The 200mm Equipment Scramble

An explosion in 200mm demand has set off a frenzied search for used semiconductor manufacturing equipment that can be used at older process nodes. The problem is there is not enough used equipment available, and not all of the new or expanding 200mm fabs can afford to pay the premium for refurbished or new equipment.

This may sound like a straightforward supply and demand issue, but behind the trend lines are unpredictable and fragmented markets, as well as some highly nuanced and conflicting business models. Demand remains strong overall, particularly from fabs that manufacture MEMS, analog and IoT chips. In fact, there are eight new 200mm fabs under construction in China. But profit margins are so thin in markets served by those fabs that it’s hard to justify paying current equipment prices, and newcomers may have a tough time making inroads.

Fig. 1: The trend line has begun climbing again for 200mm fabs as new markets and packaging approaches add new opportunities. Source: SEMI

This has set off a scramble to find used equipment on sites such as eBay. But prices continue to rise with demand, and much of the equipment available today is in bad shape. In addition, large foundries are buying up whatever equipment becomes available, sometimes letting it sit idle until they decide what to do with it, according to industry sources.

This has left 200mm fabs facing a shrinking number of options. Among them:

• Pay the premium on new or used equipment and either charge more for manufactured 200mm wafers, which in many cases is not competitive, or extend the amortization period, which is risky;
• Buy used equipment on auction sites and hope it can be repaired;
• Buy entire fabs as advanced chipmakers sell them off;
• Add new technology into existing equipment to improve the capacity and capabilities of that equipment, which is basically making incremental improvements rather than buying entirely new machinery.

No matter which approach is taken, it will not be easy for newcomers to break into this market, particularly as entrenched foundries rethink whether to upgrade their 200mm fabs to 300mm equipment. The result may be far more competition at prices that are unsustainable for new foundries.

Who wins, who loses
What becomes clear very quickly when looking at this market is that competition is getting fierce. Until several years ago, this part of the market fell under the radar because the big profits were at the latest process nodes. But with consolidation among chipmakers and an overall slowdown in , that is no longer the case.

This isn’t necessarily bad new for equipment vendors, which now are on top of the pyramid as demand for chips produced at older node grows.

“At this point, it looks like the 200mm market will continue to be strong to 2030 and perhaps beyond,” said Evan Patton, general manager of the Reliant Product Group at LAM Research. “We are seeing increasing demand for older node (28nm or above) equipment, primarily in 200mm and 300mm.”

Foundries with fully depreciated 200mm equipment and capacity already are seeing increased revenues in their 200mm business. In fact, business is so good at older nodes that many are beginning to rethink where to place their investment dollars.

“For most applications, unless you must have highest levels of performance, there may not be as compelling a business case to focus on the bleeding-edge nodes,” said Walter Ng, vice president of business management at UMC. “With so much pressure on corporate margins, selecting one process technology node over another can impact cost of goods one way or another—not just in the silicon cost, but in yield, time to market and predictability/risk. Within foundries there also continue to be considerable development efforts to broaden the support of those established nodes to more end applications. On the process side, for instance, in the power/BCD areas, new applications continue to drive new device requirements and improved RDS on performance. On the design side, some applications may require new standard-cell libraries with smaller track heights to be more competitive. So continuing to re-invest in the established nodes is commonplace for the foundries, and is typically a win-win for the customers and the foundry.”

There appears to be widespread agreement about that. “The majority of markets will continue to stay below 300mm in the foreseeable future,” said Fumihiko Kaminaga, vice president of field solutions at TEL. “There is still enough future capability in the older equipment to keep up with the expected node changes and emerging products.”

For used equipment vendors, the future direction isn’t so obvious. Demand for refurbished equipment is growing, but there isn’t enough equipment for sale. On top of that, not all used equipment can or should be refurbished.

“There are certain classes of equipment where this works well,” said Tom Salmon, vice president of collaborative technology platforms at SEMI. “So for wet benches, yes. For lithography, no. Availability of equipment is a big concern, and the secondary equipment market is tapped out with capacity. There is no equipment to do it within the cost structure they have.”

Fig. 2: China’s projected 200mm fab capacity by product type. Source: SEMI Global 200mm Fab Outlook, 2017.

What’s driving demand
And just because these are older nodes doesn’t mean the same equipment will work across a variety of applications at those nodes. Smartphones may be well understood, but requirements for 200mm equipment used for components in those phones are changing significantly.

“The new fingerprint sensors are now switching over to piezoelectric materials,” said Mike Rosa, managing director of strategic and technical marketing for Applied Materials‘ 200mm Equipment Group. “That has driven us to come up with new chamber technology to deposit new materials. And then you need to multiply that times three to six of those devices for each new phone because you need a spatial focus, new optical films, and the integration of III-V materials on silicon wafers with optical coatings.”

Even within the same devices, specifications are tightening to squeeze more performance out of the same device.

“If you look at MEMS microphones, those are moving from capacitive to piezo-based, as well,” said Rosa. “The previous spec for cones was plus or minus 0.5-degree tilt on the sidewalls. In practice, the center of the wafer has straight sidewalls, and toward the edge of the wafer they tilt out of spec. But the new spec has moved from 0.5 degrees to 0.3 degrees, so the sweet spot on the wafer has shrunk. The only way to deal with that is to put the wafer in a larger chamber.”

Applied’s approach is to sell different size chambers for 200mm equipment, which effectively changes the capabilities of older equipment. It also changes the economics of how equipment vendors can approach this market, adding component upgrades for existing equipment instead of selling new equipment. Typically, they offer a service component, as well, to monitor and maintain the equipment.

That’s one approach. Another is to build new equipment for older nodes. “Demand for older-node metrology and inspection equipment has remained strong, primarily from companies manufacturing devices at the 40nm design node and larger, often on 200mm wafers,” said Wilbert Odisho, the vice president/general manager of KT Pro Division at KLA-Tencor. “Many of these facilities are in China, serving IoT and automotive segments. At the same time, availability of core tools has continued to dwindle. We have responded to the situation by re-manufacturing legacy tools that serve customers’ needs for bare-wafer inspection, edge inspection, broadband and laser scattering-based patterned-wafer inspection, film metrology and overlay metrology.”

But piezo-based microphones and speakers are new markets. So are many of the sensors for automotive, IoT, IIoT and medical electronics, and so is 5G. And while these are considered potentially large market opportunities, they are so new that business cycles are largely guesswork. So how much a fab can invest in equipment to get a reasonable return on that investment within a specified period of time isn’t clear.

“If you look at 5G, the base stations will need high power because 5G uses a high frequency signal, which can’t be thrown that far because it gets absorbed easily, so you need to crank up the power,” said Rosa. “That means new amplifiers, new antennas and a whole bunch of repeaters. And those are all done at 200mm or below.”

In fact, chips manufactured using 200mm processes cut across almost all markets. “The exciting part of supporting 200mm and below is the sheer diversity of markets and products,” said TEL’s Kaminaga. “Demand can come from anywhere when you have to support automotive, IoT, sensors, filters, photonics, opto, discretes, analog. We suspect it’s a lot of the internal Chinese market, as well as some IoT devices and even automotive and industrial devices. The widespread use of electronics in our daily lives will continue to drive demand for products manufactured on 200mm and below equipment. Older technology is finding a new life as companies create new markets and applications for semiconductor devices.”

Packaging and materials changes
Advanced packaging and different materials should help fuel demand for 200mm equipment, as well.

On the packaging front, system-in-package and fan-outs bring the ability to mix IP developed at different process nodes. That is ideal for analog IP such as power supplies and SerDes because analog circuitry doesn’t benefit from scaling the same way that digital logic does.

“Advanced packaging does not follow the classical node evolution, and the process is far less standardized than in the FEOL,” said KLA’s Odisho. “Change in advanced packaging technology is still accelerating and new materials are being used. And process flows continue to evolve while process windows are decreasing. At the same time, packaging designs are largely node-agnostic. So packaging represents a separate dimension for adding performance or controlling cost. Advanced packaging is more a function of device type than device critical dimensions.”

So as advanced packaging becomes more popular, demand for 200mm equipment rises with it.

“Advanced packaging certainly increases the process capabilities at 28nm and above,” said Lam’s Patton. “We anticipate its growth will further increase demand for our older node systems.”

KLA also borrows technology and platforms from its front-end groups—including older node systems—and uses them in advanced packaging process control. “We have to keep several factors in mind,” said Odisho. “Substrate thickness, substrate warp, temporary carrier type, and target critical dimension are considerations when moving front-end wafer equipment into advanced packaging. In many cases, specialized wafer handling and chucking systems must be used in advanced packaging, as well as some back-end specific software and algorithm features.”

Alongside of advanced packaging—sometimes in the same package, other times not—are a variety of different materials and substrates that are being developed at 200mm.

“There has been an increase in silicon carbide for high-power switches and for sensing power levels,” said Applied’s Rosa. “There is BiCMOS for power management. Piezo materials are being used in the ultrasonic detectors for ADAS (advanced driver assistance systems). There have been a lot of predictions about SiGe for radar, but that’s the same technology as wireless LANs, which are everywhere, SiGe BiCMOS, so that may not show huge growth. But LiDAR will be big. Car companies want to push the price of LiDAR to less than $30, using one of several technologies such as a laser with oscillating mirrors or phase-array LiDAR. We’re also seeing GaN in high-power switching.”

Silicon carbide adds its own set of issues because, unlike silicon, the material is translucent. That makes it difficult for older equipment to detect the edge of the wafer.

“You have to upgrade the lasers just to handle the wafers,” said Rosa. “Deposition is basically the same, but for etch you also need new processes for ion deposition. GaN requires different processes, and at 200mm right now there is no MOCVD (metal organic chemical vapor deposition).”

Equipment vendor business model shifts
How equipment vendors should approach this market isn’t entirely clear, which is evident in the fact that there are multiple approaches in play.

“While our newest systems are certainly more advanced than our older systems, we continue to adapt new capabilities to older systems when it makes sense,” said Lam’s Patton. “We also continue to maintain our ability to manufacture new systems for older nodes, as well as provide systems with refurbished components for cost-sensitive applications. We also have the capability to provide fully refurbished systems, and we continue to invest heavily in older solutions—in some cases repurposing technology from newer equipment like our 2300 control system back for our older equipment.”

So does KLA. “Newer equipment utilizes the latest operating systems and hardware, necessary for highest throughput, computation-intensive algorithms and leading-edge performance,” said Odisho. “For the older models, KLA-Tencor’s KT Pro Division has invested in an engineering team whose charter is to actively generate solutions for implementing newer technologies on older systems. New operating systems, solid state lasers, and updated computer platforms with newer chipsets and algorithms can accelerate throughput, enhance performance and increase longevity of the older systems.”

However, in keeping with KLA-Tencor’s front-end strategy of extending the life cycle of inspection and metrology equipment through refurbishing, re-manufacturing, and upgrades, the advanced packaging group also designs products to have long lifetimes. “Advanced packaging inspection and metrology systems are designed for extendibility and upgradability to offer customers best ROI within a rapidly innovating environment,” Odisho said.

Conclusion
There is almost universal agreement that older nodes will play a larger role in many devices for years to come.

“The reasons are numerous—overall economics (design cost to silicon cost), proven design support and time to market, and predictability in manufacturing and yield,” said UMC’s Ng. “With the significant cost increase in the most leading edge nodes from a design and manufacturing standpoint and the cost it takes in volume to drive yields, it is forcing customers to consider more carefully what is the appropriate node for their end application. For most applications, unless you must have highest levels of performance, there may not be as compelling a business case to focus on the bleeding-edge nodes.”

For equipment vendors, though, figuring out what to refurbish and upgrade, and what to leave alone, is a rather fuzzy decision. “There are challenges in supporting the older equipment platforms,” said Kaminaga. “TEL does service and provide certified refurbished tool offerings on several generations of equipment, but upgrading the oldest models has limitations. Any new hardware improvement development involves factory resources and time. The challenge is to create relevant modifications on the platforms that bring the most value to the customers and have market demand.”

It’s not much clearer on the new equipment side. “The demand for 200mm and below equipment continues to be very strong, and the lack of viable cores has had an effect on the IDM’s buying preferences,” he said. “But this resurgence of new tool production comes with strong demands for pricing concessions.

Who wins in this market, and what strategy wins, so far is not obvious at this point. What is clear is that 200mm will be a hotly contested arena for years to come, which is a radical change from what most experts predicted with the introduction of 300mm equipment.

Related Stories
200mm Crisis?
Sourcing Used Equipment
China Fab Boom Or Bust?
200mm Equipment Shortfall