AI Has An Electricity Problem | Ep270: Arshad Mansoor
Will AI break the grid? As hyperscalers race to build ever-larger data centres, electricity demand is soaring. The question is no longer whether power demand will grow, but by how much, how quickly, and whether the grid can deliver.
This week, Michael Liebreich is joined by Arshad Mansoor, President & CEO of the Electric Power Research Institute (EPRI), one of the world's leading energy research organisations. Together, they tackle one of the biggest questions facing the global energy system: how much electricity will AI really need, and can our power systems keep up?
They debate the scale of AI-driven demand growth, whether the extraordinary investment by hyperscalers is justified, and how greater flexibility could allow data centres to connect faster without overwhelming the grid. They also discuss the growing public backlash against AI infrastructure, why winning local support may become the industry's biggest challenge, and whether nuclear power is truly the "get out of jail free card" many believe it to be or simply one part of a much bigger solution.
Topics included in this episode:
The race to power AI data centres
Hyperscalers' trillion-dollar AI investment
Can electricity grids keep pace with AI's rapid growth?
Gigawatt demand forecasts
Making data centres more flexible
Public acceptance and the growing backlash against AI
Nuclear power's role in the AI era
What governments, utilities and investors should do next
Leadership Circle:
Cleaning Up is proud to be supported by its Leadership Circle. The members are Actis, Alcazar Energy, Arup, Copenhagen Infrastructure Partners, Cygnum Capital, Davidson Kempner, EcoPragma Capital, EDP, Eurelectric, the Gilardini Foundation, KKR, Mitsubishi Heavy Industries, National Grid, Octopus Energy, Quadrature Climate Foundation, Schneider Electric, SDCL and Wärtsilä. For more information about the Leadership Circle, visit cleaningup.live
Links:
Arshad Mansoor’s bio: https://www.linkedin.com/in/arshad-mansoor-a2090b165/
EPRI, Electric Power Research Institute: https://www.epri.com/
Jigar Shah on Cleaning Up: https://www.youtube.com/watch?v=PCOaF-qQ_TU
Varun Sivaram & Steve Smith on Cleaning Up: https://www.youtube.com/watch?v=4kSrgRZUCwE
Rob Dunn on Cleaning Up: https://www.youtube.com/watch?v=juAyLAUmU3w&t=2s
Subscribe to our newsletter: https://cleaninguppod.substack.com/
Papers referenced:
DC Flex Mosaic: https://dcflex.epri.com/flex-mosaic/open-letter
Win Win Watts: https://winwin.epri.com/en/introduction.html
Make Electricity Cheap Again: https://mliebreich.substack.com/p/make-electricity-cheap-again
Acronyms:
GPU: Graphics Processing Unit
IPO: Initial Public Offering
NESO: National Energy Systems Operator
HVO: Hydrogenated Vegetable Oil
SMR: Small Modular Reactor
Chapters:
00:00 Intro
02:29 About EPRI
06:08 AI Power Demand
10:45 Gigawatt Debate
17:19 Hyperscaler Risk
26:15 Compute In Households
31:24 24/7 Data Power
38:28 Public Acceptance
41:39 Fear Of AI
49:39 A Nuclear Solution?
58:38 24/7 Reliable Power
01:08:00 Final Thoughts
This week, Michael Liebreich is joined by Arshad Mansoor, President & CEO of the Electric Power Research Institute (EPRI), one of the world's leading energy research organisations. Together, they tackle one of the biggest questions facing the global energy system: how much electricity will AI really need, and can our power systems keep up?
They debate the scale of AI-driven demand growth, whether the extraordinary investment by hyperscalers is justified, and how greater flexibility could allow data centres to connect faster without overwhelming the grid. They also discuss the growing public backlash against AI infrastructure, why winning local support may become the industry's biggest challenge, and whether nuclear power is truly the "get out of jail free card" many believe it to be or simply one part of a much bigger solution.
Topics included in this episode:
The race to power AI data centres
Hyperscalers' trillion-dollar AI investment
Can electricity grids keep pace with AI's rapid growth?
Gigawatt demand forecasts
Making data centres more flexible
Public acceptance and the growing backlash against AI
Nuclear power's role in the AI era
What governments, utilities and investors should do next
Leadership Circle:
Cleaning Up is proud to be supported by its Leadership Circle. The members are Actis, Alcazar Energy, Arup, Copenhagen Infrastructure Partners, Cygnum Capital, Davidson Kempner, EcoPragma Capital, EDP, Eurelectric, the Gilardini Foundation, KKR, Mitsubishi Heavy Industries, National Grid, Octopus Energy, Quadrature Climate Foundation, Schneider Electric, SDCL and Wärtsilä. For more information about the Leadership Circle, visit cleaningup.live
Links:
Arshad Mansoor’s bio: https://www.linkedin.com/in/arshad-mansoor-a2090b165/
EPRI, Electric Power Research Institute: https://www.epri.com/
Jigar Shah on Cleaning Up: https://www.youtube.com/watch?v=PCOaF-qQ_TU
Varun Sivaram & Steve Smith on Cleaning Up: https://www.youtube.com/watch?v=4kSrgRZUCwE
Rob Dunn on Cleaning Up: https://www.youtube.com/watch?v=juAyLAUmU3w&t=2s
Subscribe to our newsletter: https://cleaninguppod.substack.com/
Papers referenced:
DC Flex Mosaic: https://dcflex.epri.com/flex-mosaic/open-letter
Win Win Watts: https://winwin.epri.com/en/introduction.html
Make Electricity Cheap Again: https://mliebreich.substack.com/p/make-electricity-cheap-again
Acronyms:
GPU: Graphics Processing Unit
IPO: Initial Public Offering
NESO: National Energy Systems Operator
HVO: Hydrogenated Vegetable Oil
SMR: Small Modular Reactor
Chapters:
00:00 Intro
02:29 About EPRI
06:08 AI Power Demand
10:45 Gigawatt Debate
17:19 Hyperscaler Risk
26:15 Compute In Households
31:24 24/7 Data Power
38:28 Public Acceptance
41:39 Fear Of AI
49:39 A Nuclear Solution?
58:38 24/7 Reliable Power
01:08:00 Final Thoughts
Loading content...
Read the full transcript
Matterfact has every podcast on Apple Podcasts and Spotify: 700K+ channels, 120M+ episodes.
The candid version is on a podcast,
not the earnings call.
Matterfact turns the conversations your coverage is already having into a searchable, citable record — across every episode, not just this one.