Face it, 300mm Prime didn’t really happen. The industry, as a whole, had a chance to optimize 300mm fabs for cycle time and productivity, and maximize the investment. There was particular benefit available for high-product-mix manufacturers. Instead, 300mm Prime became Next Generation Factory, which, by over-emphasizing Predictive and Preventative Maintenance, came full-circle to the early 2000’s SEMATECH and SELETE goal of getting quality data off the tools (via e-Diagnostics). Not that these weren’t great goals, but they certainly weren’t 300mm Prime. Granted, some chipmakers used the factory modeling portion of the work to internally reduce cycle time, so there was individual benefit.
But if we really want to get the biggest factory bang for the buck on 450mm, some things have to change, and some of that change deals with getting deeper chipmaker involvement early on.
One example, as we move into the 450mm realm, there’s evidence that factory automation and planning is still getting short-changed. Some of this confirmation is the perpetuation of solely 25-wafer Front Opening Unified Pods (FOUPs). In 2018, 10nm transistor structures and 450mm wafers will give us tens of thousands of die per 25-wafer FOUP. This is great for high-volume, low-mix memory and processor manufacturers, but hardly optimal for foundries, where high-product-mix may force transport and stocking of partially empty FOUPs. Twelve or even six-wafer FOUPs may be more optimal. Take a look at the 300mm Prime Guideline #12 from 2007 for an example of how such a FOUP might be used. TSMC and Globalfoundries, supporting ARM and other processor variants, are often competing head-on against Intel in the processor space—likely more so in the future. However, with the foundries’ mix of products, they take a relative productivity hit by comparison because it takes precious time to swap recipes, reticles, and purge tools on a product change-over. Recipe management, cycle-time, stocking, distribution, are all impacted by the high-mix, high-volume nature of the foundry business. Perhaps this is all well-covered in G450C. The fact there’s only a 25-wafer FOUP, SEMI Standard for 450mm implies it’s not.
450mm Company Strategy
Just planning to “buy 450mm” when it’s available isn’t a good enough strategy—not that any chipmakers serious about 450mm are using this approach. But being involved in G450C isn’t really good enough either. G450C won’t buy very many tools, so, similar to involvement in SEMATECH programs, suppliers might do just enough to keep managers at G450C happy, saving the real goods and special deals for the real buyers, the chipmakers. Therefore, having a company strategy, of which G450C is just one facet, must part of the plan.
First of all, if G450C is really the proving grounds where decisions are to be made, you’ll need to have your company buyers convince suppliers to deliver the best tools possible to the G450C pilot line for evaluation—don’t ask them to hold the special features back for just your fab. Next, your own company’s deep involvement and leadership within the industry is required. A lot gets done in side-meetings at conferences, trade shows, symposiums, and standards meetings that won’t take place within G450C. If anything, G450C-only involvement will lull one into thinking the collective power of the group is moving everything forward. If this were true, ISMI’s 450mm program would have pushed 450mm through in 2012. 450mm is happening because chipmakers are now collaborating and leading—with their wallets—and because Intel has done a lot of work behind the scenes (in standards, government, etc.).
Cost of Strategy
Companies that have gained the most out of their consortium and association memberships are those that jumped in with both feet. At SEMI, SEMATECH, ISMI, and SELETE this was true. Unfortunately for some companies, the price of their membership was analogous to buying a ticket to watch a soccer game from the stands. In particular, Globalfoundries’ late entry into the G450C effort puts them at a relative disadvantage within this camp of experienced 450mm professionals, and they might be in danger if they don’t actively work to prevent this scene from happening.
Continuing the soccer game analogy, Globalfoundries might even have to purchase a center or striker to come up to speed quickly. Likewise, I can’t imagine a company, like Toshiba for example, even trying to enter and help set industry direction at this point in the game.
There are many billions of investment dollars and many more in revenue at stake. This is not the time to be complacent on an issue and assume it will be addressed by any other industry partner. In fact, one of the reasons it took so long for G450C to come together is that the G450C industry partners didn’t completely trust each other to begin with—everything had to be in writing. So, to the contrary, it’s time for member companies to become collectively involved in G450C and individually active in the semiconductor community, taking ownership over the particular company’s nuances (i.e. foundry/high-mix). It’s five short years until high volume 450mm production. Fine-tuning 450mm, before then, through industry groups, will take time.