Feb 13, 2025 · 45m · catalyst
The case for colocating data centers and generation
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of Catalyst, host Shayle Kann interviews Sheldon Kimber, CEO of Intersect Power, exploring how co-locating multi-gigawatt clean energy and storage assets directly with AI data centers solves severe grid interconnect bottlenecks and satisfies the surging, inflexible power demands of artificial intelligence.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Shayle holds 29.3% of the talking time here. How this is scored →
speaking balance: gold is Shayle, purple is the guest (3 minute bins)
Sheldon forcefully rejects the premise of flexible compute loads, calling power the tail of the dog and insisting nobody would idle $30B of silicon for energy savings.
Hardest push from Shayle ▶ 39:07 Historical cloud efficiency counterargumentShayle pushes back against uninterrupted load growth assumptions by pointing to two decades of data center power demand remaining flat despite massive cloud computing expansion.
Biggest teaching moment ▶ 28:00 CapEx breakdown of modern gigawatt data centersSheldon educates the audience on the economic reality of data centers, contrasting $5B of generation assets against $30B in compute silicon to prove flexibility is uneconomic.
Shayle holds their own ▶ 39:20 Challenging Jevons paradox certaintyShayle demonstrates deep market knowledge by raising the historical decoupling of compute growth from energy consumption, forcing Sheldon to clarify how autonomous AI inference differs from past web traffic.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Shayle as informed peer | Guest teaching | Guest disagreement | Shayle pushing back | Why |
|---|---|---|---|---|---|---|
| Latitude Media Newsletter Promotion and Pre-Roll Announcement | 0 | 0 | 0 | 0 | Introductory teasers, newsletter promotion, and sponsor announcements for Bloom Energy, Engie, and EnergyHub. No interactive host-guest dialogue occurs. | |
| Introduction to the Data Center Power Bottleneck and Intersect Deal | 5 | 1 | 1 | 1 | Shayle opens with a structured overview of the data center bottleneck, contrasting excess capacity hunting, nuclear additions, and on-site generation before introducing Intersect's deal with Google and TPG. | |
| Macro Dynamics, Domestic Policies, and Supply Chain Tariff Pressures | 5 | 3 | 2 | 2 | Sheldon outlines the supply-side Biden manufacturing push against the AI demand boom, while Shayle adds nuance regarding complex pre-existing tariff structures beyond recent administration changes. | |
| PPA Market Evolution, Off-Taker Risk, and Pricing Divergence | 6 | 4 | 3 | 3 | Shayle and Sheldon dig into PPA tenor shifts and merchant tail risk. Sheldon corrects the perception around tenor trends, clarifying that developers moved away from 15-year terms due to uncompensated tail risk. | |
| Mid-Roll Sponsor Break: Fuel Cells, Energy Advisory, and Virtual Plants | 6 | 5 | 3 | 3 | After mid-roll sponsor reads, Shayle asks what fundamentally drives co-location. Sheldon lays out the triad of speed, scale, and the broken grid, explaining why load and generation interconnection rules differ. | |
| Grid Interconnection Architecture and Hybrid Baseload Generation | 5 | 5 | 4 | 2 | Sheldon dismisses the traditional holy war between gas and renewables, arguing that a hybrid 80% renewable plus 20% reciprocating engine setup beats expensive new combined cycle gas turbines on cost and speed. | |
| Capital Economics and the Impossibility of Flexible Compute Load | 6 | 6 | 5 | 3 | Shayle inquires whether flexible load has any operational traction. Sheldon forcefully dismantles the idea by breaking down capital expenditure: thirty billion dollars of chips will never be curtailed for cheap electricity. | |
| DeepSeek Efficiency Hype, Jevons' Paradox, and Autonomous AI Load | 7 | 6 | 5 | 5 | The conversation debates DeepSeek and Jevons paradox. Shayle offers a historical counterexample where cloud efficiency flattened power consumption for two decades, prompting Sheldon to challenge him to a source-off and describe recursive machine queries. |