Our Global Head of Sustainability Research and U.S. Utilities Analyst discuss the rapidly growing power needs of the GenAI enablers and how to meet them.----- Transcript -----Stephen Byrd: Welcome to Thoughts on the Market. I'm Stephen Byrd, Morgan Stanley's Global Head of Sustainability Research.David Arcaro: And I'm Dave Arcaro, Head of the US Power and Utilities team.Stephen Byrd: And on this episode of the podcast, we'll discuss just what it would take to power the Gen AI revolution.It's Tuesday, April 16th at 10am in New York.Last summer, scientists used GenAI to find a new antibiotic for a nasty superbug. It took the AI system all of an hour and a half to analyze about 7,000 chemical compounds; something that human scientists would have toiled over for months, if not years. It's clear that GenAI can open up breathtaking possibilities, but you have to stop and think. What kind of compute power is needed for all of this?A few weeks ago, our colleague Emmet Kelly, who covers European Telecom, discussed the exponential growth of European data centers on this podcast. And today, Dave and I want to continue the conversation about this critical moment of powering the GenAI revolution.So, Dave, what is your current assessment of the global power demand from data centers?David Arcaro: Yeah, Stephen, we're expecting rapid growth in the power demand coming from data centers across the world. We're currently estimating data centers consume about one and a half per cent of global electricity today. We're expecting that to grow to almost four percent in 2027. And in the US, data centers represent roughly three percent of total electricity consumption now, and we expect that to escalate to eight per cent of the total US by 2027.And there will be even more dramatic impacts at the local and regional level. The data center landscape tends to be highly concentrated, and the next wave of GenAI data centers is likely to be much larger than the previous generation.So, the impact on specific regions will be magnified. To give an example, in Georgia, the utility there has previously forecasted just half a percent of annual growth in electricity use but is now calling for nine per cent of annual growth in electricity consumption, and that's largely driven by data centers.It's a dynamic that we haven't seen in decades in the utility space.Stephen Byrd: You know, what I find interesting about what you just said, Dave, is -- it is impressive to see growth go from one and a half to four per cent, but it's really these local dynamics where what we're seeing is just much more concentrated, and that's where we start to see the real issues with the infrastructure growing quickly enough.So, it's becoming obvious that the existing power grid infrastructure is not meeting the growth and capacity needs of data centers. And that's something that you refer to as the tortoise and the hare. How big of a mismatch are we exactly talking about here, Dave?David Arcaro: It's definitely a big mismatch. To your point before, the US electricity growth across the country has been flattish over the last 10 years.So, this is a step change in expectations now, from the impact from Gen AI going forward. And we're looking at over 100 per cent annual growth in the power consumed by data centers now in the US over the next four years. And for comparison, the US utility industry is growing at about 8 per cent a year.These data centers that are coming are huge. They can be 10 to 50 times as big as the last generation of data centers in terms of their power consumption. And this means it takes time to connect to the electric grid and get power. 12 to 18 months in the best case, three to five years plus in other locations, often because they might need to wait for the electric utility grid to catch up, waiting for grid upgrades and assessments and new power plants to get built.Stephen Byrd: Well, I think those delays are going to be fairly problematic for the fast-moving GenAI sec