Intel CEO will visit TSMC next month to discuss the 3nm process
According to Intel’s plan, Meteor Lake will be launched next year. It adopts a modular design, which can be stacked with modules of different process nodes, and then interconnected using EMIB technology and through Foveros packaging technology. In addition to using Intel’s own new Intel 4 process, it will also use modules made by TSMC, which is rumored to be based on the N3 process.
It is rumored that Pat Gelsinger’s trip to TSMC this time is mainly to make an “emergency revision” to next year’s production capacity plan. It is said that Meteor Lake has been delayed until the end of 2023. With the change of Intel’s Meteor Lake schedule, TSMC needs to cooperate with changing the original production plan.Some researchers say that if Intel really delays the release of Meteor Lake, there may be a big loss. Intel and TSMC are rumored to have signed an outsourcing agreement, and TSMC will elevate Intel to the same customer status as Apple, with a purpose-built R&D center and production line for the plan, including the P8 and P9 production facilities at Fab 12 and the expansion of Fab 2. If Intel is unable to mass-produce on time due to “market conditions” or “technical reasons” due to its own Intel 4 process, Intel hopes that TSMC will also delay production, but Intel is likely to bear all the losses.
It is rumored that Intel has another plan, that is, the original GPU modules will continue to be produced as planned, and at the same time, the computing modules of Meteor Lake will be outsourced to TSMC and manufactured using a 3nm or 5nm process. This approach is likely to cost Intel’s reputation, but it can “take a breather” because of the cost savings. Given that Apple, AMD, Nvidia, Qualcomm, and MediaTek all need to use these advanced process nodes, it’s unclear whether TSMC will have enough capacity to meet Intel’s needs.