AI compute costs face structural downward pressure
The thesis argued is that the economics of the AI sector face headwinds as companies shift toward lower-cost models and software efficiency, reducing the demand for brute-force hardware compute.
The argument
The speakers discussed how major companies like Microsoft are considering cheaper, less powerful models (including Chinese alternatives) to achieve 90% of their needs at a fraction of the cost. Additionally, a shift in focus from brute-force hardware scaling to software optimization could result in fewer chips being required, impacting future revenue growth for semiconductor companies.
The thesis, stress-tested
✓ What validates it
- ✓Major cloud providers officially announcing the integration of lower-cost or open-source models over premium proprietary ones
- ✓A visible reduction in capital expenditure guidance for data center build-outs in upcoming quarterly earnings reports
▸ Risks discussed
- ▸National and corporate security risks may prevent Western firms from adopting cheaper Chinese models
- ▸Mindless passive ETF flows may continue to prop up mega-cap tech valuations regardless of fundamental shifts
Hear it yourself
"They don't they're not quite as powerful, but we can kinda get 90% of of what we need to get done at a much smaller cost than what we'd get, say, on the entropics or the open AI's of the world."
00:00 / 00:15
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE