20x capacity in San José?
On the Kauppalehti forum, Uskala (@Tnokka on this forum) claimed, citing their own AI sources, that the 20-fold increase in chip capacity at Nokia’s San José plant is not true. Here is my reply to them and the ensuing dialogue:
ME: This was stated by a high-level Nokia executive (Ron Johnson) himself at the optical OFC event. Direct quote:
“Yes, 6-inch, which is 4x the capacity of a 3-inch wafer like we have today. That, along with the advancement in tools and the more tools gives us like 20x the capacity to build really, really complex things. As we build simpler things, like, for instance, lasers or inside the data center applications, it could drive many, many, many more wafers to address some critical needs in the network today.”
USKALA: That is true. But you could have explained from the beginning what “like 20x the capacity” means. It doesn’t mean 20x the output, but the change in the end products can be calculated that way.
ME: Here is Gemini’s comment on the matter:
During the ramp-up phase, yield failure and the learning curve of the new line ensure that the number of finished boxes does not immediately increase 20-fold. In the beginning, the waste percentage is always higher when moving from 3-inch to 6-inch wafers and new automation tools.
Over time, and as the process matures, this yield risk is neutralized. What physically increases 20-fold here is the total surface area of chips processed through the factory (wafer/circuit capacity). Johnson was specifically talking about this computational processing capability of the production line, which consists of pure geometry and machine power:
- Geometry (4x more surface area): The surface area of a 6-inch wafer (pi x r^2) is exactly four times that of a 3-inch wafer. Thus, in the same manufacturing cycle, you get 4 times more physical chip surface area per wafer.
- Tool efficiency (5x more wafers): The new automated 6-inch cluster tools are able to handle and process five times more wafers per hour through the line compared to the old, more manual 3-inch line.
When these two factors are multiplied together (4 x 5), the plant’s internal ability to churn out finished component surface area increases 20-fold. The output of finished and functional end products stabilizes closer to this new capacity ceiling as the yield curve reaches its optimal level.
Hopefully, the matter is now clear even to the skeptic, so there’s no need to reach for the “explain it like I’m five” tools…