Wednesday, December 24th 2025
Intel's Fab 52 is Bigger Than TSMC Arizona, Produces More Than 40,000 Wafers Per Month
Intel Foundry has massive capacity in the United States, with its size surpassing that of TSMC's Arizona campus in terms of production capacity and building space. A CNBC crew visited Intel's Fab 52 in Arizona to explore its operations, uncovering some intriguing details about wafer capacity and future production plans. According to their report, Intel's Fab 52 produces approximately 10,000 wafers per week, totaling over 40,000 wafers per month. These wafers are manufactured using the 18A node, Intel's most advanced design, featuring innovations such as a backside power delivery network and gate-all-around transistors, among others.
Compared to TSMC's facility in Arizona, Intel's manufacturing capacity and node designs are superior. TSMC currently produces about 20,000 wafers per month at the older N5 and N4 nodes, which are 5 nm class, in the Fab 21 Phase 1 stage. Although TSMC plans to expand Fab 21 to increase wafer capacity and introduce more advanced nodes, these advancements will primarily remain in Taiwan, keeping U.S.-based TSMC production a few generations behind. Intel, on the other hand, has a limited number of external partners using its nodes, with the 18A node being primarily for Intel's own products, including the upcoming "Panther Lake" processor, and some future generations.However, despite producing more wafers than TSMC, Intel's current yields are lower than TSMCs, giving Intel less usable silicon for these wafers. The "Panther Lake" design is early in its yield curve improvement, with yields on the Intel 18A node increasing by approximately 7% each month. The Fab 52 facility already includes at least one of ASML's most advanced Low-NA EUV systems, the NXE:3800E. This system features upgraded components such as enhanced wafer handlers, faster stages, and improved light sources derived from next-generation High-NA tools, enabling a production rate of 220 wafers per hour. Complementing this flagship system are three NXE:3600D scanners, each capable of processing 160 wafers per hour. Overall, Intel's Arizona operations are expected to house at least 15 EUV scanners in total.
Sources:
CNBC, via Tom's Hardware
Compared to TSMC's facility in Arizona, Intel's manufacturing capacity and node designs are superior. TSMC currently produces about 20,000 wafers per month at the older N5 and N4 nodes, which are 5 nm class, in the Fab 21 Phase 1 stage. Although TSMC plans to expand Fab 21 to increase wafer capacity and introduce more advanced nodes, these advancements will primarily remain in Taiwan, keeping U.S.-based TSMC production a few generations behind. Intel, on the other hand, has a limited number of external partners using its nodes, with the 18A node being primarily for Intel's own products, including the upcoming "Panther Lake" processor, and some future generations.However, despite producing more wafers than TSMC, Intel's current yields are lower than TSMCs, giving Intel less usable silicon for these wafers. The "Panther Lake" design is early in its yield curve improvement, with yields on the Intel 18A node increasing by approximately 7% each month. The Fab 52 facility already includes at least one of ASML's most advanced Low-NA EUV systems, the NXE:3800E. This system features upgraded components such as enhanced wafer handlers, faster stages, and improved light sources derived from next-generation High-NA tools, enabling a production rate of 220 wafers per hour. Complementing this flagship system are three NXE:3600D scanners, each capable of processing 160 wafers per hour. Overall, Intel's Arizona operations are expected to house at least 15 EUV scanners in total.

20 Comments on Intel's Fab 52 is Bigger Than TSMC Arizona, Produces More Than 40,000 Wafers Per Month
Shocking. Truly shocking. Who could have foreseen this?
If Intel produces 40k wafers a month on 18A then why didn't Panther Lake launch in Q4 2025 like they promised? Unless they plan on releasing it in the next 7 days they've lied, again.
It doesn't matter how advanced their node is on paper. Success will be judged on how many can they ship (yields) and what are the product characteristics when tested by independent reviewers in terms of power, performance and area of performance (chip size relative to it's performance).
Good video...he's wrong when it comes to aquifers though, he said once you use it, its large gone. Surface water seeps through the ground and it replenishes them. Anywho, going back to the video.
www.forbes.com/sites/emanuelabarbiroglio/2021/05/31/no-water-no-microchips-what-is-happening-in-taiwan/
Thus their water needs are naturally higher than a single fab in the US.
2) lots of tech workers already present int he area, easier to find people
3) the weather. The areas around the Great Lakes and the Mississippi are very humid, and the lakes have four very strong seasons. Because chip foundries need constant internal conditions to function, building in such an area makes it that much harder to maintain humidity and temperature levels. Arizona, OTOH, is hot and dry year round. Easy to work around. Not impossible, Intel was building that fab in Ohio, but also not ideal. That's a major reason why so many of the early fabs were in SoCal or Texas.
4) Water is 95% recycled, so once things get going the input is relatively low. On the plus side, solar is, well, EVERYWHERE, its arizona, which means these fabs can produce a ton of power with solar to keep the high energy bills in check.