Intel has decided to postpone volume production of its Meteor Lake's GPU tile using TSMC's 3nm-class technology (N3, N3E, etc.) to late 2023, according to a report from TrendForce Research. The delay has allegedly affected TSMC's investments in expanding N3-capable production capacity.
Apple and Intel were expected to be the first companies to adopt TSMC's N3 production node in the second half of 2022, with initial chips set to be delivered in early 2023. But Intel's first N3 product was projected to be GPU tile for its 14th Generation Core 'Meteor Lake' processors that are expected to be released in fall 2023. As a result, it postponed the start of its GPU tile high volume manufacturing to the first half of 2023 and recently delayed it further into 2023, according to TrendForce.
Based on Intel's new timing, we can speculate that the company might have decided to skip TSMC's N3 fabrication process and use N3E manufacturing technology instead. N3E promises an improved process window and potentially higher yields at the cost of a bit lower transistor density. We do not know whether this is the case, but it might make sense for Intel.
Earlier today, Intel affirmed Meteor Lake's 2023 launch date, repeating statements from CEO Pat Gelsinger in the company's Q2 2022 earnings call: "In 2023, we will deliver our first disaggregated CPU built on Intel 4, Meteor Lake, which is showing good health in both our and our customers' labs."
Since Intel had reportedly adjusted its 2023 outsourcing plan and notified TSMC about it, the foundry had to adjust its N3-capable capacity expansion, affecting its CapEx in 2022 and 2023.
But while Intel altered its plans for N3, other large customers of TSMC will proceed with the adoption of the company's 3nm-class nodes following their roadmap. Apple is expected to be the first and only user of TSMC's N3 node in 2H 2022 – early 2023. Companies like AMD, MediaTek, and Qualcomm intend to start using TSMC's N3E and N3P fabrication processes in 2024. As a result, TSMC will still have to invest quite a lot in its N3-capable manufacturing capacity in 2022 – 2023 timeframe.
However, remember that we are dealing with an unofficial source and take the information with a grain of salt since plans tend to change.
Stay on the Cutting Edge
Join the experts who read Tom's Hardware for the inside track on enthusiast PC tech news — and have for over 25 years. We'll send breaking news and in-depth reviews of CPUs, GPUs, AI, maker hardware and more straight to your inbox.
Anton Shilov is a Freelance News Writer at Tom’s Hardware US. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
TSMC’s 3nm yield issue and delay is well known. It is rumored that the 5th version aka N3E aka 3nm++++ will have acceptable yield. So most customers like Intel, AMD, etc. might use that since that is more cost effective. There is a chance that Apple might opt for the N3B aka 3nm+ since Apple is willing to spend more money and take the yield hit.Reply
I feel Intel must be panicking now. They are facing delays across most of their core products. The more delays they face, the more competitors are going to chip market shares away from them. I think this is the pitfall when Intel was on a roll announcing many “exciting” products/ roadmaps, but now delaying one after another.Reply
Everybody is decreasing production right now due to the economy so who is going to chip away any marketshare?Reply
A competing company would have to take an enormous risk and increase production in a time where sales are most probably going to go down just to in the end not make any difference at all to intels bottom line because whenever they are finally going to release something they are going to sell huge numbers anyway.