LLM-based applications are structurally unprofitable
Applications and services built on top of large language models are fundamentally uncommercial and cannot generate sustainable profits.
The argument
The guest argued that LLMs are probabilistic mechanisms that lack the consistency required for true automation, with studies showing low and inconsistent task completion rates. Consequently, companies cannot use them to reliably replace human labor without severely sacrificing product quality, which only monopolies can afford to do.
The thesis, stress-tested
✓ What validates it
- ✓A reduction in capital expenditure guidance for AI infrastructure by major tech monopolies
- ✓An increase in customer churn or public backlash regarding the declining quality of AI-integrated services
▸ Risks discussed
- ▸Monopoly tech firms may continue to subsidize unprofitable AI services using their core monopoly profits
- ▸Breakthroughs in LLM consistency or architecture could theoretically improve task completion rates
Hear it yourself
"And that's one of the reasons why we're starting to see a reduction in deployment of large language models, which is this chart showing adoption rates for large language models on page five of the presentation."
00:00 / 00:15
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE