AI Is Getting Cheaper. The World’s Electricity Demand Is Not

rtificial intelligence is becoming cheaper to use, but that does not mean the world will consume less computing power. In fact, the opposite may be happening. As AI models become more efficient and the cost of running them falls, companies can integrate them into more products, more services and more everyday business processes. Lower costs are creating more demand, and that demand is beginning to put pressure on the physical infrastructure behind artificial intelligence.

The International Energy Agency expects electricity consumption from data centers to continue rising rapidly through the end of the decade, with artificial intelligence becoming one of the most important drivers of this growth. The expansion is particularly significant because AI workloads require large amounts of computing capacity and increasingly sophisticated infrastructure.

This creates an important contradiction for the technology industry. AI is becoming more efficient at the individual task level, yet the total amount of AI being deployed continues to expand. A company may spend less to process a single request, but if millions of additional users and businesses begin making those requests, the overall demand for computing can still increase significantly.

The difference between software and infrastructure is becoming increasingly important. A software company can develop and release a new product relatively quickly, but the infrastructure required to support millions of users cannot be built at the same speed. Data centers require land, construction, cooling systems, networking equipment, advanced processors and reliable access to electricity. Expanding a power grid or building new generation capacity can take years.

This is turning electricity into a strategic resource for the AI industry. Access to power is becoming an increasingly important consideration when companies decide where to build new data centers and expand computing capacity. The availability of electricity can determine how quickly a project can move forward and, in some cases, whether it can move forward at all.

The implications extend beyond the largest technology companies. Energy producers, semiconductor manufacturers, data center developers and infrastructure investors are becoming part of the same economic story. The boundaries between technology, energy and infrastructure are becoming increasingly difficult to separate.

Capital is following this transformation. SoftBank’s acquisition of DigitalBridge for $4 billion is one example of how financial and technology companies are positioning themselves around the infrastructure required for the next phase of AI growth. Data centers and digital infrastructure are increasingly being treated as strategic assets rather than simply supporting services.

This could also influence where the next generation of technology infrastructure is built. Regions with reliable electricity, available land, strong transmission networks and the ability to approve large infrastructure projects efficiently could become increasingly attractive to AI companies and investors.

For years, the central question of the AI race was which company could build the most capable model. That question remains important, but another question is becoming harder to ignore. How much infrastructure can the global economy build to support the amount of AI the world wants to use?

That question could become one of the defining economic issues of the next decade.

Artificial intelligence began as a software revolution, but its future is increasingly tied to physical infrastructure. Every model, AI agent and automated system ultimately requires computing hardware somewhere, and that hardware requires electricity.

The cheaper AI becomes, the more widely it can be deployed. The more widely it is deployed, the greater the demand for computing infrastructure becomes. As that infrastructure expands, electricity becomes one of the most important resources in the entire AI economy.

The next stage of artificial intelligence may therefore be decided not only by algorithms and models, but by energy, infrastructure and capital.

AI may be getting cheaper. But the physical world required to run it is becoming increasingly valuable.