Hyperscalers will increasingly get their power off-grid, says Morgan Stanley's Stephen Byrd

Watch on YouTube ↗  |  March 09, 2026 at 22:22  |  6:25  |  CNBC

Summary

  • AI data centers are facing intense political and public backlash due to their massive power consumption and the fear that they will drive up local electricity rates.
  • To ensure security of supply and avoid grid congestion, hyperscalers are pivoting to off-grid solutions, planning to build proprietary power plants and microgrids.
  • The profit margins on AI intelligence are so high that tech giants are entirely price-insensitive to energy costs; they can easily absorb higher natural gas prices to guarantee uptime.
  • The US suffers from severe power generation and grid capacity shortages compared to China, making infrastructure "debottlenecking" the most critical investment theme in the AI supply chain.
Trade Ideas
Stephen Byrd Global Head of Thematic and Sustainability Research, Morgan Stanley 5:40
"We need renewables, but we also need a lot of natural gas... the value of the intelligence that we're essentially producing from these data centers is so incredibly high that we can absorb higher gas prices." AI data centers require massive, uninterrupted baseload power that intermittent renewables alone cannot provide. Because hyperscalers are highly profitable and price-insensitive to energy costs, natural gas demand will see a structural increase driven by off-grid data center power generation. These buyers will pay whatever it takes to keep the GPUs running, supporting higher long-term prices for the commodity. LONG natural gas as it becomes a critical, price-inelastic fuel for the AI revolution. Faster-than-expected deployment of nuclear energy (Small Modular Reactors) or significant breakthroughs in utility-scale battery storage that reduce the need for gas-fired baseload power.
Stephen Byrd Global Head of Thematic and Sustainability Research, Morgan Stanley 6:10
"Anybody in the infrastructure space, if you have a power bottleneck, a bottleneck, equipment bottleneck, the value of debottlenecking is going to go up as the value of AI goes up... the data center community will just take control of their own destiny, build their own mini grids that are completely separated." The US grid is constrained and permitting for new utility projects is notoriously slow. To ensure power security and avoid political backlash, hyperscalers are bypassing traditional utilities to build their own private power plants. Companies that manufacture electrical equipment (ETN), construct power infrastructure (PWR), and provide decentralized power generation like turbines and fuel cells (GEV, BE) will capture the billions in CapEx spent to build these off-grid ecosystems. LONG electrical equipment and power infrastructure providers, as they are the direct beneficiaries of hyperscalers taking control of their own power destiny. Regulatory intervention preventing the construction of private microgrids, or a macro slowdown causing hyperscalers to slash their AI infrastructure CapEx.
Up Next

This CNBC video, published March 09, 2026, features Stephen Byrd discussing UNG, ETN, PWR, BE, GEV. 2 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Stephen Byrd  · Tickers: UNG, ETN, PWR, BE, GEV