AI Stock Market Crash: Are We Repeating 1873? | The Week

Episode Overview

• Host George Hahn and guests examine how civil unrest and a bond market "doom loop" in France expose broader crises of fiscal sustainability and intergenerational economic fairness in both Europe and the United States.
• Jessica Tarlov and Scott Galloway analyze the geopolitical fallout of President Trump waiving Russia sanctions two days before they were set to take effect, alongside reports of Starshield speeds being throttled for Ukraine's military.
• Financial historian Liaquat Ahamed compares the 1873 railroad crash to today's AI data center buildout, illustrating how opaque debt, excess capacity, and a corporate "prisoner's dilemma" drive systemic overbuilding.
• Ed Zitron and Ethan Mollick debate the current state of AI, contrasting reports of hundreds of billions of dollars in un-installed Nvidia GPUs sitting in warehouses against the massive, underutilized practical utility of existing AI models.

Key Concepts

• France's Fiscal "Doom Loop": Around 550,000 people marched in France and nearly 2,400 were arrested as protests over school conditions (lack of supplies and teachers) and fuel prices driven by the U.S.–Iran war escalated into broader unrest and looting (which Jessica Tarlov assumes involved bad actors piling on). With an election in about 6 months—where far-right candidate Marine Le Pen polls at 37% to 38% with an edge over hard-left rival Jean-Luc Mélenchon and nobody in the political middle—France's 10-year bond yield hit 4.9% and national debt is expected to reach 122% of GDP next year. Macquarie Group strategist Thierry Wizman notes there is potentially a connection between the bond market and the streets: bond pressure forces politicians to discuss spending cuts, sparking protests that hurt business revenues, GDP, and tax receipts, which further damages debt repayment and increases bond market pressure. Resolving this requires either strict austerity or political dissolution of France's ties to the EU and Eurozone (which impose spending constraints via the Growth and Stability Pact and the Maastricht Treaty).
• Intergenerational Wealth Transfers: France's government spending is about 57% of its GDP (the highest share in the OECD), with pensions accounting for a quarter (25%) of that spending. Scott Galloway argues France kept its promise that older citizens would never have to sacrifice while breaking its promise to immigrants that they would be French. He parallels this to the U.S., where young and working-age people transfer $1.5 trillion annually via Social Security to the wealthiest generation in history, and where the U.S. currently spends more money on ICE than on children.
• U.S. Realignment on Russia and Battlefield Tech Vulnerabilities: After President Trump waived new Russia sanctions and tariffs 2 days before they took effect under a law he signed, Vladimir Putin proposed a summit with Trump and Xi Jinping without Ukrainian representation (Tarlov notes we will see if Trump says yes). This drew rebuke from The Wall Street Journal editorial board ("To Russia With Love, Signed Donald") and Andy McCarthy in National Review. Simultaneously, Ukraine's military lost speed on Starshield (the military version of Starlink); The Telegraph reported via one source that Elon Musk pitched the slowdown to Trump and Trump agreed, though both men denied responsibility. Galloway calls them "traitors to the West" for actions resulting in Ukrainian deaths, and notes that most cyberattacks come from Russia—pointing out that Tucker Carlson's engagement declined 72% the day after Yandex server farms were bombed in Russia.
• The 1873 Railroad Crash and the AI Prisoner's Dilemma: Although Microsoft CEO Satya Nadella called Liaquat Ahamed's book on the 1873 crash the "book to read right now" and claimed Microsoft is being careful, Ahamed explains that individual caution probably cannot prevent an overbuilding boom due to a prisoner's dilemma: every tech company fears losing market share if it slows construction, aiming to be one of the 2 or 3 survivors. Ahamed highlights three parallels between 1873 and AI data centers:
• Opaque Debt: Hyperscalers have shifted from funding buildouts via earnings to using debt, with Nvidia at the center essentially lending to its customers, though Ahamed notes we do not actually know how much construction is debt-financed because it is opaque.
• Excess Capacity: By the end of 1872, excess railroad capacity meant only a quarter (25%) of railroad companies could pay dividends; Ahamed supposes the modern equivalent is falling prices for data center output (tokens).
• Financing Spikes: A 1872–1873 European stock market bubble collapse caused railroad bond borrowing rates to spike from 5%–6% up to 9%.
• The AI Hardware Glut vs. Capability Overhang: A Business Insider report by Eugene Kim revealed Amazon is buying Nvidia server racks faster than it can install them, leaving them in warehouses. Wharton professor and AI skeptic Ed Zitron estimates roughly half (50%) of AI chip sales by dollar value are sitting unused, arguing hyperscalers (Microsoft, Google, Amazon, Meta) hold hundreds of billions of dollars in un-installed GPUs due to Nvidia CEO Jensen Huang's "early rack receipt program" and artificial scarcity—though Zitron admits we may never know the exact delivery mechanics without further reporting, and Ed Elson hedges against assuming Jensen manufactured this intentionally. Conversely, Wharton professor Ethan Mollick (who disclaims knowing whether Anthropic rises or falls or if data center debt is problematic) argues that focusing purely on the financial bubble misses how ubiquitous, low-cost intelligence transforms work right now without any future development.
• Sponsor & Promotional Disclosures: The episode includes sponsorships from LinkedIn Ads (citing the 2026 Dreamdata Benchmark Report claim that LinkedIn Ads generated a 121% Return on Ad Spend, the highest of major ad networks, and offering a $250 credit for a $250 spend at linkedin.com/scott) and USAFacts (which Scott Galloway discloses also sponsored his debate with Ben Shapiro on the USAFacts YouTube channel).

Quotes

• At 5:18 - "So there's some people have discussed this in the context of a doom loop, effectively a loop between what the bond market is doing and what the protesters are doing in the street, and one that could only be resolved if the state takes a strong position one way or the other." - Thierry Wizman explaining how fiscal deficits and civil unrest feed into each other in France.
• At 6:53 - "The biggest threat to social stability isn't immigration or aging, it's a system that rewards people for what they already have and punishes those who are just starting out." - Scott Galloway connecting French pension spending and U.S. Social Security transfers to generational economic inequity.
• At 19:23 - "Because each individual railroad company, in the same way that each individual tech company, is trying to be careful. It's the collectivity of their decisions that leads to overbuilding." - Liaquat Ahamed refuting the idea that individual corporate caution by leaders like Satya Nadella can prevent an AI infrastructure bubble.
• At 24:46 - "The second thing I would say for individuals using these systems is you have to try and use them for everything that you can to figure out the shape of the jagged frontier, and you need to be as ambitious as possible, you know, maximalist in how you're using these things to figure out what they're good or bad at, and then you have to make really strong choices about when you use AI and when not to." - Ethan Mollick advising listeners on how to extract real-world value from current AI tools regardless of whether financial markets crash.

Takeaways

• Map AI's "jagged frontier" through maximalist experimentation: Instead of fixating on which AI stocks or debt models will survive, ambitiously test existing AI tools across all your daily workflows to discover firsthand what they excel at and where they fail, then make deliberate choices about where to maintain human involvement.
• Use AI to augment human labor rather than replace it: Avoid firing human creatives and switching entirely to automated outputs; as Ethan Mollick notes, companies that replaced marketers with AI in January now produce uniform, nauseating "Claude speak" on LinkedIn and in articles that damages their brand differentiation.
• Monitor token pricing, borrowing rates, and un-installed inventory as AI bubble indicators: For investors evaluating hyperscalers, look past individual executives' claims of caution and track systemic indicators—specifically whether GPU purchases are sitting un-installed in warehouses, whether the price of data center output (tokens) is falling due to excess capacity, and whether credit markets are forcing borrowing rates higher.