▶AI is a foundational, transformative technology that will drive significant productivity, economic growth, and disinflation, while simultaneously disrupting incumbent business models like traditional software.May 2026

▶The market for large language models (LLMs) will consolidate around a few key players due to extremely high capital barriers to entry, making it difficult for new competitors to emerge.May 2026

▶Private markets, particularly in late-stage technology, currently offer more attractive and less competitive investment opportunities than increasingly inefficient public markets.May 2026

▶The global economy faces a significant geopolitical risk centered on the semiconductor supply chain, with over 90% of advanced chips produced in Taiwan, creating a potential collision course with China.

▶While Sundheim is bullish on AI's long-term impact, he highlights the central debate over whether the massive capital expenditures required for training models will generate attractive returns, particularly if enterprise adoption proves slower than anticipated.

▶He posits that AI will be a powerful disinflationary force, yet simultaneously warns of a potential "Great Depression"-level economic shock from a semiconductor conflict with China, presenting two powerful but opposing macroeconomic outlooks.

▶Sundheim argues that LLM API businesses are durable and have high gross margins, yet also compares the underlying models to a commodity where the primary long-term moat is personalization, not the core technology itself.May 2026

▶He predicts that large LLM companies will eventually insource their compute needs away from hyperscalers, but this is contingent on them first achieving massive free cash flow, creating a dependency in the medium term before they can challenge their suppliers.