The $15 Trillion AI Shock And China's Threat to US Valuations

"I explained to Lee that a 10 to 15 trillion dollar drawdown is plausible if the largest US technology names were to lose one third to one half of their value. The impact would not end with public companies. Once valuations on firms such as OpenAI reset, private market marks would shift even harder as investors already view them as priced above comparable public peers."

"The most destabilizing factor in our discussion was not domestic. Companies are beginning to experiment with Chinese open source AI systems such as DeepSeek. These models are royalty free and modifiable, and some early experiences suggest performance is already close to OpenAI outputs. The implications broaden if claims about Chinese chip efficiency prove correct. If their hardware can deliver comparable results without the cost and power profile of Nvidia's Blackwell line (NASDAQ: NVDA), the foundational assumptions behind US AI infrastructure demand change materially."

"From there the conversation turned to utilities. Their valuations have been elevated by expectations of long term revenue tied to data center expansions. If demand signals reverse, those stocks would reprice sharply. We also discussed how anticipated job creation attached to nationwide data center construction could evaporate. A range of fifty thousand to one hundred thousand jobs was referenced as vulnerable if AI infrastructure buildouts slow."

A 10 to 15 trillion dollar market drawdown is plausible if the largest US technology names lose one-third to one-half of their value. Private market valuations would reset more sharply after public-market repricing, particularly for firms priced above public peers. Utilities with valuations tied to data center expansion revenue would reprice sharply if demand expectations reverse. Anticipated nationwide data center buildouts could put fifty thousand to one hundred thousand jobs at risk. Chinese open-source AI systems like DeepSeek are emerging with royalty-free, modifiable models approaching Western performance, and claims of more efficient Chinese chips plus direct government capital could materially change US AI infrastructure demand.

Read at 24/7 Wall St.