Why the AI Boom May Be Unsustainable

FinanceELITE Macro Investor: How to invest as the AI BUBBLE POPS

TL;DR

Current market expectations for 25% annual earnings growth over 5 years require unprecedented productivity gains and massive household asset depletion to sustain AI capex spending, creating systemic fragility despite strong near-term earnings.

Key Takeaways

1Wall Street analysts are pricing in 25% annual earnings growth for 5 years—the highest 5-year growth in the post-WWII era—requiring 1–1.5 points of margin expansion yearly that can only come from cutting labor income or massive household savings depletion.
2AI capex of $600–700 billion annually represents only 2% of GDP; most high-value chip components are imported from Korea and Taiwan, so US GDP growth is actually driven by household desaving (3.5% income growth vs. 6–7% spending growth), not AI investment.
3Current AI revenues stand at ~$130–150 billion run-rate; reaching $5 trillion in AI-related spend over 5 years would require AI to represent one-fifth to one-sixth of the entire economy, with no current productivity data supporting such extraordinary growth.
4Frontier AI models (OpenAI, Anthropic, hyperscalers) are collectively losing money despite record capex; 70% of Microsoft's cloud revenue is from OpenAI, creating interconnected financial fragility similar to pre-2008 crisis dynamics.
5Leverage is now the marginal price setter in markets; levered ETFs, options, and securities borrowing are amplifying bull positioning, making the market vulnerable to delevering cascades (similar to recent Korean market volatility and semi-sector drawdowns).
6Attractive alternatives include 3% real-yielding TIPS (Treasury Inflation-Protected Securities) and gold, which historically delivered equity-like returns over 25 years while remaining diversifying hedges against market downturns.
7European earnings are surprising sharply to the upside and Japan's corporate restructuring is generating strong cash flows, yet these opportunities remain overlooked while concentrated AI/semi hype dominates investor attention.

Notable Quotes

“When you start to get to a point where the extraordinary is the consensus view, that's when you want to start to be cautious and find other opportunities.”
— Bob

“The thing that's so interesting about it is if you've got to deliver the most extraordinary productivity expansion in the history of man over the course of the next 5 years, you'd think it wouldn't be printing 1.1% annualized in the first half of this year. Where is my AI-driven productivity growth because it's not showing up in any of the statistics.”
— Bob

“What are the odds that we have the greatest productivity boom in the history of man and on top of it have US households erase 25% of their wealth in order to support company profits? Low.”
— Bob

“Margins are just the flip side of labor in many cases. If you're going to get margin expansion, you have to pay the workers less by an equivalent amount.”
— Bob

Chapters

1. The AI Expectations Mania

Market consensus prices in 25% annual earnings growth for the next 5 years—the highest in post-WWII history—requiring extraordinary productivity gains never achieved before and impossible margin expansion without cutting worker income.

2. The Math Behind Unsustainable Growth

To achieve 25% earnings growth with 10% nominal sales growth requires 1–1.5 points of margin expansion yearly; this forces either 5–7% labor income cuts or household savings rates to plummet from 3% to -15%, neither of which is sustainable.

3. AI Capex Is Only 2% of GDP

Hyperscaler AI capex of $600–700 billion annually contributes only tenths of a basis point to US GDP growth; most high-value components are imported from Korea and Taiwan, so real growth is actually driven by household desaving, not AI.

4. The Real Economy Problem

Current AI revenues (~$130–150 billion) must grow to $5 trillion in 5 years to justify capex; this requires end-users like Kelloggs to spend AI-derived profits, not just tech firms trading with each other in circular revenue loops.

5. Interconnected Financial Fragility

Frontier models are losing money despite billions in capex; 70% of Microsoft's cloud revenue comes from OpenAI, creating 2008-style interconnection risk where one missed revenue target cascades through hyperscalers, chipmakers, and infrastructure firms.

6. Productivity Data Contradicts the Narrative

Despite rhetoric about AI-driven productivity breakthroughs, actual productivity growth in H1 2024 was only 1.1% annualized—mediocre by historical standards—contradicting claims of the greatest productivity boom in human history.

7. Leverage as the Marginal Price Setter

Levered ETFs, options, and wealth-manager securities borrowing are driving incremental price bids; when leverage unwinds (as seen in Korean markets and recent semi-sector drawdowns), a 20% drop occurs on minimal fundamental news.

8. Overlooked Value Opportunities

European earnings are beating expectations substantially and Japan's corporate restructuring is generating strong cash flows, yet these receive minimal attention; biotech is also rebounding outside AI hype.

9. Defensive Portfolio Construction

3% real-yielding TIPS and gold each delivered equity-like 25-year returns with meaningful diversification benefits; gold trades near 2024 lows after early-year mania, making both attractive hedges in a leverage-heavy market.

10. The Path Forward

Rather than concentrated AI/semi exposure, alpha lies in diversifying away from hype and positioning for either a prolonged choppy sideways market (earnings slowly catch up to prices) or a delevering shock when households stop desaving.

Key People & Entities

Bob: Macro investor and newsletter writer; founder of alpha indexing ETF strategies; argues current AI investment is unsustainable without extraordinary real-economy productivity.
OpenAI: Frontier AI model company generating significant capex at hyperscalers; currently loss-making despite billions in compute spending from Microsoft.
Microsoft: Cloud computing provider; 70% of cloud revenue derives from OpenAI workloads, creating revenue dependence on AI model monetization.
Nvidia: GPU chipmaker; major beneficiary of hyperscaler AI capex spending; subject to leveraged retail investor positions.
Taiwan Semiconductor Manufacturing Company (TSMC): Chip manufacturer reporting 40% revenue growth; supplies high-value components for AI infrastructure, though production is off-shored from the U.S.
Anthropic: Frontier AI model company; historically loss-making or marginally cash-flow-neutral despite significant capex investment.

Glossary

TIPS (Treasury Inflation-Protected Securities): U.S. government bonds that pay a fixed real yield plus inflation adjustments; currently offering 3% real yields over their maturity, making them an alternative to equities at current valuations.
Desaving: When households spend more than their current income by drawing down savings or increasing debt; currently U.S. household savings rates are only 3% while spending grows at 6–7%, indicating aggressive desaving.
Margin Expansion: An increase in corporate profit margins (earnings as a percentage of revenue); typically achieved through either raising prices, cutting costs (e.g., labor), or increasing efficiency.
Capex (Capital Expenditure): Spending by companies on physical infrastructure, equipment, and buildings; AI capex refers to spending on data centers, GPUs, and computing infrastructure for AI model training.
Leverage: Using borrowed money to amplify investment returns; includes levered ETFs, options, margin accounts, and securities borrowing programs.
Alpha Indexing: A strategy that replicates hedge fund positioning and risk management in a low-cost ETF wrapper, combining hedge fund wisdom with liquid ETF accessibility.
Equity Risk Premium: The expected additional return from stocks over risk-free bonds; historically around 5–6% but compressed when bond yields (like 3% TIPS) are high relative to equity earnings yields.
Klei Equation: An accounting identity showing that one sector's profit or savings is another sector's deficit; illustrates why if corporations have high margins, households must be desaving.

Explore