Is IC industry heading to the $10 million photomask?
By Mark LaPedus
Semiconductor Business News
(10/07/02 06:47 p.m. EST)
SAN JOSE -- The photomask industry is moving down two separate and dangerous paths, leading some to wonder about the future viability of the industry.
At last week's BACUS Symposium on Photomask Technology in Monterey, Calif., photomask experts dropped hints that the cost of a “mask set” could double for each new technology node. So in theory, chip makers could end up paying $1.5-to-$2 million for a high-end, 90-nm (0.09-micron) mask set, and possibly $3-to-$4 million for a top-of-the-line, 65-nm (0.065-micron) reticle by 2005 or so, according to analysts.
At this rate, some wonder if the IC industry is heading towards the unthinkable: a $10 million “mask set” by the end of the decade, which clearly would be unaffordable for most--if not all--IC makers.
Then, on the other hand, amid the current and severe IC downturn right now, Dainippon Screen, DuPont Photomask, Photronics and other merchant photomask suppliers are still engaged in a bitter and messy price war in the marketplace, according to analysts.
Recently, the average selling price (ASP) for a top-of-the-line “mask set” for 130-nm (0.13-micron) chip production has reportedly fallen by 40%, from $1 million, to about $600,000, sources said. And commodity-level masks have fallen by some 20%, leaving little or no margins for what is said to represent the bulk of reticle products being sold in the market today.
Without a doubt, this is taking a toll on the bottom line. Last month, Photronics Inc. cut 9% of its workforce and shut down its mask shop in Milpitas, Calif. And last week, DuPont Photomasks Inc. said that revenue for the quarter will be lower-than-expected due to the continuing IC downturn (see Oct. 4 story ).
There is some good news for the industry, however. According to a survey from International Sematech, the annual growth rate for the photomask business is expected to be 8% over the next ten years, according to the survey. In total, the photomask industry hit $2.3 billion in terms of worldwide revenues for 2001, according to the Sematech survey.
Now, the bad news: ASIC “tape outs”--which indicates the ASIC/ASSP activity in the worldwide IC business--are expected to fall 2.3% in 2003 over 2002, according to Sematech, a chip consortium, based in Austin, Texas. And what's more, some 75% of the worldwide revenues in the photomask industry today are coming from commodity-oriented, low-margin reticles at 0.35-micron geometries and above, according to the survey.
Chip makers vs. photomask industry?
Apparently, the overall trends in the industry are causing some tension between photomask vendors and chip makers. “There is a battle going on,” declared Y.C. “Bruno” Pati, chairman of Numerical Technologies Inc. of San Jose. “Customers are saying: 'Mask costs are too high.' On the other hand, it's hard to make money [in the photomask industry],” Pati said during a keynote address at the BACUS event.
Generally, in fact, the ASP for a reticle has jumped from $260,000 for a 180-nm (0.18-micron) photomask set, to $870,000 for a 130-nm product, he said. But the ASP for a 90-nm mask set is expected to run as high as $1.5 million, he said.
Just about everyone in the industry can pinpoint the problem with the exploding theoretical costs of reticles. “The major costs are write times and inspection times,” Pati said.
But that's only half the story, especially for ASICs. Some 40% of all ASICs require only one turn in the chip-manufacturing flow, but 25% of these parts require two turns and 20% need three turns. Multiple turns in the chip-production flow spell trouble: The reticle must be re-designed, which “means we can't afford a mask,” Pati said.
And what about the 65-nm node, which is due out in 2005 or so? Some suppliers dropped hints that the price of a cutting-edge, 65-nm reticle could double over the cost of a 90-nm product, resulting in what could be a staggering $3-to-4 million mask set.
Others believe that the figure is too high. “I've heard the number at $2 million,” said Thomas Blake, vice president of marketing for DuPont Photomask of Round Rock, Tex. “But it's hard to predict and extrapolate, because there's a lot of unknowns,” Blake told SBN in an interview at BACUS.
“At 90-nm, however, we will see a $1 million mask set,” he said. “At 90-nm, some people are also looking at soft shifters at the gate level of the device,” he said, referring to “soft' phase-shift software technologies, which will also boost mask costs.
Vendors move to reduce mask costs
DuPont Photomask and other suppliers are not standing still, however. In one effort to cut the soaring costs of photomasks, a group of chip makers, EDA houses, and tool makers rolled out a replacement for the GDS-II data-format standard (see Oct. 1 story ).
“That should help to reduce mask costs,” Blake said. “It will help to reduce the file sizes. It will also help reduce write times and cycle times.”
But that's only a small piece of the photomask puzzle. Current and future electron-beams and related tools are still a problem: they are costly and slow in terms of overall write times. “Today, the most advanced e-beam tools are $15 million each,” Blake said. And with today's e-beam tools, it still takes 12-24 hours to write only one critical layer of a mask set, he said.
And the new, advanced tools are not quite ready for prime time. In December of 2002 or January of 2003, DuPont Photomask is expected to receive a new e-beam tool from its incumbent vendor--Japan's JEOL. The new, 50-KeV tool is especially geared for the 90-nm node.
DuPont Photomask also plans to shift some designs from e-beams to higher-speed, pattern-generation tools in a move to cut mask costs. Recently, it took delivery of the new, highly-touted tool from Sweden's Micronic Laser Systems AB, but the reticle supplier has not released the details about the performance of the so-called Sigma7100. “We are still working with it,” Blake said.
While mask makers are scrambling to get their tools in production, there are other major problems: defects. “Hard defects are the primary problem [in mask design],” said Kurt Kimmel, program manager for International Sematech's Mask Strategy Program. “Fifty percent of the yield loss [in a mask] are coming from hard defects,” Kimmel said in a keynote address at BACUS.
This, in turn, has given rise to the mask inspection and repair industries. Still to be seen, however, is whether or not the photomask industry can make any money amid the current and severe downturn.
The high-cost of masks and other problems are causing many chip makers to push out their 130-nm designs, leaving photomask makers to fight for the low-margin scraps at the 180-nm node and above.
It's a bad business model in need of some major breakthroughs, especially if the industry does indeed head towards the $10 million ''mask set'' by the end of this decade. |
