China is creating cheaper, 'good enough' AI models by training them on the outputs of US frontier models. This technique, called distillation, undercuts the revenue of US AI companies, threatening their ability to service the massive debt from their infrastructure buildout.
Mirroring the 2008 financial crisis, banks are packaging high-risk debt from the AI infrastructure buildout and selling it to pension funds and insurers. This spreads systemic risk into supposedly safe parts of the economy, moving it from bank balance sheets to retail investors' retirements.
A significant portion of AI industry revenue is illusory, consisting of circular payments. For instance, NVIDIA invests in a company like OpenAI, which then uses the funds to buy NVIDIA's chips. This creates the appearance of strong revenue growth while masking the industry's financial fragility.
Large AI firms like Anthropic are advocating for stringent government regulation under the guise of safety. However, these proposed rules also serve to raise the barrier to entry, making it more difficult for cheaper, open-source models and startups to compete, thus protecting the incumbents' market share.
Traditionally viewed as diversified, index funds like the S&P 500 have become concentrated wagers on AI. The top 10 companies, nearly all driven by AI, now make up over 40% of the index's value. This means passive investors are taking on significant, non-obvious, sector-specific risk.
Despite massive corporate spending on AI, an MIT study found that 95% of generative AI projects failed to produce any measurable impact on profits. This stark lack of ROI highlights a major weakness for a capital-intensive industry that is financed almost entirely on debt and future promises.