Key Takeaways
- • As of early 2026, 190 GW of hyperscale data center capacity has been announced globally across 777 projects (~148 GW planned, ~21 GW under construction, ~12 GW operational), with global data center electricity consumption projected to more than double by 2030.
- • Grid connection delays are the primary infrastructure bottleneck: while data centers can be constructed in 12-18 months, grid connections currently require 5-7 years, causing over 25% of the 110 data center projects scheduled for 2025 to experience delays due to power, permitting, and construction constraints.
- • Hyperscalers are increasingly accepting the complexity of on-site power generation to secure capacity certainty, timelines, and emissions control, fundamentally shifting the traditional data center buildout model used for nearly 30 years.
- • In April 2026, President Trump invoked Section 303 of the Defense Production Act to designate large-scale grid infrastructure as essential to national defense, authorizing emergency federal financing to expand domestic supply chain capacity for critical infrastructure components.
- • The AI data center stack encompasses six infrastructure opportunity areas: permitting and site selection, power generation, transmission and power conversion, cooling technologies, hardware systems, and software orchestration—creating a newly disrupted market after three decades of industry stability.
Summary
Bessemer Venture Partners' AI data center stack roadmap identifies a fundamental infrastructure crisis reshaping the global computing landscape. The scale is unprecedented: 190 GW of announced hyperscale data center capacity across 777 projects signals that AI has become the most power-intensive computing workload in history. This buildout trajectory—with roughly 148 GW planned, 21 GW under active construction, and 12 GW already operational as of early 2026—is forcing a reckoning with energy infrastructure that extends far beyond traditional data center economics.
The critical bottleneck is not construction speed but grid connectivity. Data center facilities can be built within 12-18 months, yet connecting to existing electrical grids currently requires 5-7 years of regulatory, permitting, and infrastructure work. This mismatch has created tangible market disruption: more than 25% of the 110 data center projects scheduled to launch in 2025 experienced delays attributable to power, permitting, and construction constraints. The consequence is that U.S. data centers will soon consume more electricity than all energy-intensive manufacturing combined, while global data center electricity consumption is projected to exceed double current levels by 2030.
In response, hyperscalers—including major cloud and AI infrastructure operators—are increasingly accepting the operational complexity of developing and managing on-site power generation systems. This represents a strategic pivot from the historical model of purchasing power from existing grid infrastructure. By building proprietary generation capacity, these firms gain certainty over construction timelines, operational capacity, and emissions profiles, effectively internalizing what was previously an external constraint.
The federal government has recognized the national security implications of this infrastructure buildout. In April 2026, President Trump invoked Section 303 of the Defense Production Act to formally designate large-scale grid infrastructure as essential to national defense. This designation authorized emergency federal financing mechanisms to expand domestic supply chain capacity for critical grid infrastructure components, positioning AI infrastructure buildout as a strategic national priority aligned with competitive positioning in the global AI race.
Bessemer's framework identifies six discrete opportunity areas within the AI data center stack: permitting and site selection, power generation systems, transmission and power conversion technologies, cooling infrastructure, hardware systems, and software orchestration platforms. This layered infrastructure model has remained largely stable for nearly 30 years, but the explosive demand for AI compute and the consequent energy requirements have triggered the first major disruption cycle in the sector's modern history.
The critical bottleneck is not construction speed but grid connectivity. Data center facilities can be built within 12-18 months, yet connecting to existing electrical grids currently requires 5-7 years of regulatory, permitting, and infrastructure work. This mismatch has created tangible market disruption: more than 25% of the 110 data center projects scheduled to launch in 2025 experienced delays attributable to power, permitting, and construction constraints. The consequence is that U.S. data centers will soon consume more electricity than all energy-intensive manufacturing combined, while global data center electricity consumption is projected to exceed double current levels by 2030.
In response, hyperscalers—including major cloud and AI infrastructure operators—are increasingly accepting the operational complexity of developing and managing on-site power generation systems. This represents a strategic pivot from the historical model of purchasing power from existing grid infrastructure. By building proprietary generation capacity, these firms gain certainty over construction timelines, operational capacity, and emissions profiles, effectively internalizing what was previously an external constraint.
The federal government has recognized the national security implications of this infrastructure buildout. In April 2026, President Trump invoked Section 303 of the Defense Production Act to formally designate large-scale grid infrastructure as essential to national defense. This designation authorized emergency federal financing mechanisms to expand domestic supply chain capacity for critical grid infrastructure components, positioning AI infrastructure buildout as a strategic national priority aligned with competitive positioning in the global AI race.
Bessemer's framework identifies six discrete opportunity areas within the AI data center stack: permitting and site selection, power generation systems, transmission and power conversion technologies, cooling infrastructure, hardware systems, and software orchestration platforms. This layered infrastructure model has remained largely stable for nearly 30 years, but the explosive demand for AI compute and the consequent energy requirements have triggered the first major disruption cycle in the sector's modern history.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.