Analyst(s): Brendan Burke, Nick Patience, Olivier Blanchard
Publication Date: March 12, 2026
Because new grid-connected power takes years longer to come online than the current pace of AI infrastructure-related capital expenditure, power delivery is emerging as the primary constraint on AI deployments. In response to this power generation gap, on-site power generation is transitioning from a temporary fix to a permanent strategy, with industry professionals expecting 33% of data centers to operate on 100% off-grid power by 2030. This constraint is also accelerating the adoption of high-efficiency hardware and architectures, such as NVIDIA’s 800 VDC systems and integrated Supermicro solutions, while simultaneously pushing inference workloads toward the network edge to maximize compute per watt.
Key Points:
- The massive capital expenditure for AI infrastructure is facing a structural power generation gap because new grid-connected power generation cannot come online as quickly as data centers are being built.
- In response to grid inadequacy, on-site power generation is transitioning from a bridge to a permanent strategy, with professionals expecting 33% of data centers to operate on 100% onsite power by 2030, using solutions such as fuel cells to circumvent multi-year utility interconnection queues.
- The power constraint is creating pressure toward efficiency improvements and network edge processing, accelerating the adoption of new hardware and architectures – such as NVIDIA’s 800 VDC systems and integrated Supermicro energy efficiency solutions – to maximize compute per watt.
Overview:
The massive capital expenditure currently directed toward AI infrastructure is facing a structural power generation gap, establishing power delivery as the primary constraint on AI deployments. The five largest US hyperscalers have committed between $660 billion and $690 billion in CapEx for 2026, with roughly 75% focused on AI compute and data centers. However, the fundamental structural problem is that data centers can be built in 12 to 18 months, while new grid-connected power generation takes three to seven years or more to come online, creating a severe and tangible bottleneck. Global data center power demand is projected to more than double by 2030, reaching 945 TWh. Compounding this issue, the US grid interconnection queue holds approximately 2,600 GW of capacity, more than twice the entire installed US power fleet. This imbalance has already resulted in one major hyperscaler disclosing an $80 billion backlog of unfulfillable cloud orders due to power limitations.
In response to this grid inadequacy, on-site power generation is transitioning from a temporary fix to a permanent component of the data center infrastructure strategy. Data center industry professionals now expect 33% of data centers to operate on 100% onsite power by 2030. This shift utilizes modular solutions such as fuel cells, which can be deployed in phases to match the IT load ramp-up, circumventing the multi-year utility interconnection queues and the manufacturing and permitting bottlenecks that plague large gas and nuclear turbine projects. Furthermore, this infrastructure buildout is increasingly debt-funded, and power constraints pose a significant financial risk by slowing the activation of completed data centers, thereby extending the Return on Investment (ROI) timeline.
Efficiency Improvements Will Alleviate These Constraints, But Not Just Yet
The power constraint is simultaneously creating intense pressure for efficiency improvements and network edge processing to maximize compute per watt. The International Energy Agency (IEA) defines a High Efficiency Case that suggests aggressive energy savings, driven by hardware and software improvements, could flatten global data center electricity demand growth by 20% by 2035. To achieve this, hardware and power distribution efficiency must be maximized at the rack level. NVIDIA is pioneering the transition to 800 Volts of Direct Current (VDC) architectures, which allows for 157% more power to be transmitted through the same copper cross-section and results in a 1% overall efficiency improvement by streamlining the power tree. Similarly, solutions such as Super Micro’s Data Center Building Block Solutions (DCBBS) integrate modular subsystems to compress the infrastructure footprint and reduce power consumption for massive AI clusters. Operators must also deploy advanced Battery Energy Storage Systems (BESS) with simulation-based software to buffer and smooth the severe, sub-second power swings caused by erratic AI training workloads, ensuring facility activation and mitigating financial risk.
Look to Edge-Based AI to Also Help Solve Power Constraints
The structural power generation gap also forces a direct pressure to shift workloads toward the network edge. Moving AI inference – the majority of daily AI operations – closer to the point of need bypasses grid congestion and operates outside of constrained utility zones. This strategy leverages the increasing installed base of sophisticated, low-power edge devices, which are orders of magnitude more energy-efficient for repetitive inference tasks than transmitting data to the cloud for processing. By distributing power consumption across millions of local devices, the industry can significantly reduce the total load on new, centralized “AI Factories,” decoupling compute from the grid and ensuring power delivery does not become the ultimate ceiling for the deployment of AI services.
The full report, “AI Grid Constraints Will Push Over 33% of Data Centers Off-Grid by 2030,” is available via subscription to Futurum Intelligence’s IQ service – click here for inquiry and access.
Futurum clients can read more in the Futurum Intelligence Platform, and non-clients can learn more in the AI Platforms Practice, the Semiconductors, Supply Chain, & Emerging Technology Practice, and the Futurum Intelligent Devices Practice.
About the Futurum AI Platforms Practice
The Futurum AI Platforms Practice provides actionable, objective insights for market leaders and their teams so they can respond to emerging opportunities and innovate. Public access to our coverage can be seen here. Follow news and updates from the Futurum Practice on LinkedIn and X. Visit the Futurum Newsroom for more information and insights.
About the Futurum Semiconductors, Supply Chain, & Emerging Technology Practice
The Futurum Semiconductors, Supply Chain, & Emerging Technology Practice provides actionable, objective insights for market leaders and their teams so they can respond to emerging opportunities and innovate. Public access to our coverage can be seen here. Follow news and updates from the Futurum Practice on LinkedIn and X. Visit the Futurum Newsroom for more information and insights.
About the Futurum Intelligent Devices Practice
The Futurum Intelligent Devices Practice provides actionable, objective insights for market leaders and their teams so they can respond to emerging opportunities and innovate. Public access to our coverage can be seen here. Follow news and updates from the Futurum Practice on LinkedIn and X. Visit the Futurum Newsroom for more information and insights.
Author Information
Brendan is Research Director, Semiconductors, Supply Chain, and Emerging Tech. He advises clients on strategic initiatives and leads the Futurum Semiconductors Practice. He is an experienced tech industry analyst who has guided tech leaders in identifying market opportunities spanning edge processors, generative AI applications, and hyperscale data centers.
Before joining Futurum, Brendan consulted with global AI leaders and served as a Senior Analyst in Emerging Technology Research at PitchBook. At PitchBook, he developed market intelligence tools for AI, highlighted by one of the industry’s most comprehensive AI semiconductor market landscapes encompassing both public and private companies. He has advised Fortune 100 tech giants, growth-stage innovators, global investors, and leading market research firms. Before PitchBook, he led research teams in tech investment banking and market research.
Brendan is based in Seattle, Washington. He has a Bachelor of Arts Degree from Amherst College.
Nick Patience is VP and Practice Lead for AI Platforms at The Futurum Group. Nick is a thought leader on AI development, deployment, and adoption - an area he has researched for 25 years. Before Futurum, Nick was a Managing Analyst with S&P Global Market Intelligence, responsible for 451 Research’s coverage of Data, AI, Analytics, Information Security, and Risk. Nick became part of S&P Global through its 2019 acquisition of 451 Research, a pioneering analyst firm that Nick co-founded in 1999. He is a sought-after speaker and advisor, known for his expertise in the drivers of AI adoption, industry use cases, and the infrastructure behind its development and deployment. Nick also spent three years as a product marketing lead at Recommind (now part of OpenText), a machine learning-driven eDiscovery software company. Nick is based in London.
Olivier Blanchard is Research Director, Intelligent Devices. He covers edge semiconductors and intelligent AI-capable devices for Futurum. In addition to having co-authored several books about digital transformation and AI with Futurum Group CEO Daniel Newman, Blanchard brings considerable experience demystifying new and emerging technologies, advising clients on how best to future-proof their organizations, and helping maximize the positive impacts of technology disruption while mitigating their potentially negative effects. Follow his extended analysis on X and LinkedIn.
