The short story is this rumor is BS and can be easily disproven on two fronts. It is in fact so easy to disprove that the fact that no one seems to have bothered to try annoys us greatly. How can you disprove it? Lets start out with the easy one, process.
Intel Meteor Lake large CPU package
Tight die tiling…
As you can’t really tell from SemiAccurate’s awful pictures, Meteor Lake is multiple dies. Intel has been very public about Meteor being their first major consumer device made on multiple nodes and coming from multiple foundries, some of which are not Intel, TSMC to be specific. While Intel has not commented publicly on what nodes will be used, SemiAccurate has said multiple time, as have many others, that the GPU die will be made on TSMC 5nm. At the recent Vision conference, this was an open ‘secret’.
Now if you go back to the original ‘report’ on this Meteor Lake delay from TrendForce, they talk about TSMC slowing down 3nm expansion because Intel delayed Meteor Lake. Sounds great but, umm, how can we put this delicately, Meteor Lake is not on, and never was on that node. Do we need to spell this out further? Why no one even bothered to check this out is beyond us… no, actually we know how these clickbait sites operate, never mind. Either way I am sure you can do parse the logic here.
The second point is that the so called ‘delay’ to Meteor Lake that Trendforce touts isn’t a delay. How do we know? Well SemiAccurate wrote that Meteor was ‘late’ 2023 on July 11, along with some specs for the CPU. As long time readers are aware, we are not shy about calling Intel on product delays, some might even say we are a tad snarky at times. Even if you carefully parse that story, you will not see the word ‘delay’ used once for Meteor or any other Intel consumer chip. There may be some in the future but none that we are aware of at the moment.
You may be forgiven in thinking that SemiAccurate was a month or so ahead of the rest on the ‘delay’ of Meteor but there is one more data point. SemiAccurate was told of this schedule in Q1 but was asked not to print it until we could get a second source. That took a bit, almost six months to be accurate, but we originally knew Meteor was Q4/23 in Q1/22. If there was a delay it wasn’t recent and nothing has changed from the first hard date SemiAccurate was given early this year. In short from our perspective there is no delay, period. Caveat: Things may change in the future but don’t read anything into our saying this, really.
So there you have it. The original report got the process wrong and blamed it for something else that may or may not be happening, SemiAccurate has no insight on TSMC’s CapEx plans. On top of this the so called delay is nothing more than the roadmap that Intel has been giving to OEMs for 6+ months now, and has not changed in that time. What others wrongly speculated and now have to revise is their problem, the real data does not support their supposition.