DeepSeek's shock sends investors running to bonds

A sudden technological shock from Hangzhou shattered the core consensus governing modern equity valuation. The unexpected release of DeepSeek's open-weights model demonstrated state-of-the-art reasoning capabilities at an estimated training cost representing a microscopic fraction of Western frontier models. The market's verdict was immediate and violent: Nvidia suffered a historic single-day market capitalization erasure of $589 billion on 27 January.

The Capital Intensity Illusion

The foundational thesis of the ongoing artificial intelligence cycle was that compute scaling laws were linear and monopolized by massive proprietary clusters. If algorithmic optimization and distillation can replicate frontier performance with radically fewer high-end accelerators, the projected hundreds of billions in enterprise GPU procurement become questionable. Hyperscaler capital expenditure trajectories that assumed perpetual hardware shortages now face the prospect of sudden efficiency deflation.

Flight to Duration and Sovereign Paper

As equity concentration unwound, panicked capital sought refuge in sovereign debt. Benchmark US Treasury yields plunged as fixed-income desks priced out inflation premia and rushed into duration, treating the algorithmic breakthrough as a disinflationary productivity shock. DeepSeek did not merely introduce a new language model; it dismantled the assumption that infinite capital expenditure guarantees a permanent proprietary technological moat, driving capital out of over-leveraged silicon into sovereign bonds.