Experts At The Table: 450mm Fab And Facilities Challenges

First of three parts: What’s needed for the fabs to handle bigger wafers; cost issues and consolidation; time frames and process nodes; who will survive.

popularity

By Mark LaPedus
Semiconductor Manufacturing & Design sat down to discuss future 450mm fab and facilities challenges with Gerald Goff, director of the project management office for fab design and construction at GlobalFoundries; Joe Cestari, president of Total Facility Solutions; Ivo Raaijmakers, chief technology officer of ASM International; and Michael Brain, senior director of the Fab Solutions Business Unit at Brooks Automation. What follows are excerpts of that conversation.

SMD: From your perspective, what are the general challenges to enable 450mm fabs?
Goff: As we start to look at our next factory, which is Fab 8.2 in Malta, the consensus is that the fab obviously needs to be 450mm-compatible. Ironically, we’ve spent a lot of time working with our design architects and engineers to qualify what that actually means. And surprisingly, we don’t believe that it means all that much. If you look at it from a dollar perspective, we think it’s almost a non-factor. The real changes that we see in 450mm factories in the future are pretty minimal. The floor loading, obviously, would increase to some degree. The load-bearing capability of the ceiling structure to support the heavier AMHS systems would increase, as well. In the envelope of the fab itself, we expect the vertical direction to increase about a meter in height. Outside of that, we really don’t see a need for change there. Given the three criteria of the taller ceiling, load bearing floor and ceiling grid, I don’t think it’s a stretch to say that a 300mm factory designed with those three criteria in mind could easily transition to a 450mm fab. On a utilities per-square-meter or square foot of clean room space, you really don’t see a lot of change in the demand. When the footprint of the tools got bigger, they required more power, water and chemicals. But because the tools got bigger, there were fewer of them that are placed in the fab. When you have the cleanroom changing in volume, it subsequently drives some increases in area exchanges from that perspective. And then you increase the enclosures themselves in the building. Outside of that, there is not a lot of driving costs there.
Cestari: Speaking not only from my business but from the industry at large, there are still a lot of unknowns with 450mm. There are some great examples of collaboration and knowledge exchange. But there are still a lot of gaps that are critical. You would think that with fewer people in the ecosystem, you would have better information transfer. But we still need much better information on what’s really required to build the best product in a timely fashion. You can also say that cost might be overriding technology or vice versa. The other challenge is that we still have a lot of cost in our ecosystem, from the component suppliers all the way up to the device manufacturers. We certainly can’t tolerate that going forward.
Raaijmakers: In 450mm, we are aiming for a decrease for the cost per square centimeter. Otherwise, no one will go in this direction. Of course, the 450mm tool itself will be more expensive. Overall, we think it could be 1.4x to 1.7x times the cost of a 300mm tool. That’s a pretty broad range. That’s because we don’t know what the tool will actually cost in high-volume manufacturing. In addition, spare parts, handling the spare parts and service of the tools are going to be a challenge. So in these areas, perhaps there are some benefits we can gain by a fab-wide solution, rather than a tool-specific solution.
Brain: 450mm is just another wafer size change. It brings new challenges, but nothing too earth shaking. The approaches are going to be more evolutionary for the most part rather than revolutionary. But there are some big changes because of the process geometries. As we are getting down to the single-digit nanometers of half-pitch feature sizes, the particles that are of concern are too small to measure. Process control becomes very challenging. It’s a whole new level of control that’s really down to the molecular level, not just for airborne molecular contaminants, but molecular-size particles.

SMD: What will it take to build 450mm fabs? Will it require more money, standards, collaboration, government help or what? And what is the overall impact?
Goff: The real question is who has the stomach and checkbook for it. You’re talking about a facility like a gigafab that is going to be somewhere north of $2 billion in just the base building construction before you put any equipment in it. To fit that out with equipment, you are talking about probably another $10 billion. So, you’re looking at an investment of between $10 billion to $15 billion to put a facility of this nature on the ground. There are just very few players that have the resources behind them to fund these activities. So the biggest challenge is the money.
Cestari: Cost is a start. Another challenge is the move towards true collaboration. We’ve used that word sloppily as an industry. We say it, but we really don’t mean it.
Raaijmakers: For every wafer size transition, starting with 200mm and going to 300mm, and then 300mm to 450mm, you see the same argument: ‘These fabs will become unaffordable. They will become very expensive.’ But the reality is that manufacturing is becoming an economy of scale. 450mm fabs will become more expensive. That’s true. So, there will be less people who can afford to make the switch. So that will naturally lead to more consolidation in the industry. You saw that with 300mm. And you will see that with 450mm.
Brain: We’re getting ready to host the America’s Cup race here in San Francisco. This is a similar contest, where it requires an incredibly large checkbook to get into the game. And once you get in, you need to take advantage of all the technology and drive it very hard. That’s because 450mm is all about cost. It’s all about driving down the cost per square centimeter. It’s going to be very challenging for equipment suppliers and it will probably lead to some of them not being able to make the jump from 300mm to 450mm, leading to consolidation. And those that will make the jump will be challenged, because we will be selling fewer tools.

SMD: When do you predict we will have 450mm fabs and at what node?
Goff: It would have to be smaller than 14nm. You might see 10nm or 7nm as the predominant ramp technologies for 450mm. Timing-wise? My not so humble opinion is that ASML runs the world, right? They will decide for the most part when they are ready to make a commitment to produce an automated litho cell that is 450mm compatible. I think 450mm will be significantly challenged by that. There may be some competitors that can help push that along. Short of someone pushing ASML’s buttons, I would suggest it’s closer to 2020 than it is to 2018.
Raaijmakers: We don’t know exactly if it’s going to be 2018 or 2020. But that’s probably a good timeframe. In terms of the technology node, it’s going to be a little more aggressive. At less than 10nm, we will start seeing 450mm.
Brain: I don’t disagree that 5nm to 10mm node will be the predominant node for 450mm, although as in other generations, there will be people making larger geometries on the large wafers, as well. I do think things will happen quicker than 2020. By 2017 or 2018, 450mm will begin to ramp up. We will see 2018 for high-volume manufacturing. But I am an optimist.