Now Playing
Multi-Foundry Shift: Google, Microsoft & Apple Redraw the Chip Supply Map
Audio is available on Spreaker — see link below.
The AI chip supply chain is fracturing as Google splits its Icefish TPU across TSMC, Samsung, and Intel simultaneously — and Microsoft and Apple are making the same move. Today's briefing covers Intel 18A yields, Samsung's 2028 profitability target, AMD's Venice performance claims, and Qualcomm's record automotive revenue.
Audio is available on Spreaker — see link below.
Google is splitting its next-generation Icefish TPU across three foundries, and that single decision captures something much larger happening across the entire semiconductor industry right now. The specifics: Google is in talks with Samsung to manufacture the memory I/O die for Icefish using a two-nanometre process.
Intel's eighteen A node is where a lot of this story pivots. Panther Lake yields have hit the fifty-five to seventy-five percent range, which is the threshold that moves a process from promising to commercially credible.
Samsung's role in this multi-foundry picture carries its own complications. The foundry chief has indicated that twenty twenty-seven profitability is unlikely, pushing the turnaround target to twenty twenty-eight.
Away from the foundry storyline, AMD has made a projection worth tracking carefully. Venice, its upcoming Zen six server CPU built on TSMC two-nanometre, is projected to deliver three point three times Nvidia Vera's per-rack performance when it launches later in twenty twenty-six.
Qualcomm is running a different kind of story. Automotive revenue hit a record one point three three billion dollars, up thirty-eight percent year over year.
The watchpoints from here are specific. Can Intel sustain eighteen A yields at commercial scale through twenty twenty-six?
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.