Over the last seven articles, we have examined what happens when institutions can no longer keep pace with the systems they govern.

They defer obligations.

Replace individual cases with categories.

Concentrate capacity.

Lose reliable measurement.

Slow their correction loops.

And eventually begin separating functions that once arrived as one institutional package.

That leaves one question.

What holds the system together after the package comes apart?

The intuitive answer is diversification.

The actual answer is more complicated.

Consider four services:

A payment provider.

An identity provider.

A cloud service.

A logistics platform.

They can be independently owned.

They can operate in different jurisdictions.

They can have different regulators.

They can compete with one another.

At the institutional layer, they are genuinely separate.

But follow them downward.

They may depend on the same fibre.

The same electricity grid.

The same compute region.

The same settlement infrastructure.

The functions separated.

Their underlying dependencies did not.

Instead, they converged on a smaller number of shared substrates.

This is the structural finding of our final article in Series I.

Institutional diversification can coexist with infrastructure concentration.

Subsea cables provide one of the clearest ways to see the problem.

The world's subsea cable systems experience roughly 200 faults each year.

That number has remained broadly stable since 2013 even as total route mileage increased from roughly 1.5 million to 2.7 million kilometres.

Measured per kilometre, the system has actually become more reliable.

Better surveys.

Deeper burial.

Better armouring.

Engineering worked.

But another variable moved in the opposite direction.

Repair response time more than doubled over the same decade.

The median global repair now takes around 40 days.

The longest recorded repair took 947 days.

This gives us two trends that should be considered together:

The system is not failing more frequently.

It is taking longer to recover.

And that distinction matters more as increasing numbers of services depend on each failure remaining short.

Repairing a cable is not simply a technical problem.

It requires specialised vessels.

Qualified crews capable of splicing fibre at sea.

Access to the damaged location.

And sometimes permission to operate in disputed or restricted waters.

Those resources exist in a commercially thin global market.

They cannot be scaled instantly.

So engineering can improve the cable while the system surrounding the cable becomes slower at restoring it.

This is the same pattern we identified at the beginning of Series I:

capacity to fail remains while capacity to recover is optimised away.

The system becomes less forgiving without becoming less reliable in the conventional sense.

Across Series I, five shared infrastructure joints kept appearing.

Settlement

Payments, savings, private money and transactions between institutions.

Identity

Access, entitlement and the ability to establish who someone is.

Compute and energy

Administrative capacity, machine assessment and increasingly the entire digital economy.

Materials

The physical inputs from which the other layers are built.

Connectivity

The layer across which the other four increasingly communicate.

These joints are not invisible because anyone is hiding them.

They are invisible for simpler reasons.

They are boring until they fail.

They are often privately owned.

And most importantly:

we count the system at the wrong level.

Regulators count licensed providers.

Procurement teams count suppliers.

Investors count companies and sectors.

Almost nobody asks:

How many of those supposedly independent services disappear together if one shared dependency fails?

We have good data about individual components.

Cable faults are counted.

Payment volumes are published.

Compute capacity can be estimated.

Refining concentration is reported.

But there is remarkably little systematic measurement of cross-layer dependency.

How many services fail together when one shared layer disappears?

That is the number required to understand actual systemic concentration.

Yet no jurisdiction we examined publishes it comprehensively.

Not because the question is impossible.

Because no institution has a mandate spanning all of the separately owned and separately regulated layers required to answer it.

That creates an important measurement gap.

And as we argued earlier in this series:

the absence of measurement can easily be mistaken for the absence of a problem.

Our base scenario is Silent Concentration — 45%.

Institutional separation continues while shared infrastructure concentration deepens largely unmeasured.

Outages are attributed to the services that stop working rather than to the common layer underneath them.

We assign 30% to Substrate Securitisation.

Governments increasingly designate privately owned infrastructure as critical and impose security, reporting and continuity requirements.

The European response to Baltic cable damage is an early example.

We assign 15% to Redundancy by Cost.

Commercial actors build alternative paths because outages become sufficiently expensive to justify capacity that earns little during normal periods.

And 10% to Cascading Substrate Failure.

A single infrastructure joint fails and removes several nominally independent services simultaneously in a region with limited alternatives.

There is little value in pretending an individual can solve global infrastructure concentration.

The practical task is narrower.

Know your exposure.

Ask:

If connectivity disappears for one week, which functions in my life disappear with it?

Payments?

Work?

Identity verification?

Critical records?

Communication?

Then maintain one offline capability for the functions you genuinely cannot afford to lose.

This is not preparation for catastrophe.

It is continuity planning against a documented recovery variable.

Take your vendor list.

Then collapse it into a substrate list.

Which cloud regions?

Which settlement rails?

Which identity systems?

Which physical connectivity routes?

Most companies will discover that the vendor list is long.

The substrate list is short.

The second is the real dependency map.

Then ask the multi-service question:

If this substrate disappears, how many functions stop simultaneously?

And where you purchase redundancy, verify it physically.

Two suppliers.

Two contracts.

Two invoices.

One underlying route.

That is not redundancy.

It is duplicated paperwork around the same failure point.

Separate two variables that are often treated as one.

Failure probability.

And:

Recovery duration.

The first is broadly stable in our cable example.

The second has deteriorated substantially.

For a leveraged company or infrastructure-dependent asset, that difference can determine whether an outage is a temporary inconvenience or a liquidity, covenant or solvency event.

Recovery time deserves to be treated as a risk variable in its own right.

Eight articles have examined very different systems.

But one correction kept appearing.

Alternatives that are genuinely separate where they are counted can become identical where they fail.

Three suppliers behind one port.

Two banks on one rail.

Twenty-seven jurisdictions on one specification.

Multiple institutions dependent on one physical cable.

That gives us the instrument we will carry forward:

How many genuinely independent alternatives exist once common origins are traced?

What stops if this fails?

How long before an alternative actually functions?

This is where Series I ends.

And where Series II begins.

Series I examined institutions.

Series II goes underneath them.

Energy and the grid.

Compute and chips.

Water.

Refined materials.

Connectivity.

Logistics corridors.

Each article will apply the same three questions to one physical substrate.

By the end, the objective is not another collection of essays.

It is a comparative matrix of the infrastructure seams underneath the modern system.

And the closing observation from Series I carries directly into that work:

The world is not becoming less forgiving because things suddenly break more often.

It is becoming less forgiving because recovery takes longer while the number of things depending on rapid recovery keeps increasing.

Most of the rising cost of the next decision is time.

Read the full analysis:

AFTER THE PACKAGE
What holds a system together once the functions have separated

THRIVE IN CHAOS
Decision Intelligence for an Uncertain World

Analysis → Forecast → Recommendations
Signal → Meaning → Action → Stability

Forecasts are probability-based analytical assessments, not certainties. This material supports independent judgment and does not constitute financial, legal or investment advice.