As AI is in the headlines nearly every day now, I have been, like many others, exploring its real-life applications like exploring CCUS and AI power load.

Below is a recent short dialogue on this taken from numerous Q&A sessions I recently had.  Q&A unedited.

My question: What will AI’s electricity consumption grow to by 2030 based on current rate of growth of its current energy demand?

AI Smith Chat GPT ver 4.0: As an AI, I don't have real-time or future predictive capabilities to forecast the electricity consumption of AI by 2030. However, I can explain that the AI energy demand depends on several factors such as technological advancement, the prevalence of AI systems, and energy efficiency measures. Currently, the computational power used in the largest AI training runs has been doubling approximately every 3.4 months (according to a 2020 OpenAI report). However, optimization in hardware, software, and algorithms can reduce power consumption significantly over time.

Keep in mind that this is a complex matter affected by many variables, so precise forecasts would require detailed information based on extensive research. For the most accurate predictions, consulting with an expert or a relevant study in the field would be recommended.

How big a contribution will very low carbon electricity produced from carbon capture, utilization, & storage (CCUS) make to support the rapidly increasing AI electricity demand by 2030?

Currently, the potential contribution of very low carbon electricity produced from Carbon Capture Utilization Storage (CCUS) to support AI's increasing electricity demand by 2030 isn't entirely known as there are numerous variables in play. However, a few points can be made.

1. Need for Low Carbon Electricity: As the use of AI expands, it's becoming more essential for us to source its power demands from low carbon emissions sources. AI's complex calculations and processing are energy-intensive, substantially increasing electricity consumption. By 2030, if the AI sector continues to grow at its current pace, its electricity demand might be significant.

2. Role of CCUS: CCUS has the potential to significantly reduce CO2 emissions from fossil fuel usage. It captures CO2 emissions produced from the use of fossil fuels in electricity generation and industrial processes, preventing the CO2 from entering the atmosphere.

3. Potential Challenges: While CCUS technology can make a big difference, some challenges limit its contribution. For example, the process is currently quite expensive, and the technology is still

being developed - we need more advancements and investment to make it a viable worldwide solution. There is also the issue of dealing with the captured CO2 - storing it can present complications, including the need for suitable locations and potential risks related to leaks.

4. Future Scenarios: Some energy scenarios see a significant role for CCUS, particularly in heavy industry and electricity generation, and believe that it could supply a sizable proportion of the world's energy demand by 2030. Still, these scenarios depend on rapid technological advances and broad regulatory support.

In conclusion, while CCUS could theoretically support a substantial part of AI's electricity demand by 2030, it's unclear how exactly this will pan out, given the challenges involved. The actual contribution will depend on advancements in technology, funding, regulatory support, and the evolution of both AI's energy demands and the wider energy mix.

But something surfaced that is worth noting:  This particular Chat GPT app is using only the inputs during its training that ended in September of 2021, so it is missing the more recent inputs of money moving into CCUS, National policies that have been implemented supporting CCUS since then and new CCUS projects announced around the globe. This is all while AI growth and power demand is far faster than expected.

So when exploring AI, one might want to know when the particular AI app quit its training and future plans to updates its data bases, when stale data is being used and consider adding updates it can use in future analyses.

All the best,

Moore to come,

Mike