AI capital concentration crowding out all other venture sectors while frontier model valuations—reaching $852B–$965B with no revenue base—decouple from fundamentals, transmitting bubble-burst risk to passive retirement capital

The article traces this pattern across railroads (1840s-1873), electricity/radio (1920s-1932), and the internet (1990s-2001), showing that even when the underlying technology fundamentally transforms the world, share prices become 'wildly exaggerated' and generate negative returns for shareholders. The current AI sector, with 10 companies valued at $27 trillion (larger than all of Europe), exhibits the same structural preconditions the article identifies as historically preceding major corrections.

The article uses this valuation comparison as a diagnostic indicator of bubble conditions, asking readers to 'think about what exactly the share price is anticipating (or indeed requires) to generate further gains from here.' The $27 trillion figure for 10 companies versus Europe's 8,000+ companies suggests either unprecedented fundamental value creation or valuation disconnection from historical norms—the article's historical analysis implies the latter is more likely.

The article documents a pattern generalizable beyond China: as AI narratives dominate investor attention, leveraged money concentrates in semiconductor and advanced manufacturing equities, compressing the gap between fundamental valuation and speculative positioning. The 84 billion yuan in net margin purchases since H2 2024 and the sharp single-day reversal illustrate how this concentration creates reflexive volatility — inflows beget inflows until a catalyst triggers rapid deleveraging. This dynamic mirrors patterns seen in US AI-linked equities and suggests a global structural tendency for AI infrastructure stocks to become vehicles for leveraged speculation rather than purely long-term investment.

The article treats China's AI funding frenzy as downstream of US benchmark-setting, not as an independent phenomenon. This pattern — where US tech valuations set global investor expectations and export speculative pricing frameworks — is a recurring structural dynamic applicable beyond China to any market with AI ambitions. The fact that Chinese model companies trade at 100x+ ARR while remaining unprofitable illustrates how thoroughly the earnings-decoupled valuation logic has been imported.

The article documents a reinforcing loop: AI companies are deeply cross-invested (Nvidia-OpenAI-SpaceX-Microsoft-Anthropic), they are entering or about to enter major indices at historically high valuations, and passive fund rules mean retail and pension capital will be automatically exposed. The 1990s comparison is analytically significant because the IT bubble caused broad retirement savings damage — the article implies the current structure is more concentrated, not less. The signal generalizes to any future wave of AI infrastructure IPOs entering indices under relaxed seasoning rules.

The article treats Q1 2026 as an inflection point where AI fundraising has moved from 'large share' to near-total dominance of global venture. This dynamic generalizes: when a single technology paradigm captures supermajority capital flows, it creates both a self-reinforcing moat for incumbents and a systemic vulnerability if the paradigm disappoints. The simultaneous valuation leapfrogging between OpenAI and Anthropic within months further suggests valuations are driven by competitive signaling rather than fundamental cash flow analysis, a pattern historically associated with late-stage bubble dynamics.