AI bottleneck shifts from compute to memory
The AI revolution has entered a second phase focused on memory and long-context agents, making memory hardware the primary supply chain constraint.
The argument
The hosts highlighted the rapid growth of the DRAM ETF, which is on track to become the fastest ETF to reach $10 billion in assets. They argued that while the first phase of AI was about compute (NVIDIA), the current 'agent' phase relies heavily on memory capacity, benefiting companies like Micron.
The thesis, stress-tested
✓ What validates it
- ✓DRAM ETF officially crossing the $10 billion asset threshold within its first 49 days
- ✓Micron reporting elevated average selling prices (ASPs) for high-bandwidth memory (HBM)
▸ Risks discussed
- ▸Cyclical downturns in the semiconductor memory market
- ▸Overvaluation of memory stocks due to speculative ETF inflows
Hear it yourself
"That's the highest since July. The thirty year treasury yield is above 5%, which is and, somebody's tweeted this out. Like, this is Trump taco territory. Yep. So around 5% yields is when Trump starts to, like, freak out, sweat bullets, and, like, do something specifically to lower yields, whatever he can. But And we know that because we saw this with the tariff scare around this time last year. It's when bond yields 30 people. This is when Trump bends. Yeah. He backed down. Right? And so will he back down on something with It won't what will he do to do to bend over? I don't really know."
11:30
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE