Direction over level
Rate of change often matters before an absolute number looks alarming.
PURPOSE & PHILOSOPHY
AI can succeed.
The investment cycle can still break.
This dashboard does not predict whether artificial intelligence will work. It tests whether demand, capital spending, cash generation and financing still reinforce one another — and warns when that economic chain begins to separate.
THE USEFUL ANALOGY
A feedback mechanism,
not a recycled crisis.
The useful part of the 2008 comparison is narrow: systems that depend on continuous acceleration can weaken before activity actually falls. In this cycle, slowing growth in AI spending, revenue or financing can matter even while all three remain historically high.
But the financial structure is different. The major buyers are profitable corporations, not highly leveraged households funded through a fragile mortgage system. A closer historical analogy may be the late-1990s telecom and internet buildout: transformative technology, necessary infrastructure, and still the possibility that investors overpay for the returns.
Technological success does not guarantee attractive investment returns.
THE REINFORCING CYCLE
AI demand→Hyperscaler capex→Chips, data centres & power→Earnings & growth→Easier financing→More investment
The dashboard looks for breaks in this chain. Its central relationship is AI-linked capex growth versus AI/cloud revenue and free-cash-flow growth. The other signals test demand, financing, competition, macro dependence and market confirmation around that core.
OPERATING PRINCIPLES
Evidence before narrative.
Confirmation over headlines
One weak reading is noise. Several independent signals weakening together is a regime change.
Economics over excitement
The monitor separates adoption from the ability of investors and operators to earn acceptable returns.
Missing is not neutral
Unverified data is withheld, never guessed or silently scored as safe.
Signals, not commands
A red regime calls for investigation and portfolio review. It is not an automatic instruction to sell.
History over snapshots
The system becomes more valuable as it builds a consistent record through multiple reporting cycles.
THE EIGHT MARKERS
One thesis,
tested from eight angles.
No marker is definitive on its own. Each has a specific job, an observable warning condition and a known limitation.
01 · CORE ECONOMICS
Capex & cash returns
Tests aggressive hyperscaler infrastructure spending against the cash those companies retain after investment.
- Why it matters
- A profitable, self-funded boom is structurally different from one whose commitments keep rising while cash returns contract.
- Healthy evidence
- Free cash flow expands while capex remains supported by revenue.
- Warning evidence
- Capex stays elevated while aggregate free cash flow contracts.
- Known limitation
- Reported capex is not exclusively AI-related and annual filings are backward-looking.
02 · FINANCING
Credit stress proxy
Tracks the market price of corporate financing and changes in risk appetite.
- Why it matters
- Credit can reveal funding stress before it becomes obvious in equity prices or company guidance.
- Healthy evidence
- Spreads are stable or tightening and issuance clears without unusual concessions.
- Warning evidence
- Spreads widen rapidly, issuance jumps, or new debt meets weak demand.
- Known limitation
- The current broad investment-grade proxy is not exclusive to AI issuers.
03 · DEMAND
Semiconductor demand proxy
Checks whether real demand for the compute layer remains strong.
- Why it matters
- Chips sit near the front of the physical investment chain. A material slowdown can travel into data centres, power and capex plans.
- Healthy evidence
- Revenue, orders and backlog remain broad and resilient.
- Warning evidence
- Growth, orders, backlog or lead times deteriorate together.
- Known limitation
- A single supplier's data-centre revenue cannot represent the entire semiconductor cycle.
04 · COMPETITION
Open-model pressure
Measures whether cheaper open or Chinese models are compressing the economic value of frontier intelligence.
- Why it matters
- AI can succeed while model access becomes commoditised, expanding usage but weakening pricing power.
- Healthy evidence
- Performance and price advantages support sustainable differentiation.
- Warning evidence
- Near-frontier capability becomes available at a steep and persistent discount.
- Known limitation
- Benchmark choice, hosted prices and model classification affect the comparison.
05 · CORE ECONOMICS
AI monetisation proxy
Compares visible cloud revenue growth with the spending required to produce it.
- Why it matters
- This tests whether economic returns are catching up with investment.
- Healthy evidence
- Monetisation grows at least as fast as the relevant investment base.
- Warning evidence
- Capex repeatedly outpaces observable revenue by a widening margin.
- Known limitation
- Cloud revenue is not purely AI revenue.
06 · PORTFOLIO
Market concentration
Measures how dependent the S&P 500 is on its ten largest companies.
- Why it matters
- Narrow leadership increases portfolio damage if AI economics disappoint.
- Healthy evidence
- Market value and earnings leadership broaden.
- Warning evidence
- A rising share of market value depends on fewer companies.
- Known limitation
- This is downstream portfolio fragility, not proof that AI demand is weakening.
07 · EXPECTATIONS
Valuation expectations
Measures how much future earnings success is already embedded in broad US equity valuations.
- Why it matters
- Demanding expectations amplify losses when delivery disappoints.
- Healthy evidence
- Valuation is supported by durable earnings growth and revisions.
- Warning evidence
- Valuation stays demanding while earnings expectations weaken.
- Known limitation
- Verified earnings-revision breadth is not yet connected and contributes nothing.
08 · MACRO EXPOSURE
Data-centre investment / GDP
Estimates how much wider investment momentum is tied to the data-centre buildout.
- Why it matters
- The larger the contribution, the larger the growth air pocket if projects are delayed.
- Healthy evidence
- Investment grows at a pace supported by utilisation and end demand.
- Warning evidence
- Exposure rises as project financing or end demand weakens.
- Known limitation
- Construction excludes chips, servers and power equipment, so it is a lower-bound proxy.
HOW TO READ THE OUTPUT
A regime is a conclusion
with conditions.
Mutually supportive
Demand, funding and earnings remain aligned. Monitor, but do not manufacture a crisis narrative.
Pressure is spreading
Multiple independent indicators are weakening. Examine exposures and the durability of the thesis.
Cross-signal confirmation
Five signals deteriorate together, the composite reaches 4.0, or an exceptional funding event occurs. Review risk; do not treat the label as a trade instruction.
THE BOUNDARY
What this is
A source-traceable early-warning system for stress in the AI investment cycle.
A disciplined way to challenge both AI euphoria and reflexive crash predictions.
What this is not
A forecast of an exact crash date, a claim that AI is a bubble, or proof that 2008 is repeating.
A substitute for security-level valuation work, portfolio construction or personal financial advice.
FROM THESIS TO EVIDENCE