AI’s role in propelling the energy sector

The energy sector stands on the brink of a transformative era, driven by the advent of AI technologies. AI’s impact on energy consumption is already making headlines with forecasts of surging electricity demand, a nuclear renaissance, a revival in gas-fired power and increased greenhouse gas emissions. However, AI’s ability to enhance how energy is produced, traded and distributed could be as groundbreaking, if not more.

AI technologies offer capabilities to forecast tomorrow’s energy needs, produce traditional energy resources in a more efficient and low-carbon manner, optimally site and operate new energy projects, deliver energy where and when it is needed, accelerate experience-led learning and the discovery of new materials, and more seamlessly trade energy commodities. One might question whether there is anything AI cannot do.

Experience shows, however, that capturing the technology’s full potential will not be easy. It requires a tight alignment between technology organizations and the workforce; addressing security and regulatory issues; formalizing relationships with multiple external entities such as industrial equipment suppliers, hyperscalers and trading partners; and getting a host of other issues right. While companies report impressive results from one-off deployments, they often struggle to achieve enterprise scale that alters overall corporate performance.

Additionally, the sheer volume and diversity of potential AI use cases can be overwhelming, making it difficult to assess how the technology might progress across the energy sector. S&P Global introduced a framework to categorize these opportunities, which can serve as a road map for industry uptake.

The following sections outline these classes of AI opportunities for the energy sector, including the magnitude and nature of their potential impacts, time to value realization and associated risk factors. The hope is that characterizing the initial outcomes and resulting changes in industry behaviors can lead to a better understanding of how AI might transform energy.

While AI presents a range of aspirational opportunities to advance and manage increasingly complex energy systems, the applications most used today tend to be straightforward and practical. They target modest improvements in operational and energy efficiency, resulting in higher productivity and lower costs and emissions. Their benefits should not be discounted, however, as cost competitiveness remains a primary barrier to advancing many clean technology solutions, and collectively, their impact can be quite large. AI is also one of the few options available for companies to deliver near-term benefits while incurring minimal additional costs.

Key opportunities include:

• Individual worker productivity. The most common expressions of AI technology today are the digital assistants deployed to support office-based staff and technical teams in their daily work, automating routine tasks.
• Operational efficiency. AI tools are applied to predict or optimize the performance of individual industrial assets or processes such as autonomous drilling, maintenance planning and scheduling. According to S&P Global Commodity Insights analysis, individual assets are realizing 10%-25% operating cost reductions and 3%-8% productivity increases through such approaches.
• Energy efficiency. AI is proving highly effective in increasing the energy efficiency of industrial equipment and processes. Machine learning algorithms calculate optimal settings for machinery such as compressors and turbines within efficiency “sweet spots” across a range of conditions without sacrificing productivity. Companies are achieving 5%-8% energy efficiency improvements through such approaches, translating directly to equivalent emissions reduction levels.

Discussions with energy companies reveal faster technological development in these areas since the advent of generative AI in late 2022. Alongside this additional investment and increased executive attention, however, are more frequent inquiries on the benefits delivered and greater scrutiny on regulatory and security compliance issues. Ultimately, the value of these AI solutions is determined by the effectiveness of the tools developed and the willingness of the technical workforce to use them; these are largely organizational, not technological, issues.

The energy transition is resulting in expanded energy value chains. More highly distributed and intermittent generation sources, increased operational interdependencies, greater market integration across fuels and regions, more regulatory and market compliance demands, and larger quantities of just about everything associated with the value chain are increasing the complexity of our energy systems. Managing these intricacies with traditional approaches may lead to the underutilization of available resources, higher prices and emissions, and greater instability and volatility.

Digitalization in general, along with more focused AI, can help manage this complexity and operate energy systems closer to their maximum physical capacities while optimizing targeted objective functions such as throughput, emissions and profit. Two opportunities stand out:

• Grid planning and management. The electric power grid is a prime example of the complexity that accompanies the energy transition. Efforts are underway to harness AI’s potential to improve the planning and operation of grids. Leveraging AI capabilities to enhance the speed and efficiency of grid planning simulations can lead to a broader range of scenarios considered and more reliable outcomes with increased renewables penetration. For example, Chile’s transmission operator is using AI to increase simulation speeds by 86%. Machine learning routines that more accurately forecast day-ahead solar and wind generation and battery discharge rates enable grid management that optimizes the full potential of the energy system’s components, thus minimizing the dispatch of more carbon-intense resources.
• Energy trading. The same capabilities that can raise grid performance are valuable within complex financial systems to deliver better energy trading outcomes. Incorporating more accurate weather patterns, consumer price elasticity behavior and other factors improves the accuracy of forecasts and trading decisions.

Large and complex systems extending across the power sector to traditional energy segments (such as the LNG value chain) and new energy domains (such as blue hydrogen production) have similar opportunities, with increasing operational, commercial and regulatory complexities and interdependencies.

In pursuing such opportunities, companies are finding that the same factors driving energy system complexity also complicate efforts to deploy the AI solutions that help manage them. Formal agreements with external entities along the energy value chain are needed to gain access to proprietary information and to take appropriate action in response to the technology’s outputs. Regulators are also justifiably cautious in approving new approaches to operating critical infrastructure, with cybersecurity and national security additional concerns. Finally, building and maintaining the models requires constant focus. These complications can slow the pace of adoption.

The advance of new forms of low-carbon energy and emissions reduction technologies such as methane sensors is hindered by their cost compared with legacy resources and the additional financial burden they place on existing operations. Ongoing innovation is needed to improve their cost competitiveness and ease investment decisions. AI’s operational efficiency gains cited previously are an important component of this equation, but significant reductions in capital expenditure for nascent technologies, such as small modular reactors and solar paired with long-duration storage, and step changes in operating capabilities and costs are needed.

The types of innovation required to realize these improvements are occurring through “learning by doing” and traditional research and development programs. The energy transition is time sensitive, however, and AI can help speed up such processes:

• New material discovery. The energy transition is being driven in part by new materials — from catalysts that reduce the cost of hydrogen production to alternative battery chemistries that allow manufacturers to diversify their supply chains and extend storage durations. At the same time, traditional methods for identifying the most promising candidates are often arbitrary and time-consuming. The pharmaceutical industry has demonstrated the value of applying AI to drug discovery, and the energy sector is starting to follow suit in the search for new chemistries that not only lead to lower costs, but display the durability, strength and other qualities needed to perform under extreme conditions (e.g., Georgia Institute of Technology’s collaboration with Meta to identify a comprehensive database of new metal-organic framework-based carbon capture materials).
• Faster learning curves. Industrial development follows a similar pattern, with subsequent iterations of an industrial asset trending toward lower costs through “learning curve” impacts. The energy sector has benefited from this effect in both traditional domains (e.g., steady cost declines in unconventional wells) and recent clean energy developments (e.g., in part, the 25% reduction in battery costs on a dollar-per-kilowatt-hour basis between 2019 and 2024). Companies are now looking to leverage AI capabilities to speed down learning curves faster in new energy areas characterized by far fewer deployments. Applying computational methods to analyze operational performance data across a range of conditions, developers can feed that information back into the next project design cycle to further optimize cost and performance.

Occidental Petroleum is taking this approach across its portfolio of upcoming direct air capture plants in the Permian Basin. According to Occidental CEO Vicki Hollub at the Energy Intelligence Forum 2022, the company intends to apply computational modeling and methods to optimize individual plant performance and “make step change differences from the first to the second, third, fourth units.”

Although promising, these AI applications are expected to have the longest runway to impact because they fundamentally alter the way most energy players operate. New R&D methods take time to ramp up and pay dividends, and closing the asset operations-design cycle through machine learning methods demands new types of relationships between operators and engineering firms.

There is an ambitious vision for AI’s potential to transform how energy is generated and distributed and how its associated emissions are abated. This stands in contrast to how the technology is being applied today. A recent S&P Global Commodity Insights survey of 36 clean technology firms informs this view, with most reporting that AI is primarily used in office-based tasks to improve workforce efficiency. While the oil and gas sector is well versed in the technology, with long-established data science programs and a range of industrial applications, a substantial gap remains between potential value and its capture.

Despite the excitement surrounding AI, its impact on the energy industry is likely to be evolutionary rather than revolutionary. This view is based on an assessment of how digital technologies often progress in industrial settings. AI is ultimately a tool used by humans, and their acceptance of the technology is based on trust that the technology will perform as expected. Trust is built over time and typically begins with modest and proven applications before expanding into larger and more complex ones. The energy sector is in the early stages of this journey.

S&P Global will closely follow the AI use cases that emerge across the entire energy value chain, including benefits delivered and organizational scale achieved. The expectation is that early deployments will be used to improve efficiency before moving on to more sophisticated applications and, ultimately, greater organizational capabilities. Signposts to watch include acceptance by regulators and the full range of data sharing, technology, operational and research partnerships established. These collective factors will largely determine the pace and character of AI’s ability to advance the energy sector.