Manias, Panics, & Crashes: Can We Learn to Avoid Great Financial Mistakes? | Global Conference 2026

Panel Context & Core Thesis: A panel moderated by James McIntosh (WSJ) discussed the book Manias, Panics, and Crashes by Charles Kindleberger in the context of the current AI investment landscape, specifically addressing whether AI constitutes a bubble or a justified technological revolution.

• While the consensus at the event leans toward "no bubble," significant disagreement exists regarding valuations, business models, and the timing of productivity gains.
• The discussion highlights a divergence between the "maximalist" view of AI as a general-purpose technology transforming the economy and the "skeptical" view of it as a capital-intensive experiment with unclear value capture.

Cathy Wood (Arc Investment) Argument: A Convergence of Technologies and Deflationary Trends

• Asserts that unlike the 1999 tech bubble where technologies were premature, AI is part of a mature convergence of five major platforms: AI, robotics, energy storage, blockchain, and multi-omic sequencing.
• Cites Wright's Law (cost declines per cumulative doubling of units) to argue that AI training costs have dropped 75% and inference costs up to 95% per cumulative doubling.
• Predicts sustainable real GDP growth in the 6–8% range driven by productivity acceleration, which would simultaneously drive inflation down.
• Claims to have invested in OpenAI, Anthropic, and xAI, noting a surge in usage where chief futurists are willing to pay $2,000/month for autonomous agents.
• Projects SpaceX valuation targets of $2.5 trillion by 2030, factoring in a potential 20–30x revenue increase from orbital data centers.

Sebastian Thrun (Stanford AI) Argument: Excellent Technology, Flawed Business Model

• Describes current AI technology as an "A-plus" innovation but foundation model business models as a "C-minus," particularly regarding user monetization.
• Notes that while models have billions of users, only ~5% pay, forcing pivots to enterprise subsidies and private equity.
• References leaked internal documents from OpenAI estimating a capital burn of $660 billion between now and 2030.
• Questions the long-term pricing power in generative AI, suggesting it may become a commodity infrastructure with no winner-takes-all dynamic due to a lack of network effects.
• Argues that unlike the internet (where ISPs didn't capture value) or mobile (where telcos didn't capture value), AI foundation builders may face a similar "commoditization" risk without clear pricing power.

Austin Hill (S&P Global) Argument: The "Chicken vs. Egg" of Productivity and Fed Policy

• Distinguishes between a "mania" (exuberance) and a "bubble" (popping), suggesting AI could be a justified mania if productivity gains are real.
• Warns that if productivity growth is "expected" by markets now, the Federal Reserve may need to raise rates to prevent overheating, rather than cutting them.
• Draws a parallel to the 1990s: Alan Greenspan kept rates low initially, but when productivity became undeniable, the Fed tightened aggressively, which arguably contributed to the late-90s bubble burst.
• Highlights the risk of "circular deals" between hyperscalers, compute suppliers, and model builders, which could mask true solvency and contagion risks.

Benedict Evans Argument: Fundamental Uncertainty in Technology Trajectories

• Argues that unlike previous platform shifts (PCs, internet), there is no theoretical model for how LLMs work or what their limits are, making "vibes forecasting" inevitable.
• Notes that AI investments are massively behind demand, unlike fiber optics which were ahead of demand, but questions how long efficiency gains can outpace exponential demand growth.
• Emphasizes the structural difference between a "good technology" and a "good company," suggesting many AI companies may fail to capture value despite technical success.
• Raises the possibility of "proof of humanhood" becoming a regulatory or consumer requirement, which could slow the momentum of agentic AI deployment.

Zoe Chafkin (Morgan Stanley) Argument: Liquidity Risk and Valuation Timing

• States that while valuations (P/E multiples in high 30s) resemble past bubbles (2000, 2007), the critical risk factor is "timing" and liquidity availability.
• Warns that the global sovereign debt-to-GDP ratio is at historical highs, and if hyperscalers must raise massive capital in the public markets, it could trigger a yield curve inversion similar to the Liz Truss UK gilt crisis.
• Advocates for strict risk management and stress testing, noting that the 2008 crisis was exacerbated by a shift from "mark-to-model" to "mark-to-market" without adequate liquidity.
• Highlights the danger of "circular deals" in private markets where companies are desperate to participate in funding rounds, preventing short sellers from exposing inflated valuations.

Sebastian Thrun & Austin Hill: Specific Industry Observations

• SaaS Dynamics: Thrun notes the "SaaS-pocalypse" theory is flawed; enterprise software providers may survive by embedding AI (e.g., Claude) rather than being replaced by it.
• Autonomous Driving: Wood predicts a "winner-take-most" scenario in transportation favoring vertically integrated players like Tesla over Waymo due to lower cost structures; Evans counters that autonomous driving predictions have been "next year" for 15 years, indicating low predictive value in current roadmaps.
• Hardware Shifts: Agentic AI is increasing the CPU-to-GPU ratio, driving demand for CPUs (Intel, AMD) back to parity with GPUs, contrary to previous assumptions of GPU dominance.
• Healthcare Convergence: Wood points to single-cell sequencing and AI-driven drug discovery (curing sickle cell and beta thalassemia) as a concrete, near-term application of the tech convergence.

Future Outlook & Risks

• Inequality & Labor: The panel discussed the "K-shaped" potential of AI, with Evans dismissing the "feudalism" scenario as overplayed, though acknowledging potential wage divergence for those who embrace vs. reject the tech.
• Private Market Illiquidity: Chafkin warns that private markets lack the shorting mechanism of public markets, creating "club deals" where high valuations cannot be challenged, increasing systemic risk.
• IPO Wave: A massive wave of IPOs is anticipated from private AI companies (OpenAI, Anthropic) and SpaceX, which could saturate public markets and expose investors to the risk of public skepticism toward private valuations.
• Key Indicators for a Bubble Burst: Panelists identified watching sovereign yield curves, secondary market valuation discounts (where paper value exceeds trade value), and the ability of hyperscalers to secure capital as the primary signals of a potential crash.