Key insights
- The author expresses concern over the massive CapEx investments in LLMs, suggesting it's a potential trap due to accuracy limitations for enterprise applications. They highlight a shift towards deterministic AI, evidenced by a Milken Conference panel featuring ASML and Google execs. The author believes the market is overvaluing generative AI infrastructure and that the real B2B opportunities lie in AI systems that guarantee accuracy, implying a bearish outlook for current LLM-focused tech stocks.
looking at the recent earnings and the sheer amount of CapEx big tech is dumping into scaling LLMs is making me nervous. pouring hundreds of billions into probabilistic models that basically just guess the next word is a wild bet when enterprise clients need 100% accuracy. you cant run a power grid or logistics network on a model that might hallucinate because of a weird prompt was checking out the speaker notes for the Milken Conference to see what the institutional guys are focusing on right now. its pretty telling that the ASML and Google execs are doing a panel with Logical Intelligence entirely focused on deterministic AI (the brief is here https://logicalintelligence.com/milken). seems like the smart money is quietly pivoting if the industry is already moving toward architectures that understand actual mathematical constraints and logic, then pricing in a permanent monopoly for current generative AI infrastructure feels like a mistake. The real b2b money is going to flow into systems that physically cannot hallucinate. just feels like retail is blindly chasing the LLM trade while the actual builders are already looking for the off-ramp.