AI revolution needs algorithms: Infrastructure behind stargate

Artificial intelligence often looks like software. You type a question into ChatGPT, receive an answer and move on. Behind that simple interaction, however, sits an enormous physical infrastructure system involving computer chips, data centres, electricity, cooling systems, fibre networks, construction and billions of dollars in investment.

Project Stargate is one of the clearest examples of this shift.

Announced in January 2025, Stargate was presented as a joint venture involving OpenAI, Oracle and SoftBank, with MGX as a financial backer and Microsoft, NVIDIA and Arm involved as technology partners. The announced ambition was up to $500 billion in U.S. AI infrastructure by 2029, targeting roughly 10 gigawatts of computing capacity.

But the most interesting part of Stargate may not actually be the headline number.

It is the physical machine being built underneath the AI economy.

AI needs a physical foundation

The modern AI boom depends heavily on specialised processors known as GPUs. Originally developed for graphics and gaming, GPUs became extremely valuable for AI because they can perform thousands of calculations in parallel.

Stargate's flagship Abilene, Texas campus illustrates the scale involved. The project is designed around enormous computing capacity, with one OpenAI–Oracle arrangement envisioning infrastructure capable of eventually housing roughly two million AI chips. Every one of those chips requires power, cooling and networking.

That changes how we should think about AI.

The “cloud” is not really a cloud.

It is buildings filled with machines.

The surprising bottleneck may be electricity

It is easy to assume that the biggest limitation facing AI is access to better chips. Project Stargate suggests another problem may be even more important: power.

The book describes electricity as the potential real currency of the AI infrastructure race. A gigawatt-scale AI data centre can consume electricity on a scale comparable to a large power plant or roughly 100,000 average households combined. Stargate's Abilene campus was engineered toward nearly one gigawatt of power.

That creates an unusual problem.

AI companies can order more chips. But building substations, transmission lines and generation capacity takes time.

The book points out that power infrastructure permitting can take a decade or more in some markets, while AI companies are trying to expand in years rather than decades.

This could make electricity one of the defining constraints of the next phase of artificial intelligence.

Stargate is also a construction story

The infrastructure race is already visible on the ground.

Stargate's Abilene campus spans more than 1,000 acres and roughly four million square feet. Additional sites have been announced across several U.S. states, with individual developments running into the billions of dollars.

That means the AI economy is creating demand far beyond software engineers.

Electricians, welders, HVAC specialists, fibre technicians, construction workers, engineers, security personnel and facilities teams all become part of the AI supply chain.

The announced sites are projected to create more than 25,000 direct jobs, with additional employment expected through supporting industries.

Why this matters beyond America

Africa is not currently a Stargate-scale hosting location. The book points to power-grid limitations as one of the major reasons. But that does not mean Africa is irrelevant to the story.

If global compute becomes cheaper and more available, entrepreneurs and businesses can access increasingly capable AI tools without owning the data centres producing that computing power.

That could matter enormously for African SMEs, freelancers and technology businesses. The opportunity may be less about building the next Abilene and more about using the intelligence produced by infrastructure built elsewhere to solve African problems locally.

The bigger lesson

Project Stargate changes the question we should ask about artificial intelligence.

Instead of only asking:

“What can AI do?”

we should also ask:

“What does it take to make AI possible?”

The answer involves capital, land, chips, electricity, construction, cooling, networks and people.

That is the story Project Stargate explores — not AI as an abstract technology, but AI as an industrial infrastructure race.

Read the full book:
Project Stargate: Understanding the AI Infrastructure Race

Frequently Asked Questions

What is Project Stargate?

Project Stargate is a large-scale AI infrastructure initiative announced in 2025 with an ambition to invest up to $500 billion in U.S. AI infrastructure by 2029.

Why does AI need so much infrastructure?

Advanced AI requires enormous computing capacity, which means specialised chips, data centres, networking, cooling and electricity.

What is the biggest infrastructure challenge for AI?

Project Stargate identifies electricity and grid capacity as a potentially greater constraint than chips as AI infrastructure expands.

How many jobs could Stargate create?

The announced sites are projected to create more than 25,000 direct onsite jobs, particularly in skilled trades and infrastructure-related roles

Does Africa need to build its own Stargate?

Not necessarily. The book argues that Africa can benefit from increasingly accessible AI tools even without hosting gigawatt-scale data centres, while continuing to consider its own digital sovereignty and compute strategy.