Yet the rapid rise of artificial intelligence is exposing the limits of both analogies.
Modern AI systems require enormous computing power. Training advanced machine-learning models involves tens of thousands of specialised processors operating simultaneously inside vast data centres designed to run continuously. These facilities are extraordinarily energy-intensive.
According to the International Energy Agency, data centres consumed around 415 terawatt-hours of electricity globally in 2024, roughly 1.5% of total global electricity demand. As AI workloads expand, that figure is expected to rise sharply. Some projections suggest that data-centre electricity consumption could exceed 1,000 terawatt-hours by 2030 ” roughly comparable to the current electricity consumption of Japan.
The scale of infrastructure required is striking. Training frontier AI models can involve tens of thousands of specialised graphics processing units (GPUs) running simultaneously for weeks. These computing clusters draw enormous amounts of power and generate intense heat, which in turn requires energy-intensive cooling systems.
Electricity is only one constraint. Large data centres also require substantial water resources for cooling. Major facilities can consume millions of gallons of water per day, in some cases comparable to the water usage of a small town. Even routine digital interactions carry a hidden environmental footprint. Some estimates suggest that a single AI query may indirectly consume several hundred millilitres of water once data-centre cooling requirements are included.
Technology companies are attempting to address the resulting emissions through carbon-removal purchases and renewable-energy investments. Microsoft, for example, has committed roughly $1.7 billion to carbon-removal contracts, including projects that capture methane from agricultural waste and prevent it from entering the atmosphere.
Such efforts aim to offset emissions generated by the growing energy demands of digital infrastructure.
Yet these measures highlight the scale of the underlying challenge. Carbon removal can compensate for emissions, but it does not eliminate the rapidly expanding electricity demand created by AI systems.
Data alone does not produce intelligence. It must be processed by machines that require vast amounts of electricity, advanced semiconductor chips and massive computing clusters. The true engine of artificial intelligence is not simply data, but computation ” and computation ultimately depends on energy.
Much of that energy still comes from fossil fuels. Despite rapid investment in renewable power, large portions of the global electricity grid remain dependent on coal, oil and natural gas ” fuels formed over millions of years from ancient biological matter.
In other words, the digital revolution remains deeply connected to the geological past. Beneath the sleek interfaces of chatbots, recommendation engines and AI assistants lies an infrastructure powered, at least in part, by the compressed remains of prehistoric life.
The slogan that data is the new oil was always a metaphor. But in a curious way, the age of artificial intelligence has brought the comparison closer to reality than its creators might have imagined. The digital economy may run on algorithms and silicon, but much of its energy still comes from the ancient carbon stored beneath the Earth’s surface.
In the age of artificial intelligence, the machines of the future still run ” quite literally ” on dinosaurs.
