SK Hynix is projecting 576% year-over-year profit growth on 78% operating margins as HBM4 dominates AI memory supply — but the post-2028 supply cliff is real. Today's briefing also covers Intel's agentic CPU comeback, NVIDIA custom-silicon pressure, and why financing costs are now a binding AI infrastructure constraint.
Audio is available on Spreaker — see link below.
Intel's AI revenue just grew seventy percent year-over-year, and the story behind that number is more interesting than the headline suggests. The reason is agentic AI.
Shift to memory, and the numbers are extraordinary. Market consensus is projecting SK hynix's third-quarter twenty-twenty-six operating profit at seventy-six to seventy-seven trillion Korean won.
The substrate bottleneck adds another layer to this. Japanese substrate suppliers are taking a cautious line on capacity expansion.
Financing is now moving to the center of the infrastructure constraint story. CoreWeave's unsecured bond yields are near fifteen percent.
On competitive dynamics, custom silicon is no longer a fringe concern. Google's TPU, Amazon's Trainium, and OpenAI's own chip efforts are actively eroding NVIDIA's market share.
The through-line across all of this: the AI hardware story is no longer just about who makes the best chip. It's about supply walls in memory, substrate capacity decisions made by cautious Japanese manufacturers, financing costs that are already binding for smaller cloud players, and a CPU demand driver that most investors haven't fully priced.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.