The true cost of AI is beginning to surface

The true cost of artificial intelligence is starting to become apparent. Initially, AI appears affordable due to monthly subscriptions and user-friendly interfaces. However, as the technology evolves, its true expense is hidden within the complex systems it relies on, such as data centers, electricity, water, chips, cloud providers, and compliance processes.

The next stage of AI development involves agents that perform multiple tasks like searching, planning, calling tools, and checking work. This shift changes the economics from a per-prompt pricing model to an infrastructure-based pricing model. This transition is evident in GitHub's decision to move Copilot from a flat-rate to a usage-based billing system, which will be calculated through AI credits based on token consumption.

While individual AI tasks may seem cheap, the underlying computational steps can involve hundreds of operations. This discrepancy between user perception and actual computational effort highlights the beginning of AI's true cost structure. AI's adoption is akin to a subsidy phase, with large tech companies competing for market share and users, offering low-cost entry points to accelerate adoption.

However, this phase may lead to an eventual increase in total costs as AI becomes embedded in various sectors like finance, healthcare, and logistics.

The infrastructure required to support AI, such as data centers and electricity consumption, is growing at an unprecedented rate. The International Energy Agency predicts that global electricity consumption from data centers could more than double by 2030, accounting for nearly 3% of global electricity usage. This surge in power demand could strain existing power grids, creating a new set of infrastructure challenges that may fall on technology companies or be spread through electricity rates, tax incentives, or public investment.

In summary, the true cost of AI extends beyond subscription fees to encompass the entire infrastructure it depends on, including data centers and electricity. As AI becomes more integrated into various industries, understanding and managing these costs will be crucial to avoid potential disruptions and ensure sustainable growth.