Memory and CPUs emerge as AI bottlenecks
The guest argued that hardware bottlenecks are shifting to high-bandwidth memory (HBM) and enterprise CPUs, making their suppliers prime beneficiaries of the agentic AI wave.
The argument
While GPUs generate tokens, agent loops and harnesses run primarily on CPUs, driving unexpected demand for enterprise processors. Additionally, memory remains a severe physical bottleneck, leading the guest to suggest Micron could surpass Meta in market value within 12 months.
The thesis, stress-tested
✓ What validates it
- ✓Micron's market capitalization surpassing Meta's within the next year
- ✓Intel or AMD reporting significant revenue growth specifically from enterprise CPU sales driven by AI agents
▸ Risks discussed
- ▸Oversupply of HBM as manufacturing capacity catches up
- ▸GPU architectures integrating more native CPU and memory capabilities
Hear it yourself
"Is that bad or good for you that they learn from you and adapt? It's both good and bad in the sense. I knew this, like, around end of twenty twenty four this is gonna happen. So it never caught me by surprise at all. It was just a matter of time. I still am surprised that the quality is still not there because I regularly test every product out there. But I'm happy that, honestly, they changed Google to be what it should be. I believe that the frontier is where the money is. The frontier in AI is not about answering questions anymore. It's about actually going and doing work for you."
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE