Source item

CoWoP has recently taken the spotlight in the AI server industry. While it is a promising technology that merits ongoing attention, it also comes with significant uncertainties and challenges surrounding mass production and commercialization. Much has already been written online about its technical advantages and manufacturing hurdles. Here, I offer two additional perspectives for consideration: 1. A Comparison with Apple’s SLP Adoption Timeline According to Zhen Ding's annual report, it appears that Apple began investing in SLP R&D as early as 2013, with mass production finally kicking off in 2017 for new iPhone models (X, 8, and 8 Plus). Over those four years, Apple collaborated with material suppliers, manufacturers, and equipment vendors to overcome both development and production challenges. This wasn’t merely the development of a single technology—it was an industry-wide ecosystem upgrade. Although today’s PCB industry is far more advanced than it was a decade ago, Nvidia does not necessarily possess stronger control over technology and the supply chain than Apple did at the time. Moreover, implementing SLP in CoWoP presents far greater challenges than Apple's use case. A rough comparison shows that CoWoP SLP involves approximately 10,000 times the system power, less than half the line width and spacing, more than three times the layer count, and 100 times the area. Without concrete test data, the expectation that CoWoP will reach mass production and be deployed in Rubin Ultra by 2028 seems overly optimistic. 2. The Simultaneous Mass Production and Commercialization Challenge of CoWoP and CoPoS TSMC is also developing another next-generation packaging technology called CoPoS, with mass production similarly targeted for post-2028. While CoWoP offers theoretical improvements in transmission efficiency and supply chain simplification, CoPoS is designed to tackle practical manufacturing efficiency bottlenecks. From a business standpoint, CoPoS should reasonably take priority over CoWoP. In reality, the attempt to introduce two major, unproven technologies within the same year carries considerable risk—posing yet another potential challenge to CoWoP’s 2028 mass production goal.