Quick tidbits from the call:
Needless speculation: which CPUs are made in Intel's Oregon fabs? ...13th gen CPUs with via oxidation problems?
- Lunar Lake fabrication was "almost entirely outsourced".
- On-package memory also will hurt Lunar Lake margins.
- Biggest CCG margin killer: Meteor Lake was designed to ramp on Oregon fabs. But Intel made a late-breaking switch to Ireland fabs: problematically, Ireland fabs have "a much higher wafer cost" and costs will stay higher in the next quarter.
- A lot of would-be 18A customers were waiting for the PDK and did not necessarily commit.
Now that I've read through the transcript, here's my updated takes on these points:
- Nothing new to add.
- CFO said that they are offering the on package memory at cost so that hurts margins. Still seems weird that they can't get accretive margins out of the package as a whole. I really don't understand their play here. They say there is a lot of interest for LNL, so they are going to be going much higher volume than originally planned, but at the same time, saying it's a very expensive part and will hurt margins. I could understand a volume play if it was made in house, but being on TSMC, I don't see the logic? Maybe ARL isn't getting the interest they had hoped for?
- OK, for this one, it's not that Ireland has much higher wafer costs in general, it's that ramping the production in Ireland sooner than expected is causing a short term increase in wafers costs. Basically, they had planned a transition from production in Oregon to Ireland in a certain time frame, but then the demand for MTL was higher than expected, so they accelerated the time frame and ramped Ireland quicker than expected, which is causing higher wafer costs. Most likely yield issues related to trying to ramp Ireland more quickly than planned.
- Pat reiterated that their prior $15B in lifetime deals for the fab is still in place, but that most of the money is on the packaging side right now. I know the PDK for 18a (or at least some updates customers were waiting for) was delayed and combine that with the comments from the call about at least a portion of the reduced CapEx being due to installing less equipment than originally planned for customer production, it seems to me that 18a isn't drawing in as many customers as they had initially hoped.