China’s AI crash is warning sign for the AI bubble | Dean Baker
AI can be useful without justifying today’s valuations; test the investment case by asking who will pay, how much, and what measurable value they get.
- Evaluate AI investments by identifying paying customers, realistic prices, and returns—not projected usage alone.
- Treat AI like a business expense: compare cheaper models and keep only uses that deliver value.
- Watch whether demand and revenue come from non-AI businesses; spending by AI companies alone cannot sustain the sector.
The interview is mainly an economic argument; the notes capture its core points without a practical demo. · Best for: Investors and business leaders assessing AI spending, valuations, or infrastructure plans.
Dean Baker argues that AI may continue spreading even if AI stocks fall, but current valuations depend on profits and broad customer demand that have not yet been demonstrated. He sees cheaper competition, uncertain returns, and infrastructure costs as reasons to question investment plans.
Bubble risks and the China comparison0:49
- Baker calls the Chinese AI-stock surge a bubble, while stressing that a financial crash would not erase the technology.
- He contrasts the much smaller Chinese AI market with US AI-related valuations, while noting that his comparison is rough.
- The internet bubble is his example of financial excess collapsing without stopping adoption of the underlying technology.
Why bubbles turn—and price competition7:47
- Baker says he cannot identify a definitive trigger for the Chinese sell-off; shifts in investor sentiment can accelerate a decline.
- The host raises price cuts and token-price competition; Baker says high valuations require companies to sell AI at prices that generate substantial profits.
- Cheaper models may meet business needs, making it difficult for frontier-model providers to sustain high prices.
Customers may treat AI like any other expense11:21
- Baker expects companies to compare AI costs much as they compare suppliers for ordinary inputs.
- He says companies that encouraged employees to use more AI have begun asking what they get for their AI bills and whether cheaper tools will do.
Market spillovers and infrastructure costs12:58
- Baker sees limited direct spillover from China’s market decline because markets are segmented and the Chinese sector is smaller.
- In the US, he expects a bubble collapse to slow investment rather than end it; data-center opposition and costs may also affect projects.
- He argues that AI investment can crowd out other investment and says local resistance could give companies reason to reconsider expensive build-outs.
The demand test for AI valuations23:27
- Baker says AI needs paying customers outside the AI sector; companies buying from one another do not establish broad, sustainable demand.
- He sees no evidence yet of the economy-wide productivity shift that AI promoters’ valuations assume.
- His practical test: ask who will buy the product, what they will pay, and whether the value justifies the investment.
Notes are AI-generated from the video's transcript and may miss nuance.