As is well known, Intel is trying to pivot to a foundry business fabbing chips for external customers like TSMC or Samsung. Compared to those, however, it currently lacks a portfolio of older, lower-cost technologies. For those, Intel has paired up with Taiwan’s UMC some time ago, aiming to develop an inexpensive easy-to-use 12nm process. But it now appears that this collaboration could later also produce an affordable 3nm process.
According to the analyst firm Funda, the joint Intel–UMC 12nm process project is progressing and is being prepared for availability at Intel’s Arizona fab. This year, the tools and Product Design Kits are expected to be completed and ready, while the first customer tape-outs are anticipated next year. The first customer chips could enter mass production before the end of 2027. Naturally, this assumes no problems or delays occur—there is a fairly high chance that this timeline may prove to be overly optimistic.
However, Funda also reports something even more interesting: Intel’s collaboration with UMC has allegedly been expanded to include a joint 3nm process project. It appears this is not simply a matter of UMC offering Intel’s existing 3nm process unchanged. Instead, it is said to be a technology that Intel itself will also offer, implying that this joint 3nm process will actually be different and enhanced with contributions from UMC. According to Funda, the goal is for the process to be roughly competitive in performance with TSMC’s 3nm process while likely being less expensive for customers.
Intel already has a 3nm process (designated Intel 3), which it uses to manufacture Granite Rapids server processors (Xeon 6900P) as well as a small portion of the lower-cost Core Ultra 200 processors (the rather peculiar Arrow Lake-U family). This process represents an improvement over Intel’s 4nm process and is probably more comparable to TSMC’s 5nm or 4nm technologies, whereas TSMC’s 3nm technology is one generation more advanced.
Intel 3 was originally also intended to become part of Intel’s offering for external foundry customers, but it has apparently failed in that role so far. The collaboration with UMC could remedy this failure if the program includes modifications to the technology itself, its design rules, and customer support to make the process easier and more accessible for external customers. This is an area where UMC has considerably more experience.
The question is whether UMC’s contribution will also make it possible to reduce the cost of the Intel 3 technology. In theory, that could make this process an ideal technology for a large number of products for which the continuously rising cost of advanced manufacturing technologies has become a major issue. If a process could be developed that breaks away from this trend in a positive way—with a significantly better performance-to-cost and power efficiency-to-cost ratio—it could slow or even reverse the overall rise in the prices of computers, mobile devices, and various components.
At the same time, UMC would receive a major boost from this technology. Until now—or more precisely, until then—the most advanced process in its manufacturing portfolio is expected to be the 12nm process developed together with Intel, which, although based on FinFET technology, van already be considered a mature process compared to today’s leading-edge nodes. It is intended primarily for embedded chips, network adapters, and similar devices.
A 3nm process is still close to the cutting edge today, though, and could be used more universally for mainstream computing and mobile products, for example for Arm-based SoCs in more affordable smartphones. It would also move UMC technologically ahead of where Chinese foundry company SMIC currently stands (or rather, where it is today due to its lack of access to EUV technology). If this project succeeds, it could significantly strengthen UMC’s position in the semiconductor industry alongside Intel’s position as a foundry service provider. At the same time, it could also greatly benefit the entire market and end users alike. Whether this potential can actually be realized, however, remains to be seen.
Source: Jukan
English translation and edit by Jozef Dudáš
⠀
