Why AI is the Most Crowded Trade in History (Since 1870) | Saurabh Mukherjea x Ishaan Chitale

Nuggets

Why AI is the Most Crowded Trade in History (Since 1870) | Saurabh Mukherjea x Ishaan Chitale · Nuggets

The Anatomy of an AI Bubble & The Lessons of FOMO [00:02:46 - 00:05:39]
- The Isaac Newton Precedent: Isaac Newton, a highly successful investor in the 17th/18th century, initially evaluated valuations rationally during the South Sea Bubble of 1720 and exited early with tidy profits. However, unable to withstand the psychological pressure of watching his peers accumulate immense wealth, he succumbed to Fear of Missing Out (FOMO), reinvested at the absolute peak, and was completely wiped out [00:03:40].
- The Madness of Man: Newton’s post-crash reflection—“I could calculate the motion of heavenly bodies, but not the madness of man”—serves as a warning to modern institutional and retail investors caught up in the "first day, first show" market mania [00:04:30].
- The Gartner Hype Cycle: Every major technology boom over the past 300 years follows the exact same trajectory: a technology trigger initiates a vertical rise toward the Peak of Inflated Expectations, which inevitably faces a correction as economic reality sets in. Current sentiment matches the peak of this curve [00:05:01].
The Capital Expenditure Arms Race [00:05:39 - 00:08:27]
- Hyper-Scaler Spending Explosion: Since OpenAI launched ChatGPT in November 2022, hyper-scalers (Google, Meta, Elon Musk's xAI, Anthropic, Amazon, and Microsoft) have entered a massive infrastructure arms race. Total capital expenditure (CapEx) for these giants is projected to reach $700 billion to $800 billion in 2026 alone [00:06:24], [00:08:07].
- The Massive Weight of Infrastructure: Constructing one gigawatt of modern data center capacity costs approximately $50 billion, with $35 billion allocated exclusively to purchasing specialized GPUs [00:07:17].
Indices and the Illusion of Diversification [00:08:27 - 00:12:36]
- The Shovel Seller Wins: In any classic gold rush, those who make reliable fortunes are not the frantic prospectors digging for gold, but the entities selling the picks and shovels. Nvidia, led by Jensen Huang, is the ultimate shovel provider, driving an 11x expansion in net profits over three years [00:08:50].
- Index Concentration Traps: Everyday investors purchasing traditional, market-cap-weighted index products like the S&P 500 or Nasdaq-100 believing they are achieving diversification are actually placing highly concentrated momentum bets on a single technological storyline [00:11:49].
- Active Portfolio Underweighting: Recognizing that the theme occupies roughly half the practical weight of major indices, Marcellus maintains a disciplined stance by holding only a 30% exposure to the AI and semiconductor ecosystem within its global portfolio [00:11:14].
Extreme Valuations & Historical Parallels [00:12:36 - 00:17:01]
- The Perfection for Ever Pricing: The global market has ceased pricing in a few strong cyclical years for hardware and component manufacturers. For example, memory semiconductor suppliers like SK Hynix historically run on cross-cycle operating margins of 5% to 10%. They are currently executing at 80% to 90% operating margins, and their valuation multiples assume these peak numbers will persist permanently without cyclical interruptions [00:13:46].
- The 1870 Railroad Mania Parallel: Present-day asset concentration in a singular segment has surpassed the peak of the 2000 dot-com bubble. To find a similar scale of sectoral concentration where 40% to 50% of total market cap comes from one thematic infrastructure build, financial history must go back 150 years to the US Railroad Mania of 1870 [00:15:58].
The Risk of Commodity-Level Compression [00:17:01 - 00:26:49]
- Productivity Gains Latency: Emerging corporate data indicates that while enterprises are pouring billions of dollars into AI API tokens, they are struggling to realize the proportional internal productivity or efficiency gains originally expected to justify the costs [00:21:15].
- Pricing Power Degradation: Large corporate clients have begun pushing back against LLM developers, forcing price cuts on token delivery. As distinct foundational large language models begin to look functionally identical, the space faces rapid commoditization and margin compression [00:23:23].
- The True AI Gap: Citing Sebastian Mallaby’s book The Infinity Machine (focused on Demis Hassabis and Google DeepMind), true general intelligence requires an elegant combination of both Deep Learning (neural networks) and Reinforcement Learning (learning directly from errors over time). The accessible tools driving current corporate CapEx have not yet cracked this combination, accelerating the commoditization of simpler operational AI model wrappers [].
Portfolio Positioning & The Value of Discipline [00:26:49 - 00:38:23]
- The 2000 Bubble Comparison: A critical difference between 2000 and the present is that dot-com leaders were cash-burning startups, whereas today's tech leaders are highly profitable, cash-positive powerhouses [00:30:10]. A total 80-90% wipeout is unlikely; instead, a long, painful rotation out of glamorous names and toward neglected quality is expected [00:31:14].
- Strategic Allocations: Marcellus's global strategy focuses on under-owning the crowded mega-cap tech space and reallocating capital into small-and-mid-cap compounders across Europe and America that steadily grow earnings in the high teens at sensible valuations [00:32:29].

Why AI is the Most Crowded Trade in History (Since 1870) | Saurabh Mukherjea x Ishaan Chitale

1. Executive Briefing (TL;DR)

References

More nuggets

BofA: From sports to pop culture: Prediction markets $1 trillion bet | 14 Jul 2026

The Bridge Ep. 13: Good Credit Holds. Bad Credit Breaks | iCapital

Risk, Resilience, and Relationships Ft. Alona Gornick (MD & Senior Investment Strategist, Churchill Asset Management) |16 Jul 2026 | S&P Global Market Intelligence

Why AI is the Most Crowded Trade in History (Since 1870) | Saurabh Mukherjea x Ishaan Chitale

1. Executive Briefing (TL;DR)

References

More nuggets

BofA: From sports to pop culture: Prediction markets $1 trillion bet | 14 Jul 2026

The Bridge Ep. 13: Good Credit Holds. Bad Credit Breaks | iCapital

Risk, Resilience, and Relationships Ft. Alona Gornick (MD & Senior Investment Strategist, Churchill Asset Management) |16 Jul 2026 | S&P Global Market Intelligence

2. Speaker Profiles & Context

3. Thematic Deep Dives

The Anatomy of an AI Bubble & The Lessons of FOMO [00:02:46 - 00:05:39]

The Capital Expenditure Arms Race [00:05:39 - 00:08:27]

Indices and the Illusion of Diversification [00:08:27 - 00:12:36]

Extreme Valuations & Historical Parallels [00:12:36 - 00:17:01]

The Risk of Commodity-Level Compression [00:17:01 - 00:26:49]

Portfolio Positioning & The Value of Discipline [00:26:49 - 00:38:23]

How Chef Daniel Boulud scaled a restaurant empire with intention | 9 Jul 2026 | Capital Group