SEMICON WEST REPORT: Smoother, unsynchronized cycles help industry, analysts agree

by Bob Haavind, Editorial Director, Solid State Technology>

The semiconductor equipment market is in a pause right now, agreed analysts at a SEMICON West Bulls&Bears session, but it will not be severe and some even see a nice pickup next year.

“We’re definitely in a pause, probably at or near the bottom, a U but not a V,” said analyst Robert Maire of Needham. He expects the market to be “flattish” next year, rising in single digits, perhaps 5%.

There was agreement that the wild cycle swings of the past have moderated, and major equipment buying sectors are not as synchronized as in the past.

“Logic companies don’t build their field of dreams any more,” pointed out Jay Deahna, JPMorgan analyst. “Book-to-bill used to swing between 0.5 and 1.5,” he added, but this has moderated.

“Cycles are out of synch, but we’ve gotten lucky not more rational,” according to Brett Hodess, Merrill, Lynch analyst. He believes there will still be cycles, but more muted. Chipmakers can plan better because of shorter delivery time on tools, he believes. “Going from 200 days to 100 days helps,” he added.

A wider product spectrum, especially across consumer markets, also may be helping to smooth cycles, suggested Moderator Thomas Caulfield, EVP at Novellus Systems.

More coordinated planning might also be at work. While the industry has not consolidated, as some predicted, it has clustered since the ’90s, pointed out Maire. “There are five or six ‘clustomers’ vs. 50 or 60 (buying fabs) in the ’90s,” according to Maire. Some chipmakers are now willing to avoid overbuilding, he said, even if some capacity goes to competitors. He also agreed with Caulfield’s point about more end product diversity — “It used to be the new Microsoft OS that drove the industry,” he commented.

Tim Acuri, Citigroup analyst, believes that the current downturn will be shorter, but not necessarily shallower, than in the past. Memory, including flash, provides about half of the business, and the high level of buying there may be sustained for another year. But Acuri sees a secular change in the logic/foundry sector because of the “fab lite” trend, with major chipmakers turning over much more future manufacturing to foundries.

Much ado about memory

There was some lively discussion about the outlook for memory. Deahna of JPMorgan believes that there may be some speculative capex spending based on consumer shifts, such as the iPhone, 3G phones, and Microsoft’s Vista operating system, all of which will use lots of memory, he pointed out. “Samsung sees a large increase in the average DRAM per box, like 1.3 to 1.6Gb, maybe even 2Gb,” he said.

Maire of Needham suggested that there is no rush to Vista, and that PC memory will top out at 2Gb, because there is no need for 4 or 6Gb. He cited Dell’s offering an option for new customers to go back to XP if they wished.

Acuri of Citigroup pointed out that NAND in notebooks is having a much bigger impact on the memory market than Vista, which he sees as only a short-term issue. Maire agreed that solid-state drives (perhaps 32Gb) will be an even bigger memory market.

Caulfield cited studies that suggest that there will need to be a million wafer starts/year to satisfy Vista demand, but Jay Deahna thinks that “you will need to double that in the next couple of years,” citing another study that found only 8% of companies planning to shift to Vista this year, while 47% say next year, and the rest undecided. As a result, he believes that memory makers will have to double DRAM starts by July of next year.

Brett Hodess agreed that next year will be a big PC year, and that small- and medium-sized companies will have to change over to Vista. “But you don’t need more than 1Gb of memory,” he said, adding that more memory won’t really improve the PC performance. Hodess said that he recently bought a PC for his daughter, who wanted lots of memory, but all he could find were systems with 1.1-1.3Gb — he could not even find a 2Gb PC.

While unit demand has been rising 10%, according to Caulfield, lower ASPs (particularly for DRAM) have held down memory revenues. But that could change next year, Deahna pointed out, because memory demand will be rising without more capacity, so he sees profits increasing.

While Maire said he felt that flash was the real driver and the amount of DRAM is irrelevant, Acuri disagreed, pointing out that DRAM was finding many new applications and that there was a lot more specialty DRAM that could command higher prices.

“Do the math on Micron,” Maire responded, “It would be better to put your money under the mattress.”

Complexity driving 300mm surge, but not consolidation

Maire believes, however, that we will see an increase in capital intensity over the near term that could boost process tool markets. There was agreement that most remaining 200mm fabs are not profitable, because of lower ASPs, and those chipmakers are being forced to convert to 300mm rapidly.

At the same time, “complexity is going up exponentially,” Maire said, who counted 65 masking and 1000 wafer processing steps, while line widths are shrinking and wafer size is increasing. “We are seeing the last vestiges of the 300mm shift, with some ability to raise productivity left in the DRAM sector,” he believes. Beyond that, there is less chance to compensate for the complexity increase. He sees capital spending in the near term rising above the long-term 20%-22% trend line.

The major challenges he sees are in adding new materials to the process flow, and, in the near term, coping with the huge cost for lithography. He sees the industry struggling to find a way around having to spend $30 million or more for EUV tools.

The problem, according to Tim Acuri of Citigroup, is that “the return on invested capital is not pretty.” He sees a secular decline in ROIC, which he believes will force further consolidation in the industry. This brought a round of disagreement.

“There is only one mega-merger in this industry that really worked out — KLA-Tencor,” responded Deahna, although he granted there have been successes with some strategic plug-ins.

The industry won’t shrink to 8-10 companies, believes Hodess, because of rapid technology change that can be best handled by specialty companies in a particular sector. The industry has already consolidated to some extent, he said, although customers don’t seem to want three contenders.

“When you do get three, the third player drops out,” Hodess commented.

He also said that the complexity of the technology along with the efficiency of many of the tool companies doesn’t give the opportunity that private capital or deal-makers are looking for to boost revenues or improve operations.

Maire agreed that few acquisitions have worked out, while there are hundreds of small companies. Five or ten years from now, he expects, there will be even more companies in the sector, although the top ten firms will do about 75% of the business.

Not enough broad support for 450mm

Caulfield from Novellus raised the prickly topic of a transition to 450mm wafers, a subject that generated renewed debate at this year’s SEMICON West. “That’s an esoteric discussion, with only two companies asking for it,” Acuri responded. “Tool companies are not interested, especially if they aren’t big at Intel or Samsung.”

Maire pointed out that the shift to 300mm was very painful, and a move to 450mm will be ten times worse. He suggested that the industry must go back to the Intel model for the 200mm transition. The tool companies face a tough economic model, he explained, and if only a small base of chipmakers want 450mm wafers, they must provide the money.

“I promised my wife I wouldn’t stay around for another wafer size change,” Hodess confessed. He believes that 2018 might be a reasonable compromise target for a transition, with work starting at the turn of this decade.

Hoping for a “perfect storm”

Hodess suggested that the push might come with a “perfect storm” upward in the industry, with strong markets for DRAM and flash, and a surge for the foundry/logic sector at the same time. “If applications came along and all the chipmakers needed equipment for technology changes, this could happen,” he predicted. If it did happen, the tool companies would outperform the overall stock market.

The law of big numbers dictates against such a scenario, suggested Maire. Even a new application area like the Internet wouldn’t have a big enough impact to move the market like this. Macroeconomic factors are more important for driving the market now, he believes.

Acuri pointed out that investors know that the market is cyclical, and would discount any boom in the tool market, so there would not be a big impact on stocks.

With even a hint at a perfect storm in the upward direction this time, the session ended with an upbeat feeling in spite of the current slowdown. — B.H.


Easily post a comment below using your Linkedin, Twitter, Google or Facebook account. Comments won't automatically be posted to your social media accounts unless you select to share.