Energy / North America
The Evolving Relationship Between Energy and AI
Investment in artificial intelligence is rapidly increasing, with significant implications for the energy sector. Major tech companies are expected to exceed global investments in oil and natural gas production by 75% this year, indicating a shift in capital allocation towards AI technologies.
Source material: Podcast episode: The relationship between energy and AI is evolving rapidly
Summary
Investment in artificial intelligence is rapidly increasing, with significant implications for the energy sector. Major tech companies are expected to exceed global investments in oil and natural gas production by 75% this year, indicating a shift in capital allocation towards AI technologies.
The International Energy Agency (IEA) is reassessing the energy-AI relationship due to advancements since their last report in April 2025. AI applications have evolved from basic text queries to more complex tasks, affecting energy consumption and infrastructure requirements.
AI hardware efficiency is improving rapidly, with annual gains of 50-60%, while model efficiency can increase by two to ten times each year. However, certain AI services, especially generation, are significantly more energy-intensive, contributing to an overall rise in energy demand.
The energy sector is facing significant bottlenecks due to increased demand for transformers and gas turbines, resulting in the highest global orders for gas turbines since 2000. Concurrently, the tech sector is grappling with supply shortages, particularly in high bandwidth memory, which is critical for advanced AI data centers.
Perspectives
Proponents of AI in Energy
- Highlight the rapid increase in investment in AI technologies, indicating a shift in focus towards AIs potential in the energy sector
- Argue that AI can significantly enhance energy efficiency and reduce consumption by 2035
Neutral / Shared
- Acknowledge that while AI has the potential to optimize energy systems, current challenges such as data accessibility and skill shortages remain
- Recognize that the relationship between data center growth and electricity prices is complex and context-dependent
Metrics
capex
$400 million USD
capital expenditures last year
This figure highlights the rapid growth in AI investment compared to previous years
Last year that figures did at $400 million
75%
expected growth in capital expenditure this year
A 75% increase signifies a dramatic escalation in AI-related investments
this year is expected to grow by 75%
500 teravot hours kWh
total electricity consumed by all data centers globally last year
This represents a significant portion of global electricity consumption
Last year, about 500 teravot hours of electricity was consumed by all data centers globally.
50%
growth in electricity consumption from AI-specific data centers
This indicates a rapidly increasing energy footprint for AI technologies
AI data centers saw their electricity consumption growing by 50% in this one year period.
70%
increase in global orders for natural gas turbines
This surge indicates a critical demand for energy infrastructure to support AI technologies
we saw global orders for natural gas turbines increased by an astonishing 70%
highest ever level since the year 2000
historical context of gas turbine orders
This highlights the unprecedented demand for energy solutions in the current technological landscape
total orders in 2025 at the highest ever level since the year 2000
65%
percentage of corporate procurement of renewables in the US by the tech sector
This indicates the tech sector's pivotal role in driving renewable energy adoption
the tech sector is responsible for 65% of corporate procurement of renewables in the United States
25 gigawatts
conditional off-take agreements for small modular reactors
This indicates a growing interest in innovative nuclear solutions to meet future energy demands
we tracked about 25 gigawatts of what we call conditional off-take agreements
Key entities
Key developments
Phase 1
Investment in artificial intelligence is rapidly increasing, with significant implications for the energy sector. The International Energy Agency is reassessing the energy-AI relationship due to advancements and evolving demands.
- Investment in AI is rapidly increasing, with major tech companies capital expenditures expected to exceed global investments in oil and natural gas production by 75% this year
- The International Energy Agency (IEA) is reassessing the energy-AI relationship due to significant advancements since their last report in April 2025, emphasizing the importance of staying ahead of developments
- AI applications have progressed from basic text queries to more complex tasks, such as coding and managing long-duration processes, which affects energy consumption and infrastructure requirements
- Hyperscalers, the largest tech companies, are anticipated to invest over $700 billion in capital expenditures by 2026, largely driven by the demand for data centers, surpassing historical investments like NASAs Apollo program
- The energy efficiency of AI varies significantly across different services, complicating the quantification of energy usage for specific tasks
Phase 2
The relationship between energy and artificial intelligence is evolving, with significant increases in energy consumption driven by AI technologies. As AI hardware and model efficiencies improve, the energy demands of AI applications are also rising, particularly in data centers dedicated to AI.
- A simple AI query, like a text generation request, consumes about 0.3 watt hours, which is similar to the energy used by a TV during the same time
- AI hardware efficiency is improving rapidly, with annual gains of 50-60%, while model efficiency can increase by two to ten times each year
- Certain AI services, especially video generation and long-duration tasks, are significantly more energy-intensive, contributing to an overall rise in energy demand for AI
- In 2025, electricity consumption from data centers rose by 17%, with AI-specific data centers seeing a 50% increase, indicating a growing energy footprint
- AI dedicated server racks are expected to consume as much electricity as 65 households, underscoring the concentrated energy demands of AI infrastructure
Phase 3
The energy sector is facing significant bottlenecks due to increased demand for transformers and gas turbines, resulting in the highest global orders for gas turbines since 2000. Concurrently, the tech sector is grappling with supply shortages, particularly in high bandwidth memory, which is critical for advanced AI data centers.
- The energy sector is experiencing significant bottlenecks, particularly with increased demand for transformers and gas turbines, leading to the highest global orders for gas turbines since 2000 and extended delivery wait times
- A shortage of high bandwidth memory is creating new challenges for the technology sector, particularly affecting advanced AI data centers, with supply constraints expected to last two to three years
- Disruptions in the Middle East are impacting energy supply chains, notably affecting helium supplies essential for advanced chip manufacturing, which could have major implications for data center operations
- In response to these challenges, the energy sector is innovating, with projections indicating that renewable energy power purchase agreements will cover half of data center electricity consumption in the coming years
- The tech sector is also advancing nuclear energy, with significant agreements aimed at extending the operational capacity of existing nuclear plants and supporting the development of small modular reactors
Phase 4
The energy sector is increasingly integrating artificial intelligence to enhance efficiency and resilience, despite facing significant challenges such as data accessibility and skill shortages. AI has the potential to reduce global energy consumption significantly by 2035, but current utilization is hindered by regulatory and interoperability issues.
- The energy sector is complex and capital intensive, generating large amounts of data but facing challenges in fully utilizing AI due to issues with data accessibility and a lack of skilled professionals
- AI applications in the energy sector include enhanced weather forecasting for renewable sources and improved battery performance, with the potential to significantly lower global energy consumption by 2035
- Research shows no clear link between the electricity consumption of data centers and rising energy prices, indicating that efficient use of existing infrastructure can accommodate increased demand without driving costs higher
- The analogy of airline capacity utilization suggests that effective infrastructure use in energy can absorb additional demand from data centers without raising prices, unlike scenarios of overcapacity
- Regulatory and interoperability challenges impede the integration of AI into energy systems, restricting real-time communication between devices and the grid, which is crucial for optimizing energy usage
Phase 5
The relationship between energy and artificial intelligence is evolving, with significant increases in energy consumption driven by AI technologies. On-site power generation for data centers is becoming a notable trend due to slow grid connections and rising electricity demands.
- The report notes a significant increase in on-site power generation for data centers in the U.S, driven by slow grid connections, with many projects currently proposed
- On-site power generation adds complexity and costs for data centers, which require high reliability, leading to the need for overbuilt generation facilities and battery storage integration
- There is no clear correlation between electricity consumption and data center growth with rising electricity prices, although demand pressures can still influence pricing due to infrastructure constraints
- A notable trend in the adoption of industrial robots and physical AI is identified, indicating that future industrial competitiveness will hinge on effectively deploying AI technologies to boost productivity
- By 2030, electricity consumption from data centers is projected to rise significantly, highlighting the need for strategic planning to address energy demands and infrastructure development
Phase 6
Electricity consumption from data centers is projected to double by 2030, yet will only represent about 3% of global electricity consumption. The growth in demand may be constrained by supply chain bottlenecks in IT and energy equipment.
- By 2030, electricity consumption from data centers is expected to double, yet will still account for only about 3% of global electricity consumption, highlighting significant supply chain constraints
- The increase in electricity demand from AI and data centers may face limitations due to bottlenecks in IT and energy equipment, as well as grid capacity, despite the potential for more AI applications
- Future trends in electricity consumption remain uncertain; advancements in AI efficiency could either stabilize data center demand or lead to continued growth if monetization aligns with investments
- While energy demand from data centers is largely established, the potential for AI to enhance energy efficiency across the sector is largely untapped, with digital meters and connected appliances currently representing a small portion of total usage
- AI applications in the energy sector could potentially mitigate the rise in energy demand, with analyses indicating that energy savings from AI might surpass total demand growth through 2035