AI Needs More Than Chips | Presented by CME Group
Why It Matters
Energy constraints will dictate AI deployment speed, making power‑infrastructure investments critical for competitive advantage.
Key Takeaways
- •AI growth limited by rising electricity demand, not algorithms.
- •Data center power use could double by 2026, hitting 6‑12% US electricity.
- •Morgan Stanley predicts 49 GW power shortfall by 2028.
- •Companies invest in efficient chips, on‑site generation, and modular nuclear reactors.
- •Hyperscalers plan $1 trillion energy infrastructure spend in 2025‑26.
Summary
The CME Group presentation warned that the greatest obstacle to the AI boom is not algorithmic sophistication but the sheer amount of electricity required to power ever‑larger models.
Global data‑center consumption is expected to more than double by 2026, potentially accounting for 6‑12 % of U.S. electricity by 2030. Morgan Stanley forecasts a 49 GW capacity shortfall by 2028. At the same time, chip manufacturers are squeezing more performance per watt, while firms are adding on‑site solar, wind and small modular nuclear reactors to offset demand.
CME highlighted a quote: “The boom adapts. It does not bust,” underscoring that hyperscalers are already earmarking over $1 trillion for energy‑infrastructure projects in 2025‑26. These investments illustrate a shift from pure compute spending to parallel power‑supply development.
For investors and corporate strategists, the message is clear: AI growth will be tightly coupled with the rollout of new, low‑carbon power assets. Companies that secure reliable, affordable electricity will gain a competitive edge, while regions lagging in grid upgrades risk missing AI‑driven revenue streams.
Comments
Want to join the conversation?
Loading comments...