I’ve been staring at the sheer volume of capital pouring into AI data centers lately, and honestly, the math hasn’t been mathing for me. Valuations are sky-high. Burn rates are even higher. It begs the question: Is this sustainable?
The latest course on EverdayCPE, “AI Booms, Busts, and Irrational Exuberance,” digs into a fascinating thesis from FT Alphaville. I went down the rabbit hole, cross-referencing the video transcript with the original FT research and some data from McKinsey.
The conclusion isn’t what you might expect. It’s not about which stock to pick. It’s about whether these AI companies are actually worth more to society dead than alive.
Here is my look at the economics of the AI bubble and what I learned from the course.
The “Dead Than Alive” Theory
The central argument Brian presents is historical. He compares the current AI “YOLO economy” to the railroad boom of the 1800s or the telecom bubble of the 1990s.
In both previous cases, massive amounts of private capital were incinerated. Investors lost their shirts. Companies went bankrupt. But what was left behind?
- Railroads: A national transportation network that powered the industrial revolution.
- Telecom: Miles of dark fiber that eventually enabled the modern internet.
The FT Alphaville report suggests we are seeing the same pattern with AI. We are currently in a phase of “irrational exuberance.” Investors are funding billions in infrastructure—GPUs, data centers, and power grid upgrades.
If the bubble bursts and these AI startups fail, the physical assets don’t disappear. Society inherits a modernized power grid and massive compute infrastructure at pennies on the dollar.
The Data: Burning Cash for the Public Good?
To understand the scale of this, I looked at the numbers discussed in the course materials.
- Projected Investment: We are looking at roughly $1 trillion in capital expenditures.
- Price Compression: A McKinsey paper referenced in the research notes that “neoclouds” (GPU rental companies) are pricing services up to 85% cheaper than hyperscalers like Amazon or Microsoft.
This price war creates a race to the bottom. It’s terrible for profit margins but excellent for innovation. It means cheap compute for researchers, universities, and small businesses.
The US strategy seems to be letting private capital take the hit. The White House is essentially stepping back and letting companies “burn” money to build out national capabilities.
US vs. China: Two Different Gambles
One of the most interesting parts of Brian’s analysis was the divergence in global strategy.
- The US Approach: Encourage aggressive risk-taking. Let companies “YOLO” their way into massive debt to build the best frontier models. If they crash, the infrastructure remains.
- The China Approach: Embrace commoditization. Instead of spending billions on Capex to be first, they are focusing on open-source models. They are drafting behind the US leaders, staying one or two steps back but avoiding the massive financial risk.
Risks to Watch
While the “society benefits” theory is optimistic, the short-term financial reality is messy. The course highlights a few critical risks identified by Fitch Ratings and Barron’s:
- Speculative Building: Developers are breaking ground on data centers without tenants signed up.
- Power Shortages: There isn’t enough juice. If every ordered Nvidia chip went online tomorrow, the grid would crash.
- Corporate Debt: Companies are starting to borrow against their equity to buy GPUs. That is a dangerous leverage game.
Key Takeaways
After reviewing the transcript and the data, here is what stuck with me:
- Infrastructure Outlasts Equity: Even if AI companies go bust, the hardware and power improvements remain a net positive for society.
- The “YOLO” Factor: We are in a phase of reckless spending, driven by a fear of missing out, which drives down the cost of compute for everyone else.
- Commoditization is Coming: The moat for AI models is shallow. The real value may eventually be in the physical infrastructure, not the code.
Want to earn CPE for this topic?
- Compare Options: See how we stack up against others in our 2025 Flexible CPE Guide
- Understand the Format: Read how Nano-Learning works for CPAs.
- Check Your State: Ensure you are compliant with our State Requirements Guide.
- What is EverydayCPE?
Related Courses:
