China has won the AI war & US firms will go bust | Andrew Neil x Steve Keen

Overview

The video argues that the U.S. AI boom is at serious risk of turning into a major financial and economic crash—potentially rivaling past “boom-and-bust” episodes. Meanwhile, China is positioned to win the AI race through cost advantages, energy and infrastructure readiness, and a more open (or public-utility-like) approach to AI.

Main arguments and analysis

1) China is outcompeting the U.S. on cost-to-performance

• The video claims the biggest competitive threat to U.S. AI companies is cheaper, more “commoditized” AI from China.
• These models are said to achieve 80–90% of the performance of leading U.S. models at a fraction of the cost.
• It contrasts much higher U.S. valuations (e.g., OpenAI/Anthropic) with lower Chinese valuations despite similar or comparable capabilities.
• This valuation gap is framed as “systemic risk”: valuations may depend on pricing power that disappears once cheaper models spread widely.

2) U.S. AI investment is financed with opaque, fragile “private credit”

• A central claim is that U.S. AI expansion is driven not only by equity and normal debt, but also by private credit—off–balance-sheet, less regulated lending via special purpose vehicles and other mechanisms.
• The video cites a study attributed to the Japanese stock exchange/Nikkei, arguing there is hidden off-balance-sheet debt for major U.S. tech firms.
• It frames this as a possible trigger for a broader financial meltdown:
  • If AI companies fail to generate enough revenue, they may not service the debt.
  • Losses could cascade through the financial system, reflecting patterns seen in earlier crises—especially timing breakdowns tied to private leverage.

3) Energy and infrastructure constraints could stall AI expansion in the U.S.

The video argues that U.S. AI data centers face a severe electricity and grid-connection bottleneck:

• The power grid is portrayed as outdated and insufficient for rapid buildout.
• Even beyond electricity, constraints include:
  • Transformers
  • Substations
  • Transmission lines
  • Skilled labor shortages
• It adds a water dimension using Texas examples:
  • Data centers are claimed to require very large water quantities for cooling, potentially competing with regional water needs.
• The video asserts hyperscalers and AI builders can’t rapidly resolve these constraints, and major new power projects take too long (e.g., years for nuclear; SMRs not soon available).
• It also discusses delays in turbine capacity (for gas/combined-cycle generation), suggesting AI demand is outpacing available generation hardware.

4) Combined “triple whammy” risks could crash AI incumbents even if AI technology survives

• The core thesis is not that AI “fails” as technology, but that the companies and financial structures behind it may collapse.
• The video compares this risk to prior technology cycles where “vanguard” firms went bankrupt (railways, electricity, aviation, dot-com), even as underlying innovations continued.
• It suggests that most U.S. AI companies may not survive—possibly only about 1 in 10.

5) Steve Keen’s macro-financial framing: private debt dynamics drive booms/busts

Professor Steve Keen (University College London) endorses the “private credit is systemic risk” view, linking it to Minsky’s financial instability hypothesis:

• Private leverage fuels overinvestment during booms.
• If cash flows don’t materialize quickly enough, firms borrow again.
• The process continues until timing breaks, triggering a cascading downturn.
• Keen argues mainstream attention on government debt overlooks the private debt mechanism capable of producing crashes.
• He suggests the trigger timeline is likely within the next two years, and that additional global shocks (conflict disruptions and declining capacity in physical production) make debt servicing harder.

6) Chinese strategy: energy readiness + cost optimization + openness

Keen’s rebuttal of U.S. assumptions includes:

• China is said to have already built a large-scale higher-capacity electrical transmission system, including 800-volt infrastructure rather than 400-volt, reducing AI-era energy friction.
• Chinese engineering culture is portrayed as prioritizing reducing the cost of AI computation, achieving strong results via more efficient training/inference rather than brute-force scaling alone.
• Chinese models are described as more open, potentially loadable on less centralized infrastructure, contrasting with U.S. reliance on massive data centers.

7) Potential AI diplomacy is viewed as unlikely; “open vs proprietary” is the real divide

• Even if U.S.–China talks are rumored, Keen doubts they would resolve competition.
• The deeper conflict is framed as:
  • China: open/public utility orientation (more accessible models, broadly available)
  • U.S.: proprietary, profit-driven control (argued to worsen pricing-pressure/bubble dynamics)

Overall conclusion

The video’s message is that the U.S. AI boom reflects classic—and amplified—valuation and financing fragility:

• High expectations
• Heavy leverage via private credit
• Infrastructure constraints (especially energy)

Together, these are said to create crash conditions. China is framed as better positioned to win due to:

• Lower-cost model performance
• More prepared energy infrastructure
• A potentially more open distribution model

Presenters or contributors

• Andrew Neil (presenter)
• Professor Steve Keen (University College London; contributor/interviewee)