Google & Alibaba: Why Does the AI Investment Cycle Keep Going?
Because stopping AI CapEx can be more dangerous for a Big Tech company than continuing it, even when the near-term ROI is uncertain.
This is the core of the AI capital cycle.
1. The strategic logic: “If I stop, my competitor wins”
Imagine Alibaba spends $10B on AI while Tencent, ByteDance, Microsoft, Google, Amazon, etc. continue spending.
If Alibaba stops:
Alibaba stops AI CapEx
↓
Less compute
↓
Slower model/product improvement
↓
Developers/customers move elsewhere
↓
AI ecosystem weakens
↓
Future market share lostThe cost of under-investing is difficult to measure today.
So management tends to think:
“The ROI may be uncertain, but the cost of falling behind is potentially enormous.”
That creates a powerful FOMO mechanism.
---
2. AI is unusually capital-intensive
Traditional software:
Hire engineers
↓
Build software
↓
Marginal cost ≈ low
↓
Scale globallyFrontier AI:
More users
↓
More inference
↓
More GPUs
↓
More electricity
↓
More data centers
↓
More networking
↓
More GPUsSo success itself creates additional CapEx requirements.
That's important.
If AI demand is genuinely strong, companies need to keep investing simply to serve that demand.
---
3. The incumbent's dilemma
For Alibaba, the decision isn't simply:
“Should we invest $80B in AI?”
It's closer to:
“Should we invest $80B, while everyone else is investing $100B+?”
That changes the psychology.
Even if management believes the expected AI return is only moderately attractive, the strategic option value of maintaining AI leadership can justify continued spending.
---
4. The really interesting part: CapEx becomes self-reinforcing
This creates a loop:
Competitor CapEx ↑
↓
Management fears falling behind
↓
Own CapEx ↑
↓
Supplier revenue ↑
↓
AI infrastructure narrative strengthens
↓
Investors expect more AI investment
↓
Competitor feels pressure
↓
CapEx ↑So you don't necessarily need executives to be irrational.
Everyone can be acting rationally.
Yet the collective outcome can still be excessive investment.
This is very similar to an arms race.
---
5. Why management has a particularly strong incentive
There is another layer.
A CEO can easily justify:
“We invested heavily in AI and AI didn't work.”
The explanation can be:
“The technology evolved differently than expected.”
But imagine the opposite:
“We didn't invest, while our competitors did, and AI became the dominant technology platform.”
That is potentially a career-ending decision.
So management faces an asymmetric risk:
| Decision | If AI succeeds | If AI fails |
|---|---|---|
| Invest aggressively | 🟢 Hero | 🔴 Wasteful CapEx |
| Don't invest | 🔴 Strategic failure | 🟢 Capital discipline |
The career risk of under-investing can therefore be greater than the career risk of over-investing.
---
6. This is where executive compensation matters
Suppose the CEO owns significant stock.
Then:
AI spending
↓
Revenue growth narrative
↓
Higher valuation multiple
↓
Stock price ↑
↓
CEO wealth ↑The CEO receives the upside relatively quickly.
But the downside from a failed AI investment may arrive years later through:
- depreciation
- lower ROIC
- excess capacity
- restructuring
- debt
- dilution
This creates a potential time-horizon mismatch.
The executive can capture today's narrative value while shareholders absorb tomorrow's capital-allocation consequences.
That doesn't mean the executive is behaving improperly.
It means the incentive structure matters.
---
7. Why Alibaba raising equity is particularly interesting
This is why the Alibaba situation is worth watching.
If a company has enormous AI ambitions but doesn't want to slow CapEx, it has several choices:
Option A — Use existing cash
Cash → AI CapExProblem: reduces liquidity.
Option B — Borrow
Debt → AI CapExProblem: increases leverage and interest burden.
Option C — Issue equity
New shareholders
↓
Cash
↓
AI CapExProblem:
Existing shareholders are diluted.
Alibaba chose the third mechanism for this HK$80B raise.
That tells us something about the scale of the investment cycle.
---
8. The dangerous version
The real danger is not:
“Companies are spending heavily on AI.”
The dangerous version is:
AI CapEx ↑
↓
AI revenue ↑
↓
But AI FCF does not keep up
↓
Need external financing
↓
More CapEx
↓
More financing
↓
Higher valuation
↓
More financing capacity
↓
More CapExAt that point:
The financing system begins supporting the AI investment cycle rather than AI cash flows supporting the financing system.
That's the line I would watch.
---
9. The strongest counterargument
There is an important reason not to assume this is a bubble.
AI could have a very unusual payoff curve.
For example:
2025–27:
CapEx >> AI FCF
2028–30:
AI adoption accelerates
2030+:
AI productivity + monetization explodeIf that happens, today's enormous CapEx could be completely rational.
Amazon's early AWS investment looked excessive at certain points.
The same could eventually be true for AI.
---
10. Therefore, the key metric isn't CapEx
This is the most important point.
Don't ask:
“Why are they spending so much?”
Ask:
“What happens to incremental return on invested capital as spending increases?”
Watch:
AI CapEx ↑
↓
AI Revenue ↑?
↓
AI Gross Profit ↑?
↓
AI FCF ↑?
↓
ROIC ↑ or ↓?Healthy cycle
CapEx ↑
Revenue ↑↑
FCF ↑↑↑
ROIC ↑Bubble cycle
CapEx ↑↑↑
Revenue ↑
FCF →
ROIC ↓↓
Debt ↑
Equity issuance ↑The second pattern is what would make me seriously bearish.
---
Bottom line
They maintain the investment cycle because stopping is strategically dangerous.
But that creates a paradox:
The fear of missing the AI revolution can cause companies to invest so aggressively that the industry collectively overbuilds the AI infrastructure needed for the revolution.
And that's why the Alibaba event is interesting.
The important signal isn't simply:
“Alibaba fell 10%.”
It's:
“The market may be starting to distinguish between AI investment and AI returns.”
If that distinction spreads to NVIDIA, Microsoft, Amazon, Google, Meta and the private AI ecosystem, that would be a much more significant signal that the AI capital cycle is reaching its stress point.