Overview of BiggerPockets Money: Aswath Damodaran on the $10 Trillion AI Bet
This episode clips a conversation with valuation expert Aswath Damodaran about the scale of AI spending, what kind of revenue would be needed to justify it, and how investors should think about AI exposure in their portfolios. His central argument: the AI boom should be viewed as the construction of a massive “factory” built by one set of companies, while another set hopes to profit from the products and services that factory produces. The big question is whether the eventual AI business will be large enough to justify the $2T+ already being invested.
Main Thesis: AI Is a Factory, Not Just a Stock Story
Damodaran reframes AI as a collective industrial buildout:
- Architecture/factory builders: Nvidia, power companies, data center developers, electrical equipment firms, real estate providers, etc.
- Product/service users: Meta, Alphabet, Amazon, Microsoft, OpenAI, Anthropic, and others trying to monetize AI applications.
His point is that you should not lump all these companies together just because they’re “AI-related.” Some are selling the picks and shovels; others are trying to turn the factory’s output into profit.
The scale of investment
- He estimates $2T to $2.5T has already been invested in AI infrastructure.
- He traces the AI spending race back to November 30, 2022 — the public launch of ChatGPT.
- In his view, this is the largest buildout in business history.
What Revenue Would Justify the Spend?
Damodaran says current AI product/service revenue is nowhere close to what would be needed to justify the capex.
Current scale
- Best-case combined AI product/service revenues: about $250B over the last 12 months.
- That is far below the level needed to support a $2T+ infrastructure build.
His rough long-term threshold
To make the math work, he believes AI may eventually need to generate something like:
- $8T to $10T in revenue
That figure depends on margins, operating costs, taxes, and profits. In other words, AI cannot just be a niche tool; it would need to become a truly enormous global business.
His reality check: the 3P test
He uses a simple framework:
- Possible? Yes.
- Plausible? Maybe, but strained.
- Probable? He says low probability right now.
He is careful not to call it impossible or automatically call it a bubble because the data is still too incomplete.
Why Side Businesses Matter
A major part of the discussion is how much of the AI buildout is being subsidized by very profitable non-AI businesses:
- Alphabet’s advertising business
- Meta’s advertising business
- Microsoft’s cloud and enterprise businesses
- Other durable revenue streams that fund AI investment
Damodaran says those businesses should be separated out when evaluating the AI bet, because they can hide the real risk. If the AI factory turns out to be unprofitable, losses may be absorbed by those profitable businesses — but shareholders still bear the cost through lower returns and write-offs.
Who gets hurt if the bet fails?
- Alphabet/Meta/Microsoft shareholders: likely write-downs and opportunity cost
- Highly levered players like CoreWeave: bigger systemic risk if debt can’t be serviced
His concern is that debt-heavy AI players could spread pain beyond investors if the economics fail.
AI Economics: Tool vs. Replacement
Damodaran also challenges the idea that AI can simply charge huge sums as a “tool.”
If AI is just a productivity tool
- Revenue potential is limited, because firms won’t spend unlimited amounts on a tool layered onto existing labor costs.
If AI replaces workers
- He notes total global employee compensation is around $26T annually.
- That is the theoretical upper bound for AI’s addressable market.
But he says that scenario creates a macroeconomic problem:
- If AI replaces too many workers, those workers lose income.
- If they lose income, they stop buying goods and services.
- Then the companies using AI have fewer customers.
So the “replace everyone” scenario breaks the broader economy.
What This Means for Investors
Damodaran’s advice is not to panic, but to be intentional.
Avoid extreme market timing
He cautions against:
- dumping all AI-related holdings
- fleeing into utilities/consumer staples as a “safe” play
- moving everything to cash and waiting for a perfect entry
His warning: even if a correction is coming, waiting on the sidelines can cost more than the decline itself.
Time horizon matters
- If you’re close to retirement, reduce stock exposure and move to safer assets like bonds or T-bills.
- If you’re young with a long horizon, a future AI correction should not radically alter your investing plan.
Sleep test
He endorses a simple rule:
- If your portfolio keeps you awake at night, it’s too risky.
- Change it until you can sleep.
Damodaran’s Personal Investing Approach
He shares how he manages his own money:
Current approach
- Owns about 50 companies
- Portfolio built over 40+ years
- Globally diversified, across sectors
- Does not need index funds for diversification
What he sold
- Tesla: sold because the company has become too political for his taste
- Nvidia: sold in stages because it became too expensive, even though it was a huge winner
What he still owns
- Five of the Magnificent Seven, though he thinks most are now fairly valued or only slightly overvalued
His rules
- No single stock over 15% of his portfolio
- Sells in stages, not all at once
- Uses limit orders when emotions might interfere
- Will buy when prices are attractive, but only after doing enough homework to distinguish a great business from a falling knife
Practical Advice for Concentrated Holders
For listeners heavily concentrated in Tesla, SpaceX-linked exposure, or tech megacaps, his suggestion is:
- Don’t try to fix everything at once
- Sell in stages
- Consider doing it in tax-advantaged accounts first
- Use an autopilot plan if needed, so emotion doesn’t prevent action
He stresses that the goal is not to eliminate all regret — some sold stocks will keep rising — but to reduce dangerous concentration risk.
Key Takeaways
- AI’s current buildout is enormous: $2T+ in capex
- Current product/service revenues are still far too small to justify that spend
- Long-term AI revenues may need to reach $8T–$10T to make the math work
- The biggest risk is not just “AI is overhyped,” but that the economics may be much smaller than the infrastructure being built
- Investors should separate AI infrastructure winners from AI monetization bets
- Concentration risk matters more than being perfectly right about the theme
Bottom Line
Damodaran’s core message is that AI could still become massive, but the market is already pricing in a very ambitious future. He’s not calling it impossible — just hard to justify today. For investors, the bigger lesson is to understand what you own, avoid overconcentration, and make portfolio changes in a disciplined way rather than emotionally reacting to the AI narrative.