Let’s Talk Artificial Intelligence: Is it helping or hindering the energy transition?

On a recent visit to the US Silicon Valley, I discussed the fast improvements of AI-based solutions with many tech companies. I have no doubt that AI will transform how we work, operate assets, or make decisions. In other words, we get a fantastic number of new tools in our toolbox. However, we are still trying to understand where and how to use it best. So many screws are lying around that we are sometimes unsure if we have the right screwdriver in hand.

The potential of AI is huge – something which is also reflected in today’s media landscape. You cannot scroll far here on LinkedIn without coming across a post on the power of AI.? What I have not seen sufficiently addressed, however, is the power it needs to run AI. Because behind the technology lie data centers that operate 24/7, 365 days a year, and need a considerable amount of power to keep them running.

Take this as an example: If you generate a single image with generative AI, it consumes as much energy as charging your phone. As a result, it can be estimated that AI already uses as much energy as a country (2% of total global energy demand, such as Indonesia). By 2026, the IEA estimates AI to be equal to the amount of electricity used by the entire country of Japan (6% of total global electricity demand).

These projections are based on a lot of assumptions, and the actual demand might vary significantly. However, one thing is clear: we will need more power. The costs for this power will also be a variable that determines whether certain AI-based models offer an interesting business case or not. The added electricity demand AI requires could change our view of how infrastructure for the generation and the distribution/transmission part of electricity will develop over the next few years. Particularly if the ramp-up of the additional demand comes fast – which is never something that is easy in infrastructure such as electricity.

In addition, tech companies running data centers have very tight net-zero targets and, therefore, have a vast appetite for decarbonization efforts. Helping data centers become more energy efficient and decarbonize their energy demands is, therefore, both a rapidly growing need and an opportunity.?An opportunity that we at 西门子能源 are acutely aware of.

At present, it is a common habit for data centers to purchase carbon credits to offset their unavoidable emissions. This is a start, but not really tackling the problem head-on. In the case of a quick ramp-up of power needs, we will see a rising demand for natural gas-based solutions. However, we have to ensure that we stay focused on transitioning these to decarbonized solutions, whether by operating these on carbon-free fuels or by faster pushing renewables and storage solutions.

My guess is that the phone or laptop you are reading this from is ever so slightly warm. Multiply this by a 1000, and you might be able to imagine the heat generated and cooling required in a data center. Every single interaction you have with ChatGPT requires a large bottle of water for data center cooling. To put this into perspective: ChatGPT had around 200 million active users last week. Employing energy-efficient technologies to utilize data centers' waste heat using for example heat pumps, is therefore important to achieve higher energy efficiency and to assist companies in meeting their sustainability targets.

This brings me to the important aspect: how AI is actually boosting efficiency and accelerating innovation in the energy industry. One of the most impactful ways AI has already been improving the energy market is in predicting supply and demand. Developing a greater understanding of when power is available can be complicated for renewable technologies since the sun does not always shine, and the wind does not always blow. That is where machine learning comes into play: to match variable supply with rising and falling demand. This maximizes the financial value of renewables and allows them to be more easily integrated into the grid.

A further example comes from our own production facilities: At Siemens Energy we used Artificial Intelligence in the turbine and rotor design of the world’s most powerful gas turbine. We also use AI to automate the manual production process in order to improve production quality in our wind turbine blade factories.

Further down the line, AI-based models allow us to carry out remote diagnostics and predictive maintenance on customer infrastructure, which maximizes efficiency, optimizes monitoring, and keeps the flexibility for dispatchable power during fluctuations. And on a more human note, Artificial Intelligence assists us in analyzing safety reports collected across our company to identify incident patterns, in turn supporting us in our constant strive for workplace safety.

We should not be oblivious to the fact that AI’s hunger for electric power is an issue and needs more discussion. A world concerned with reducing carbon emissions cannot ignore the energy demand of multiple virtual countries as well. The modernization of grids and decarbonization of data centers is, therefore, a major piece of the puzzle if we want to see AI’s advancement – something worth advocating for. AI is already bringing enormous advancements to the energy industry and opening new avenues for efficiency and innovation. Whether we are talking about improving the quality of products, solutions, and services or accelerating decision-making and increasing productivity in our daily work, I think it is a powerful technology that will help us drive the energy transition forward and amplify human creativity … so, what is your guess? How much did AI support me in writing this article?