Podcast Summary: Catalyst with Shayle Kann
Episode: “AI scaling pathways: on grid, on edge, off grid, off planet”
Date: March 12, 2026
Host: Shayle Kann (Latitude Media)
Guest: Jake Elder (SVP, Research & Innovation, Energy Impact Partners)

Episode Overview

This episode explores the various pathways for scaling AI data centers to meet surging compute demands, with a view toward energy and decarbonization constraints. Shayle Kann and Jake Elder break down four main configurations: hyperscale grid-connected data centers, edge (smaller, distributed) centers, off-grid (standalone power), and orbital (space-based) centers. The discussion navigates technical, social, regulatory, and economic barriers as well as theoretical scalability and future potential.

Key Discussion Points & Insights

1. Framing the Challenge (02:21–05:40)

• Unprecedented Demand: As AI and compute demand rise, the core challenge lies not just in compute but in delivering enough energy to support it.
• Shayle sets the premise: "Let’s assume compute demand continues to scale … the demand for watts to deliver that compute." (04:21)
• Main focus: Comparing frameworks and constraints across the four main pathways.

2. Hyperscale, Grid-Connected Data Centers (05:40–11:17)

Main Constraints

• Transmission Speed: Connecting large data centers to new power supply can take 5–7 years, with interregional transmission timelines bordering on “infinite years.” – (Shayle Kann, 08:04)
• Power Quality: Engineering challenge, but manageable compared to transmission.
• Social License: “Blanket bans on new data center developments… because of community pushback.” (Jake Elder, 06:24)
• Confusion Over Hybridization: Many ‘off-grid’ projects are actually grid-connected with backup, not truly islanded.

Notable Quotes

“The timeline to get a new transmission line built … is essentially infinite years. In the United States, at least in recent history, we just aren’t doing it.”
— Shayle Kann (08:00)

3. Edge (Distributed, Smaller) Data Centers (11:18–17:47)

Definitions and Categories

• Latency as a Red Herring: While proximity is supposed to reduce latency, most enterprise and AI uses don’t demand sub-millisecond delays (autonomous vehicles do most compute onboard).
• Categories:
  • On-Device: True edge (e.g., phones, vehicles).
  • Mini-Hyperscale: Sub-50MW, smaller form factors but similar in function.
  • Micro: Hundreds of kW at sites like substations or commercial buildings.

Main Challenges

• Economics: Hard to beat the efficiency and cost of massive, centralized centers.
• Speed: Theoretically, edge can allow for faster roll-out, but orchestrating hundreds of small sites adds complexity.

Notable Quotes

“Latency is a bit of a red herring … not for zero applications, but for very few does it seem that you need such low latency …”
— Shayle Kann (14:12)

“If you end up in a cost game … you’re subscale relative to bigger sites.”
— Jake Elder (13:37)

4. Off-Grid Data Centers (17:47–28:47)

Advantages

• Freedom from Transmission Bottlenecks: “If you remove the grid as a constraint, we have plenty of land available.” (Shayle Kann, 17:47)
• Potential for Cost-Competitive Renewables + Storage: Foundational studies show high solar+storage viability in the American Southwest.

Constraints

• Reliability Engineering Is Hard: The grid is a “giant shock absorber”—going fully off-grid means building that yourself.
• Current Reliability Gaps: Early off-grid projects struggling to achieve even 90% uptime.
• Supply Chains Remain a Limiting Factor: Even off-grid, you still need turbines, transformers, switchgear—“the supply chain problem.”
• Location Still Matters: Labor, water, workforce factors still constrain truly remote development.

Notable Quotes

“If you go off grid … you have to build the whole shock absorber yourself … and that’s not just expensive, it’s really complicated.”
— Jake Elder (19:34)

“It should be solvable. It is a real engineering challenge.”
— Shayle Kann (21:06)

5. Orbital (Space-Based) Data Centers (28:47–39:03)

Theoretical Appeal

• “Scalability thing”—in theory, infinite compute possible off-Earth—no environmental or social bottleneck.

Overhyped Proposals

• Elon Musk claims orbital centers could be “the cheapest” within years—Kann and Elder strongly disagree.
• Heat Dissipation Challenge: Space-based computing accumulates heat; a single gigawatt data center would require “a radiator the size of a small town.” —Jake Elder, (31:52)

Practical Constraints

• Space Debris: A 4-sq-km asset could be hit by debris “every hour”—ongoing risk and maintenance nightmare.
• O&M Infeasibility: On-Earth, hardware is replaced constantly; in orbit, that’s infeasible barring major advances in robotics.
• Economic Disadvantage: Solar is more productive in space, but energy is only 5–15% of data center costs; chips/maintenance are bigger and get worse with orbital logistics.

Notable Quotes

“If you scale that up to a single floating thing that’s 4 sq kilometers … you can basically expect a piece of space debris hitting that data center every hour.”
— Jake Elder (32:29)

“If it breaks at least today, you’re kind of stuck with it broken … until then, it just adds economic drag on the overall project.”
— Jake Elder (33:19)

6. Comparing Off-Grid vs. Off-Planet (34:49–37:41)

• Key Limiting Factors:
  • Off-grid: Supply chain for turbines, transformers, etc.
  • Orbital: Launch capacity (mostly Starship)—“Starship has to launch a lot, like a lot a lot…” (34:49)
• Neither is “eminently more scalable” at global scale in the short term.

Notable Quotes

“If we’re going straight to space, I’m surprised that we’re not stopping at a waypoint along the way of doing a lot of off grid. I’m surprised that hasn’t happened.”
— Shayle Kann (37:24)

7. Competitive Market Shares: Ten-Year Outlook (39:03–42:54)

Jake Elder's Prediction:

• Hyperscale Grid Connected: 50–60%
• Off-Grid: 10–15%
• Edge: ~15%
• Off-Planet: 5–10%

Shayle generally agrees, but is even more bullish on off-grid relative to edge.

“If a listener wants to convince us of edge, I would welcome it. Jake and I both. But yeah, we’re struggling to find the … reason why [edge] ….”
— Shayle Kann (42:16)

Memorable Moments & Quotes (with Timestamps)

• Heat dissipation in space:
  “The whole International Space Station … rejects less than 100 kilowatts of heat in total. And they have a radiator the size of a soccer field.”
  — Jake Elder (31:00)
• Space debris as existential risk:
  “You can basically expect to have a piece of space debris hitting that data center every hour.”
  — Jake Elder (32:30)
• On Edge definitionally struggling:
  “It’s the most obvious and theoretically fastest way to deploy compute. … It should be the right answer, but I agree with you.”
  — Jake Elder (42:01)
• Call to listeners:
  “If a listener wants to convince us of edge, I would welcome it.”
  — Shayle Kann (42:16)

Conclusion

Shayle and Jake expertly situate the buzzing excitement over grid limitations, off-grid strategies, and “crazy” ideas like orbital computing in a sober engineering and economic context. Their consensus: evolution will remain slow, centralized, and earth-bound—until, perhaps, a leap by necessity or breakthrough hurdles are cleared. Off-grid is the next frontier before orbit. Edge, in theory a quick fix, remains unproven at the scales required.

For more detail

• See Latitude Media (latitudemedia.com) for links and show notes
• Event note: Live podcast at Transition AI conference, San Francisco, April 13

Produced by Latitude Media, supported by Prelude Ventures. Host: Shayle Kann, Guest: Jake Elder.