Conference Presentation, Panel
AI Infrastructure at Scale: Financing the Power-Constrained Grid | Global Conference 2026
Milken InstituteBrian Sullivan, Lucy Heintz, Douglas Kimmelman, Karl Kuchel, Rajit Nanda, Michael Zeltkevic, Doug Kimmel, Carl Cuscelli, Lucy Hines
- Infrastructure sector capital expenditure rates are projected to reach a trillion-dollar figure rapidly.
- U.S. power prices are anticipated to rise significantly, while U.S. data center utilization rates are expected to peak within the next year.
- Global electricity demand growth outside the U.S. is driven by industrialization and cloud migration rather than AI, which accounts for only 5% of demand growth in those markets.
- The AI infrastructure focus has shifted from speed to market to speed to power, prompting relocation to regions with reliable power at scale.
- The AI training phase is expected to decline sharply, shifting industry attention to the inference phase.
- AI is projected to account for approximately 30% of data center demand in five years, though physical AI applications are expected to be significantly more data-intensive by that time.
- Less than 10% of companies currently have AI in production despite 88% prioritizing it, with ROI for current spending potentially realized by 2030.
- Supporting current CapEx levels may require $1.5 trillion in AI-related revenues by 2030, while smaller non-hyperscale AI players may face consolidation in a few years.
- Data center reliability standards may shift from "five nines" (99.999%) to lower tiers such as "one nine" (99%) or "four eights" (99.888%).
- Saudi Arabia plans to increase subsea landing stations from 17 to 24 within 1.5 years to connect 5.9 billion people across Africa, Europe, and South Asia with 30-40 millisecond latency.
- Gas turbine costs have tripled, and availability is limited for five years.
- Battery technology costs are forecast to decrease by another 34%, with emerging iron oxidation technologies expected to provide 100 hours of storage.
- Broadband modem costs are anticipated to follow a steep downward trajectory similar to electric vehicles within two years.
- The AI stack comprises segments with 4-5 year cycles and a significant segment with a 30-40 year cycle.
- Regulated U.S. utilities may attempt to build natural gas plants with a 40-year outlook, though revenue contracts may only secure 7-15 years.
- Nuclear, solar, and batteries are expected to dominate the energy infrastructure landscape over the next 100 years.
- Current wind, solar, and battery costs in India are five cents, whereas nuclear fusion is projected to cost between 25 and 40 cents.
- Investors face risks with new battery technologies that currently have a 0% financing success rate due to a lack of operating history.
- Hyperscalers may not require the highest reliability standards, potentially lowering barriers to entry for new infrastructure.