The AI energy paradox: Balancing consumption and efficiency
As New Zealand pursues ambitious renewable energy goals, AI could emerge as a critical tool for transforming our electrical grid's efficiency and sustainability.
However, AI technologies require a lot of power to operate, putting pressure on electricity grids. Since the launch of ChatGPT in 2022, the possibilities of AI systems have become widely acknowledged by both the public and industry, leading to an explosion of discussions, research, use cases, and ideas. Balancing AI’s pressure on the country’s power supply are the opportunities that advances in AI bring for increased energy efficiency and effective renewable integration in the electrical grid.
AI’s intensive consumption of electricity
AI systems and software are energy-intensive technologies, with high energy requirements occurring both in the development process and operations. The electricity demand is split into two distinct phases: the training phase and the operational phase. In the training phase, massive datasets are processed by the AI system, analysing huge amounts of data with the objective of making the AI smarter.
OpenAI, the company behind ChatGPT, is currently training a new AI model, Orion. In six months of training Orion, the training process is already estimated to have consumed the equivalent of the entire annual US steel industry electricity consumption (11 billion kWh a year). The costs for OpenAI are not insignificant either; they have spent an estimated 500 million USD on electricity for the first Orion training run alone. This was for a single AI model.
The second phase of electricity demand from AI systems is the operational or running phase. This is the phase where users interact with the AI model. In February 2023, the chairman of Alphabet (the parent company of Google) indicated that a single AI search query will “likely cost 10 times more than a standard keyword search.” Morgan Stanley (2022) estimated that Google’s 3.3 trillion search queries cost roughly a fifth of a cent (USD) each.
What does AI mean for energy demand?
The International Energy Agency estimates that electricity consumption from data centres, AI, and cryptocurrencies could double by 2026, reaching roughly the same level as the electricity consumption of Japan. This is likely to be due to a massive increase in investment in data centres, with existing data centres being insufficient to meet demand and a massive increase in AI use.
New Zealand, in the short term, is unlikely to see an uptick in electrical demand from data centres. This is largely due to one factor: none of the major AI companies have a data centre in New Zealand. Amazon and Microsoft, while planning to build in New Zealand, are either still at the planning stage or have quietly put the projects on hold.
What is AI’s potential for providing and enabling sustainable and renewable energy?
It has been shown that AI is a heavy user of electricity; however, the use of AI technology can be adopted to improve the efficiency of the electricity network and prevent grid failures. Real-time AI data analysis could be used to more accurately predict the demand for electricity within the grid – reducing energy waste and increasing the stability of the grid. Modern electrical grid management is a constant balancing act, striving to provide enough supply to meet demand. If too little energy is supplied, blackouts occur; if too much is provided, energy is wasted.
Additionally, the International Energy Agency predicts AI will enable the effective integration of renewable energy into the grid. If this prediction is true and AI is integrated into the electrical grid, New Zealand could see improvements in energy supply, efficiency, and sustainability, improving outcomes for consumers and electricity suppliers alike. One reason New Zealand may be especially set to benefit is our goal to achieve 100 percent renewable energy by 2035 and net zero emissions by 2050. As of 2024, 42.8 percent of New Zealand’s total primary energy supply came from renewable sources.
Government and industry could work together to facilitate the adoption of AI technology in the energy sector
The successful integration of AI into New Zealand's energy sector demands collaborative action between government and industry. By developing clear regulatory frameworks, investing in infrastructure, and supporting research, both sectors can unlock AI's potential for grid management and renewable energy integration. While challenges around data privacy and technological uncertainty remain, a coordinated approach could position New Zealand as an innovative leader in sustainable energy systems. The key lies in balancing cautious implementation with strategic investment, ensuring that AI becomes a tool for enhancing energy efficiency and supporting our national renewable energy goals.