For the past few years, the artificial intelligence investment story has been dominated by one them - chips.

As companies rushed to develop and deploy generative AI, demand for advanced semiconductors exploded. Graphics processing units (GPUs), particularly those produced by Nvidia, became the foundation of the AI boom and some of the most sought-after pieces of technology in the world.

But as AI moves beyond experimentation and into large-scale adoption, another challenge is becoming impossible to ignore.

Building smarter models is only one part of the equation. Running them requires enormous physical infrastructure: data centres, electricity, cooling systems, networking equipment, and reliable grid connections capable of supporting industrial levels of power demand.

AI is no longer just a software story. It is rapidly becoming an infrastructure story.

The first race was about computing power. The next one may be about everything required to support it.

AI's Industrial Buildout Has Begun

The scale of investment flowing into AI infrastructure is extraordinary.

Microsoft, Alphabet, Amazon, and Meta are committing hundreds of billions of dollars towards expanding AI capacity. New data centres are being built, cloud networks are expanding, and companies are spending heavily on servers, specialised equipment, and the infrastructure needed to keep everything running.

A modern AI data centre looks less like a traditional technology asset and more like a digital factory.

Thousands of processors operate continuously, supported by storage systems, high-speed networking, power equipment, and advanced cooling. These facilities require years of planning and billions of dollars before they begin generating meaningful returns.

This marks an important shift.

For decades, technology companies benefited from the economics of software: create a product once and scale it globally with relatively little physical investment. AI changes that model. The companies leading this revolution are increasingly dependent on real-world assets.

Algorithms still matter. Intellectual property still matters. But access to infrastructure is quickly becoming a competitive advantage.

Electricity Is Becoming the New Bottleneck

Nothing highlights this shift better than energy.

The International Energy Agency expects global data-centre electricity consumption to more than double by 2030, reaching around 950 terawatt-hours annually. AI-focused facilities are expected to be a major contributor as models become larger and require more computing power.

However, the challenge is not simply producing more electricity.

The bigger problem may be delivering it.

Across many regions, electricity grids are struggling to keep pace with demand. Transmission projects can take years to complete, connection queues are growing, and shortages of critical equipment such as transformers have created additional pressure.

Power is no longer just another operating cost. It is becoming a strategic resource.

A company can own thousands of advanced AI chips, but without enough electricity those chips cannot create value. Computing power is only useful when the infrastructure behind it exists.

The AI industry may have mastered some of the world's most complicated digital problems, but its next challenge could come from the physical world.

The Opportunity Is Moving Beyond Semiconductors

None of this means the semiconductor story is over.

Advanced chips remain the foundation of artificial intelligence, and demand for leading processors continues to be incredibly strong. But the AI opportunity is becoming much broader than chips alone.

Utilities are attracting attention as technology companies search for reliable long-term power supplies. Data-centre operators are expanding aggressively. Manufacturers of electrical equipment, transformers, and power-management systems are becoming increasingly important parts of the AI supply chain.

Networking is another crucial area.

As AI systems become larger, moving information efficiently between thousands of processors becomes essential. Faster connections can improve performance and reduce bottlenecks, which is why technologies such as optical networking and photonics are gaining more attention.

Cooling is also becoming a major focus.

AI servers generate significant heat, especially as more computing power is packed into smaller spaces. Advanced cooling solutions, including liquid cooling, are becoming increasingly important to improving efficiency and performance.

The first stage of AI rewarded companies building the brains of the system. The next stage may reward those building the infrastructure that keeps those brains running.

Some of The Next Winners May Be Measured in Megawatts

Technology companies understand the challenge and are already searching for solutions.

Chip designers are increasingly focused on energy efficiency, not just raw performance. Better chip designs, advanced packaging, and new technologies aim to increase computing power without creating the same increase in electricity demand.

Future efficiency gains may reduce the amount of energy required for each AI calculation, although overall electricity demand could continue rising as AI adoption expands and more powerful systems are deployed.

At the same time, companies are looking closely at energy supply.

Renewable energy will play an important role, while natural gas remains a key source of reliable generation in many regions. Nuclear power has also returned to the discussion as companies search for dependable, carbon-free electricity that can operate around the clock.

But generating electricity is only one piece of the puzzle.

Transmission lines, substations, transformers, and grid connections all need to expand. A gigawatt of power has little value if it cannot reach the data centres that require it.

That is why AI increasingly represents the meeting point between technology and infrastructure.

The evolution is not moving beyond semiconductors. It is building on top of them.

Chips remain essential, but the next phase of AI growth may create winners in industries that historically sat far outside the technology spotlight: utilities, energy providers, electrical equipment companies, and data-centre operators.

There is an irony in this. One of the most advanced technologies ever created may ultimately depend on some of the oldest economic foundations e.g. energy, infrastructure, and capital.

The race to build more powerful AI is far from finished. But the next winners may not only be determined by who has the best algorithms or the fastest chips.

They may also be determined by who can keep the lights on.

References

1. www.iea.org
2. www.iea.org
3. www.iea.org
4. www.reuters.com
5. www.reuters.com
6. www.reuters.com