Global Data Center Demand Will Double by 2030
Rising AI token usage and complex reasoning models will force a rapid expansion of global data infrastructure.
Updated on Sept. 23, 2026 in Data Centers
Live Poll
Will the surge in AI data center construction prove to be a sustainable long-term investment?
Global data center demand is projected to double by 2030 as AI token consumption reaches 118 quadrillion tokens. This expansion will be driven by the emergence of reasoning models that consume 15 times more tokens per request than traditional generative AI.
Why it matters
The shift toward agentic AI models requires significant increases in token production capacity and localized inferencing workloads. These infrastructure demands are forcing a fundamental evolution in how data centers are built to handle greater computational loads.
AI model performance and complexity increased one million times between 2016 and 2026, while costs per token have decreased by a factor of 10 annually. By 2030, inferencing is projected to account for 60% of all data center workloads.
The players
Data center operators
These entities manage the physical infrastructure and computational facilities required for large-scale cloud and AI processing.
The details
Data center operators are pivoting toward standardized and modular designs to accelerate deployment speed and efficiency, incorporating liquid cooling and DC power. Because inferencing must occur near where tasks are executed, the architecture is now divided into five layers covering energy, chips, infrastructure, models, and applications.
Timeline
Performance grew one million times between 2016 and 2026.
Massive build-outs emerged in Europe and North Africa in 2024.
Global data center demand is expected to double by 2030.
The Tech Race
The transition to modular data centers marks a departure from legacy, custom-built facilities that could not scale at the pace required by modern AI. This race is now defined by the ability to rapidly deploy standardized hardware to meet the exponential growth in global token demand.
As data centers localize inferencing tasks to reduce latency, users may experience faster, more responsive AI interactions in their daily applications. However, the current bottleneck in U.S. data center capacity could lead to temporary service deployment delays or varied AI performance by region.
The takeaway
The move toward modular data centers highlights that physical infrastructure has become the primary constraint on AI development. To keep pace, operators must prioritize energy-efficient designs that can handle the massive token-per-request demands of reasoning models.
Further reading
For more information on infrastructure requirements, visit Data Centers.
Live Poll
Will the surge in AI data center construction prove to be a sustainable long-term investment?
