Intel has announced that the 18A process (1.8nm equivalent) is progressing well and exceeding expectations. Consequently, the 20A process (2nm equivalent) version of the Arrow Lake processor has been canceled in favor of using an external foundry (TSMC).

There are reports that Broadcom is testing the Intel 18A process, but the results are not promising and may even lead to the cancellation of their foundry cooperation.

At the Deutsche Bank 2024 Technology Conference, Intel CEO Pat Gelsinger stated: "I'm very excited to tell you that the 18A process has a very healthy defect density (D0) of less than 0.4 at this stage."

Defect density refers to the number of defects per square centimeter of wafer. Generally, less than 0.5 is considered good. If the 18A process indeed achieves less than 0.4, it is undoubtedly quite impressive.

Especially considering that the 18A will enter mass production in a few more quarters, its defect density will undoubtedly be further reduced.

The TSMC N7 7nm and N5 5nm both had defect densities of about 0.33 in the first three quarters of volume production, with the latter dropping to 0.1 in volume production. The former, however, continued to have a relatively high defect density in volume production and took a couple more quarters to reduce it to 0.1.

Intel's first consumer product to use the 18A process will be Panther Lake, expected to be named the Core Ultra 300 series. The first data center product will be Clearwater Forest, expected to be named the Xeon 7 series. Both are set to enter mass production and be released in 2025.

The 18A process is not only a crucial step for Intel to surpass TSMC in process technology but also a significant opportunity that will be made available to external foundries.

Intel revealed that potential foundry customers, such as Microsoft and the U.S. Department of Defense, are already highly interested. It is expected that eight 18A chips, including Intel's own products and those of external customers, will be completed by mid-2025.