Now Playing
Enterprise AI Boom, CPU Inversion & Memory Shortage Through 2027
Audio is available on Spreaker — see link below.
Nvidia's new ACIE segment reveals enterprise and sovereign AI nearly matching hyperscaler spend, while Intel flags a radical CPU-to-GPU ratio flip for agentic workloads. Plus Micron's HBM shortage warning, TSMC outsourcing, and AMD's $10B Taiwan bet.
Audio is available on Spreaker — see link below.
NVIDIA just redrew the map of AI demand. The company split its Data Center reporting into two segments, and what that split reveals is the story of the moment.
Intel's CEO Lip-Bu Tan dropped a number this week that deserves attention. In training workloads, the standard ratio was roughly one CPU for every eight GPUs.
Micron's warning at the JPMorgan TMC Conference deserves to be taken seriously. HBM, DRAM, and NAND tightness will persist past twenty twenty-six.
Micron's response to that pressure is a notable strategic concession. The company is outsourcing the base die for HBM4E to TSMC's three-nanometer process, while upgrading its core DRAM to one-gamma with EUV.
AMD's commitment of over ten billion dollars to Taiwan for advanced packaging, assembly, and foundry partnerships with TSMC and Powertech is the clearest statement yet that packaging capacity is now a strategic moat. AMD, Micron, and Intel are all deepening Taiwan dependencies simultaneously, which concentrates risk but also concentrates capability.
The metrics worth tracking from here are relatively clear. Watch whether NVIDIA's ACIE segment maintains that thirty-one percent quarterly growth rate or whether it pulls back as easy enterprise wins get absorbed.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.