Is the AI Bubble Just Getting Started?
Everyone seems to agree AI is a bubble that's about to burst. But what if everybody's wrong? This week starts with a contrarian column arguing that roughly $1 trillion in AI build-out since 2024 is only about 2% of GDP, while past build-outs like the railroads and telecom reached closer to 25% before they peaked. At today's pace, AI wouldn't get there until around 2030. The hosts test the comparison against the Panic of 1873. Railroad track roughly doubled in the years after the Civil War, and financier Jay Cooke's bank funded long-term railroad loans with short-term borrowing. When a crash in Vienna spread across the Atlantic and depositors wanted their money back, the money wasn't there. From there, they look at today's circular AI financing, where chipmakers, cloud providers and hyperscalers are each other's biggest customers, and at Oracle's large lease commitments tied to OpenAI. Their conclusion: yes, there's a bubble, and nobody knows when it pops. The practical answer is to know your exposure, diversify, and benefit along the way without going all in.
Next up is Meta's new agentic AI, Muse. Colin tested it by asking it to review the podcast. It watched recent episodes and came back with notes on thumbnails, chapters and editing. The hosts walk through Muse's pitch: goal-first instead of a blank prompt, end-to-end encryption from the cryptographer behind WhatsApp's encryption, and passwords kept in a separate vault. They also cover the trade-off underneath it. Letting an assistant shop, schedule and watch on your behalf means handing over access in exchange for ease, and metadata still flows back to the parent company. That leads to a debate over whether AI is really the fastest-adopted technology in history, or whether most people are using it without choosing to every time they run a search.
Then a viral post from a San Francisco tech executive listing her non-negotiables in a partner, including that he be "post-economic," meaning wealthy enough never to worry about money again. The hosts run the odds, then turn to the more useful question of what role financial stability should play in a relationship.
Before the wrap, research showing that only about 23% of S&P 500 stocks outperformed the index over ten years. That's roughly a one-in-four chance for each pick, and a reminder of why buy-and-hold stock picking is harder than the success stories make it look.
The episode closes on a difficult client story. An inheritance of about $100 million lost roughly $40 million to estate taxes because no estate planning was done. The previous advisors hadn't even named beneficiaries on the accounts. Updating beneficiaries takes minutes, and at that level of wealth, the basics should never be missed.
Sources:
"Everybody's Wrong" column, JC at Paradigm (referenced on the recording; not independently verified)
Trivalent / Adam Parker research on 10-year S&P 500 constituent performance (referenced on the recording; not independently verified)
Meta Muse launch coverage and Mark Zuckerberg's AI paper (referenced on the recording; not independently verified)
The Rogers Diffusion of Innovations framework
Follow Us:
Instagram: https://www.instagram.com/compoundgrowthpod
Wheeler’s LinkedIn: https://www.linkedin.com/in/wheeler-crowley-0a63933b/
Colin’s LinkedIn: https://www.linkedin.com/in/colin-walker-mba-6099a038/
Credits:
Created By: Wheeler Crowley and Colin Walker
Production, Editing and Post-Production: Tori Rothwell
