The Physical Internet: What Exists Behind the Cloud

The internet feels weightless.

You open an app.

A page loads.

A video starts playing.

An AI answers your question.

A file appears on your screen.

From your perspective, almost nothing happened.

But underneath that moment is an enormous physical machine.

Buildings.

Power stations.

Servers.

Fiber.

Submarine cables.

Routers.

Cooling systems.

Data centers.

Land.

Satellites.

Ports.

Factories.

And thousands of people maintaining infrastructure that most users will never see.

We call it the cloud because the physical reality is inconveniently complicated.

The cloud is not in the sky.

It is somewhere.

And increasingly, where it is becoming one of the most important questions in technology.

Start with the cables

Take something as ordinary as sending a message across the world.

You type.

You press send.

The message appears almost instantly.

But if the person receiving it is on another continent, your data may travel through an extraordinary physical network before reaching them.

Fiber-optic cables carry enormous amounts of internet traffic across oceans.

According to TeleGeography's 2025 submarine cable map, there were 597 active or under-construction submarine cable systems and 1,712 cable landings represented on the map.

These aren't abstract connections.

They are physical cables sitting on the seabed.

They connect countries.

They connect continents.

And they connect the data centers where much of the world's digital activity actually happens.

This means something important:

The global internet is partly an infrastructure network built on the ocean floor.

Then there are the buildings

Your request eventually needs to reach a computer capable of processing it.

That computer may sit inside a data center.

A data center is essentially a highly engineered building designed to house enormous quantities of computing equipment.

Inside are racks of servers.

Storage systems.

Networking equipment.

Power systems.

Cooling systems.

Backup generators.

Uninterruptible power supplies.

And connections to external networks.

The International Energy Agency describes data centers as facilities housing servers, storage, networking equipment and the supporting infrastructure required to keep them operating.

The important word here is supporting.

The computers aren't the whole system.

A server without electricity isn't useful.

A server that overheats isn't useful.

A server without network connectivity isn't useful.

A data center without sufficient grid capacity may simply be an expensive building full of equipment that cannot operate at its intended scale.

That is why the physical infrastructure around computing matters almost as much as computing itself.

The hidden dependency: electricity

This is where the story becomes much bigger.

Technology companies often talk about computing as if the primary scarce resource is intelligence.

More GPUs.

Better models.

More parameters.

More inference.

More users.

But every computation ultimately requires energy.

And AI is increasing the importance of that relationship.

The IEA estimates that global data-center electricity consumption reached about 415 TWh in 2024, around 1.5% of global electricity consumption at the time. Its base case projected consumption to roughly double to about 945 TWh by 2030.

More recent IEA analysis says data-center electricity consumption increased 17% in 2025, while electricity consumption from AI-focused data centers grew even faster.

The interesting part isn't simply the size of the number.

It is the geography.

Electricity cannot be moved around the world as effortlessly as information.

A developer can deploy software globally in minutes.

You cannot deploy a 500-megawatt power connection globally in minutes.

You need land.

Transmission.

Transformers.

Generation.

Permits.

Construction.

Cooling.

Equipment.

Time.

This creates a strange asymmetry at the heart of modern technology:

Information moves incredibly quickly. Physical infrastructure does not.

The cloud has a location

Consider what happens when you use an AI service.

You don't see its location.

You see an interface.

But somewhere, physical machines are processing your request.

And the location of those machines matters.

Power prices matter.

Grid reliability matters.

Network latency matters.

Climate matters.

Land availability matters.

Tax policy matters.

Regulation matters.

Access to fiber matters.

Access to chips matters.

Political stability matters.

Water availability can matter for cooling.

And increasingly, access to electricity itself can become a constraint.

The IEA notes that data centers are geographically concentrated, meaning their effects on electricity systems can be much larger locally than their global percentage might suggest.

This changes how we should think about technology companies.

A technology company isn't necessarily just a software company anymore.

It may depend on an enormous physical supply chain.

The invisible stack

Think about the modern internet as layers.

At the top:

You

The person using an application.

Below you:

Applications

Chatbots.

Social networks.

Search engines.

Streaming platforms.

Banks.

Marketplaces.

Then:

Software infrastructure

Cloud platforms.

Databases.

APIs.

Authentication.

Developer tools.

Then:

Computing

CPUs.

GPUs.

Memory.

Storage.

Servers.

Then:

Data centers

Buildings.

Cooling.

Power distribution.

Backup systems.

Network equipment.

Then:

Networks

Fiber.

Internet exchanges.

Routers.

Content-delivery networks.

Submarine cables.

Then:

Energy

Generation.

Transmission.

Distribution.

Fuel.

Storage.

And beneath some of that:

Physical industry

Semiconductors.

Steel.

Copper.

Aluminium.

Construction.

Transformers.

Industrial machinery.

Logistics.

Real estate.

The further down you go, the less visible the system becomes.

Yet the lower layers determine what is possible at the top.

This is why AI is becoming an infrastructure story

The first phase of the AI boom was largely about models.

Who has the best model?

Who has the best benchmark?

Who has the best reasoning?

Who has the best product?

Those questions still matter.

But another competition is developing underneath them.

Who has enough compute?

Who can secure chips?

Who can build data centers?

Who can obtain electricity?

Who controls the network?

Who can finance the infrastructure?

Who can build it quickly enough?

The IEA reported in 2026 that capital expenditure by five large technology companies exceeded $400 billion in 2025 and was expected to increase further in 2026, while also highlighting bottlenecks involving transformers, gas turbines, chips, grid connections and approvals.

That tells us something important.

The AI race isn't occurring entirely inside laboratories.

It is also occurring in:

power plants, factories, warehouses, construction sites and transmission networks.

The geography of intelligence

This may become one of the defining questions of the next decade.

Where does computing happen?

Historically, the internet made geography feel less important.

A company could be based in one country and serve customers everywhere.

But AI may bring geography back into technology in a different way.

Because large-scale computing requires physical infrastructure.

And physical infrastructure requires resources.

That means countries have incentives to develop their own computing capacity.

Not necessarily because they want to disconnect from the global internet.

But because computing capacity increasingly has strategic value.

A country that controls energy, data centers, networks, chips and technical talent has a different position from one that simply consumes services built somewhere else.

This is why the infrastructure layer deserves much more attention.

The paradox

The internet has spent decades making the world feel increasingly digital.

But the next phase of the internet may make the physical world matter more than ever.

AI requires chips.

Chips require factories.

Factories require energy.

Data centers require electricity.

Electricity requires generation and grids.

Networks require fiber and cables.

And all of it requires land, capital and physical construction.

The more digital the economy becomes, the more important certain physical systems can become.

That is the paradox.

The digital economy is becoming increasingly dependent on physical infrastructure.

And Africa is part of this story

This isn't only a story about Silicon Valley.

Africa's position in the physical internet matters.

The continent is connected to an expanding network of submarine cable systems, while new data-center, cloud, fiber and connectivity investments are changing where digital infrastructure can be located.

But the opportunity isn't simply:

"Africa needs more internet."

The larger question is:

Where will Africa's computing infrastructure be built?

Where will the power come from?

Where will the data centers sit?

Which cities become connectivity hubs?

Which countries develop meaningful local compute?

Which companies build the infrastructure?

Which industries become dependent on it?

And perhaps most importantly:

Who owns the physical layer?

That question will matter far beyond technology.

The cloud is becoming a map

We tend to imagine technology as a collection of products.

Apps.

Models.

Platforms.

Websites.

But underneath them is a map.

A map of:

power plants

data centers

fiber routes

submarine cables

internet exchanges

chip factories

industrial supply chains

ports

land

capital

And the companies and countries positioned around those physical systems may have enormous influence over what can be built above them.

The next time an AI model answers your question in two seconds, remember:

There is no cloud.

There is a building.

Inside that building are machines.

Those machines require electricity.

That electricity comes through infrastructure.

The data arrives through networks.

Those networks eventually connect to physical cables and facilities around the planet.

And behind all of it is an industrial system that is becoming one of the most important parts of the digital economy.

The internet may look weightless.

Its foundations are not.

Omniv editorial metadata

Category: Technology / Infrastructure / AI Format: Analysis Reading time: ~8–10 minutes Tags: AI, infrastructure, cloud computing, data centers, energy, submarine cables, internet, Africa, technology

Related Omniv articles to publish next:

Why Energy Could Become the Bottleneck for Technology

The New Geography of Computing

What Happens When Countries Start Treating Data Like Oil?

The Infrastructure Wars Nobody Is Talking About

AI Is Not Just a Software Revolution

That sequence is deliberate: physical internet → energy → geography → national strategy → AI infrastructure. It gives Omniv a coherent editorial universe rather than a collection of disconnected tech articles.

Absolutely. I'll continue the entire AI / computing / infrastructure universe in the same Omniv editorial style. These should read as connected investigations rather than generic SEO posts.