree or another, GlobalFoundries, Powerchip, TSMC and UMC have 300mm foundry capacity for specialty processes. GlobalFoundries, TSMC and UMC also are pursuing the leading-edge digital market. “We really think there are two types of SoCs,” said GlobalFoundries’ Noonen. “There is the SoC at the bleeding edge. Then, there is the SoC that is ‘Moore than Moore.’ Both are interesting.”

Late last year, GlobalFoundries embarked on a strategic initiative, called “Vision 2015,” which involves its fab operations in Singapore. In Singapore, GlobalFoundries owns both 200mm and 300mm fabs, which were primarily used for mixed-signal and logic production, respectively.

Under the new initiative, GlobalFoundries will expand the 300mm fab in Singapore for analog/mixed-signal production. In the fab, it has begun installing 300mm tools, which were recently acquired from Taiwan’s ProMos. The 300mm plant, dubbed Fab 7, will be capable of making nearly 1 million wafers per year, up from 600,000 wafers a year right now. The expansion is expected to be completed by mid-2014.

All told, GlobalFoundries believes its analog/mixed-signal processes will enable foundry customers to compete more effectively against the IDMs. “We really want to accelerate our mixed-signal technology at 300mm,” Noonen said. “The idea is to tap into technology that is not only competitive, but is better than what other people have in their captive fabs.”

Meanwhile, in May, United Microelectronics Corp. (UMC) implemented a similar strategy. UMC turned its 300mm fab in Singapore from a leading-edge logic plant into a specialty process production facility. In the fab, UMC will make CMOS image sensors, embedded memory and high-voltage chips. The fab will “enable UMC to tap into new market opportunities,” said W.Y. Chen, chief operating officer at UMC.

200mm is good enough
Meanwhile, the specialty foundries own fabs with 200mm wafer sizes and below, ostensibly putting them at a disadvantage. “We do not see 300mm as a significant threat today,” said Marco Racanelli, senior vice president and general manager at TowerJazz. “In the specialty analog markets we serve today, what differentiates us versus our competitors is performance and service and not wafer size.”

Long term, TowerJazz is also looking at 300mm. The company and IBM are part of a group that is bidding for a 300mm fab project in India. “We do see over the next several years a subset of analog products that include low-voltage power, CMOS image sensor and RF that can take advantage of the larger wafer size, and thus we are looking at creating this option for our customers,” Racanelli said.

Most specialty foundries have no plans to build a 300mm fab. For example, an executive from MagnaChip Semiconductor said the chances the company will build a 300mm fab are “zero.” MagnaChip’s competitive edge is process development, not wafer size, according to the executive.

MagnaChip has two 200mm fabs capable of making 114,000 wafers a month. It is one of several vendors that compete in perhaps the hottest mixed-signal market: bipolar-CMOS-DMOS (BCD). BCD combines the advantage of several processes, including bipolar for analog, CMOS for digital, and DMOS for power and high voltage.

In BCD, MagnaChip has four technology platforms: deep trench isolation (DTI), junction isolation, non-epi and SOI. The flagship process is a 0.18-micron, 80-volt technology based on DTI. It is also developing 0.13- and 0.11-micron BCD technology. “The larger geometries are not going away,” said Francois Hebert, vice president of engineering at MagnaChip. “0.18- and 0.35-micron will stick around for awhile.”

It has also added a silicon-on-insulator (SOI) process on its roadmap. Based on SOI and DTI, MagnaChip’s new 0.35-micron, 16-volt process is targeted for OLEDs. Soon, it will release a 0.18-micron version. “I love SOI,” Hebert said. “It can get rid of all of the natural problems (relating to) capacitance and low-voltage.”

STMicroelectronics, meanwhile, has been in production with a 0.32-micron BCD process based on SOI. For 70-, 100-, 140- and 200-volt applications, the chipmaker is ramping up a 0.16-micron version. Meanwhile, for 8-, 40- and 60-volt applications, STMicroelectronics is also ramping up BCD9S, a 0.11-micron, BCD technology not based on SOI. One of the applications is for braking systems in automobiles, said Jean-Marc Chery, general manager of the Embedded Processing Solutions Segment at STMicroelectronics.

On its roadmap, STMicroelectronics plans to offer a 90nm BCD technology within its 300mm fab by 2014 to 2015. In doing so, the chipmaker and others will likely drive down the costs of BCD, making it more attractive. By then, however, the analog/mixed-signal industry in general could also be swimming in overcapacity amid a push toward 300mm fabs.