Four of five dimensions have crossed their historical thresholds.
Each dimension is measured against the level that preceded past infrastructure buildouts that ended in writedowns, from 1840s rail to the 2000s telecom cycle. The fifth, utilization, sits on the line.
What this reading changes.
Every reading opens with the same seven fields, in the same order, so two readings can be compared line for line. The charts sit below this layer, not above it. A position on a chart is a measurement, not a verdict.
Credit issuance crossed its historical threshold, taking the count to four of five. Utilization did not move off the line.
Utilization is the dimension that converts installed capacity into cash flow. Every prior cycle in the reference class cleared or failed on that conversion, not on the size of the buildout.
Whether the current buildout clears at installed cost, or the assets transfer to second owners at a discount.
High on the crossings themselves, which are measured against cited figures. Medium on the reading as a whole: two of the five thresholds are analyst judgment rather than a cited line.
- NarrowedFinancing structures that assume utilization holds above the writedown line.
- OpenedDiligence on second-owner economics: what the installed base is worth to an operator who did not pay to build it.
- UnchangedAnything resting on end demand for compute. This reading measures the financing and utilization of supply.
- ClosedNone this reading.
Utilization printing clearly above the 50% line at the next sweep. A revision to either judgment threshold. Capex guidance reversing rather than holding.
Next monthly sweep. Utilization is the dimension under watch.
Each field is backed by entries in the same claim records the evidence layer produces, each carrying its sources, an independence grade and a stated falsifier. Thresholds are pre-registered in the Register. Construction and grading rules: the method. Machine-readable: /data/readings.json.
Three objects, kept separate.
Most monitoring collapses these into one number. Keeping them apart is what allows a reader to disagree with the third without disputing the first.
Observed signal
What changed in the measured world. Distance from a threshold lives here and nowhere else.
Analytical implication
How that change bears on the pivotal uncertainty, argued in writing and open to dispute.
Decision impact
Which options opened, narrowed, closed or stayed put, for a named decision holder.
It is never presented as decision latitude. A marker twice as far past the line is not twice as consequential, and the scope above says nothing on its own about which choices remain open. A named analyst makes that connection in the decision layer, in prose a reader can challenge.
Five dimensions, measured against their thresholds.
The scope above renders these five tracks in polar form. Here they are laid flat, each on its own baseline-to-ceiling scale, with the pre-registered threshold marked. Nothing is normalised into a shared index, because the dimensions do not share units and pretending otherwise would hide the judgment in the weighting.
Utilization is under revision. The September evidence update explains why no new crossing has been recorded for it, and Utilization Is Not One Number sets out the six-stage construction that replaces the single threshold prospectively. Sources: company capex disclosures, NVIDIA Q3 FY26, Sequoia Capital, and research reports from Morgan Stanley and J.P. Morgan. Two of the five thresholds — cost and utilization — are analyst judgment anchored to a benchmark rather than to a base rate, and carry at most medium conviction. Both are labelled as such in NCB-004.
What each signal does to the others.
Signals are not read one at a time. Each relationship asserted below carries its own confidence grade, separate from confidence in the underlying observations, because a well-measured fact can support a weakly argued relationship.
| Relationship | Mechanism | Confidence |
|---|---|---|
| Credit › Utilization | Debt-financed capacity carries a fixed coupon. Where utilization does not convert capacity into cash flow, the coupon is met from balance sheet rather than operations. | High |
| Competition › Cost | Concentration of purchasing in four buyers holds supplier pricing up, which sustains capital intensity even as unit compute costs fall. | Medium |
| Profitability › Credit | A widening gap between capex and AI revenue lengthens the payback period, which pushes financing from equity toward debt. | High |
| Utilization › Ownership | Sustained shortfall against fixed obligations is the mechanism by which assets transfer without ceasing to operate. This is the relationship the historical reference class is selected on. | Medium |
Three cycles that rhyme.
Infrastructure overbuild followed by ownership transfer is one of the older shapes in capital. The asset class is new each time. The mechanics are not. The class is named here so a reader can dispute the choice rather than absorb it.
Habsburg silver and sovereign debt
American silver financed the largest empire in Europe, collateralized through Genoese and German banking houses. Four sovereign defaults in 51 years. The mines kept producing. Operational control transferred to Dutch and Genoese financiers.
Capex without yield discipline
The Railway Mania
Parliament authorized over 7,000 miles of new railway between 1844 and 1846. About 6,000 miles got built. Most operating companies were insolvent by 1850. The track stayed, and became the backbone of the late-Victorian economy under consolidated successor firms.
Utilization shortfall meets fixed debt
Fiber and telecom buildout
Roughly $500B of investment laid the long-haul fiber spine of the modern internet. WorldCom, Global Crossing and 360networks all failed. By 2002 an estimated 95% of installed fiber was dark. It went on to carry the cloud era under new owners at cents on the dollar.
Credit cycle inversion
Infrastructure stays. Owners change.Darśan · historical orientation
Disanalogies carried with the class: none of the three prior cycles had a single supplier capturing the majority of capital expenditure, and none had an asset with a depreciation schedule as short as current accelerator hardware. Both weaken the analogy on the recovery side and are recorded with it.