Fireside Chat, Interview
AI: What Investors Should Know
- China is projected to potentially dominate the AI model layer as a low-cost producer, while the US is favored to lead the chip layer and its associated major applications; the application layer's winner remains contingent on enterprise cost savings.
- Consumer AI adoption is robust with over one billion daily active users, whereas enterprise adoption has been disappointing and requires significant disruption of the labor force to justify infrastructure investments.
- Global capital expenditures are expected to exceed $3 trillion by the end of 2026, driven by the need for wholesale enterprise adoption to provide a return on invested capital.
- Current enterprise challenges include a disconnect between executive perception and line worker utility, difficulties in structuring diverse data sets, hallucinations, and the risk of losing experienced judgment as entry-level roles are displaced.
- The economics of AI are predicted to rely on replacing a significant portion of the workforce rather than merely enhancing existing employee efficiency, though mass replacement is not expected; instead, technology will aim to improve the efficiency of current staff.
- Public markets currently exhibit an earnings bubble due to over-earning, while private markets show valuation bubbles for companies with unproven business models and robust valuations.
- Circular investing carries substantial risk until business economics improve, and several major AI companies may remain unprofitable for four to five years, creating potential repercussions if a large IPO fails.
- Skepticism in the AI sector is considered beneficial for market insulation, with a prediction that only a couple of dominant companies will emerge in each of the US and China regions.
- Technology trends are shifting toward Small Language Models (SLMs) rather than single large language models, with SLMs demonstrating progress, reduced hallucination, lower power consumption, and greater efficiency.