Will US debt burst the AI bubble? With Ruchir Sharma

Episode Summary

Executive Summary: Rishir Sharma argues the AI boom already shows classic bubble traits—high valuations, broad ownership, heavy investment, and rising leverage—and that the decisive threat may not be AI itself but US Treasury yields. With the government and AI firms competing for scarce capital, a sustained 10-year yield above 5% could raise financing costs, pressure tech investment, and puncture the boom unless growth and inflation improve quickly.

Main Topics: AI bubble diagnostics using the 'four O's' (Priority: 5/5): Sharma applies his bubble framework—overvaluation, over-ownership, over-investment, and over-leverage—to AI and concludes that all four are present to varying degrees, with leverage still the least acute but worsening. US national debt as the hidden trigger (Priority: 5/5): The episode’s central thesis is that America's swelling public debt, not AI fundamentals alone, may be the macro pin that pops the AI bubble by competing for capital and pushing yields higher. Rising Treasury yields and debt sustainability (Priority: 5/5): The conversation details how higher interest rates raise the government’s debt-service burden, worsen fiscal dynamics, and set a floor for borrowing costs throughout the economy. Financing gap in the AI build-out (Priority: 5/5): AI infrastructure spending is far ahead of AI revenue, forcing companies to rely on capital markets. This turns the AI story into a financing story dependent on cheap capital. The 5% 10-year yield threshold (Priority: 4/5): Sharma identifies a sustained move above 5% in US 10-year Treasury yields as a regime change that could make funding AI expansion materially harder. Policy constraints and investor strategy (Priority: 3/5): The discussion ends with limited policy options—lower deficits, higher growth, lower inflation—and Sharma advises investors to diversify away from concentrated AI exposure.

Key Arguments: AI exhibits the hallmark signs of a bubble because valuations are extremely stretched, ownership is unusually concentrated, investment spending has surged, and leverage is rising fast. US public debt is now large enough that government borrowing competes directly with AI companies for the same pool of savings and market funding. If Treasury yields remain elevated or break decisively above 5%, they could make AI expansion harder to finance and reduce investor risk appetite. The AI boom is no longer just a technology story; it has become a capital-markets story driven by access to debt and equity funding. The government’s debt-service burden is escalating to the point that it is absorbing a growing share of tax revenue and could crowd out private borrowing. A sustained productivity boom from AI could offset debt pressures, but there is not yet strong evidence that AI is lifting productivity enough. History suggests bubbles usually end with a monetary tightening event, and higher rates are the modern equivalent of the pin that pops them. For investors, the key risk is concentration: many portfolios are exposed to AI both through equities and through AI-company debt.

Data Points: AI bubble risk score: 7 to 8 out of 10 - Sharma’s rating of how close the AI bubble is to bursting US national debt: about 100% of GDP - Current debt level by standard GDP comparison US gross debt: well over 100% of GDP - Alternative measure cited for total government debt US debt-to-GDP at start of century: about 37% - How much the ratio has risen since 2000 Annual debt interest payments: more than 3% of GDP / nearly $1 trillion - Current annual cost of servicing US debt US defense spending: about $800 billion - Used to compare with interest expense Tax revenue share spent on debt service: 1 in 5 - Up from 1 in 10 at the start of the century Tax revenue share spent on debt service at start of century: 1 in 10 - Historical comparison 10-year Treasury yield at start of year: just under 4% - Average level cited in mid-February 10-year Treasury yield currently: about 4.8% - Current benchmark borrowing rate discussed AI infrastructure spending: around $1 trillion per year - Estimated global annual spend on AI build-out AI services revenue: about $200 billion per year - Generous estimate of annual AI revenue AI financing gap: at least $800 billion per year - Difference between AI spending and revenue US budget deficit: around 6% of GDP - Current deficit cited for the federal government US government annual borrowing: about $2 trillion - Approximate borrowing requirement implied by the deficit Potential downturn deficit: nearly 10% of GDP - Sharma’s estimate if a mild downturn adds about four percentage points to the deficit Threshold for concern: 5% 10-year yield - Level Sharma says would create a problematic regime shift

Pivotal Quotes: "we're pretty much at seven to eight" — Rishir Sharma: His rating of how close the AI bubble is to bursting "the AI story, which was like a very big technology story, has now increasingly become a capital market. Market story, a financial market story" — Rishir Sharma: Explaining why AI depends on external financing "the pin that brings this all crashing down has nothing to do with AI at all" — Chris Giles: Framing the episode’s core thesis about US debt as the trigger

Implications: Investors should treat AI as a crowded, rate-sensitive trade. If US yields stay high or exceed 5%, financing costs could rise, valuations could compress, and the AI capex boom may slow sharply unless growth and inflation improve.