China Is About To Pop The AI Bubble

Summary

Key Points

• —The US stock market, including retirement funds, is currently propped up by the narrative that American tech companies will generate trillions in profits from AI, a story that may be ending. 0:00
• —A significant lack of trust in AI exists among businesses and governments, highlighted by incidents like Anthropic being ordered to cut off foreign access to its models and concerns over data ownership, security, and AI "hallucinations." 1:23
• —The AI business model is fundamentally broken because, unlike traditional software, every user interaction incurs direct costs (electricity, chip wear), meaning increased customers lead to linearly increasing expenses rather than expanding profit margins. 12:15
• —Major AI players like OpenAI are burning billions of dollars annually, and even Nvidia's sales are partially self-funded through complex arrangements with "NeoClouds," indicating a lack of genuine, diverse market demand. 15:17
• —Hyperscalers like Microsoft, Google, Amazon, and Meta are investing trillions in AI but are not disclosing AI-specific revenues, suggesting these ventures are losing money and their overall growth is driven by other business segments. 15:44
• —China presents a formidable and significantly cheaper alternative, outspending the US by almost 10:1 in AI development while offering models that are 7-12 times more cost-effective and nearly comparable in quality, often by "distilling" or copying US-developed models. 17:27
• —The AI bubble's potential pop is unlikely to await companies stopping capital expenditures; instead, it could be triggered by a major tech CEO announcing a moderation in infrastructure investment, which would give others permission to follow suit. 23:01
• —While bond market credit spreads currently appear calm, they were similarly tight before the 2008 crisis, indicating they measure belief rather than objective truth, and Michael Burry points to chip stocks at peak valuations and declining AI token prices as concerning signs. 26:57
• —The market is already showing a shift towards these cheaper AI models, evidenced by a 20% decline in the price of AI tokens and the emergence of highly capable AI from unexpected Chinese companies like a food delivery service. 29:42