Podcast Summary: Inevitable (MCJ Podcast)

Episode: AI Data Centers That Help the Grid with Emerald AI

Date: February 10, 2026
Host: Cody Simms
Guest: Dr. Varun Sivaram, CEO & Founder, Emerald AI

Episode Overview

This episode dives into how Emerald AI is enabling AI data centers to become flexible, responsive participants within the power grid, turning them from “villains” stressing electric infrastructure to valuable allies enhancing grid stability and efficiency. Dr. Varun Sivaram explains how Emerald’s software platform—Conductor—lets data centers dial their electricity use up or down, optimizing for grid needs without sacrificing compute performance. The discussion investigates the unique challenges and opportunities posed by the explosive growth of AI workloads, the historical context, product mechanisms, practical demonstrations, and the broader implications for energy policy and the industry.

Key Discussion Points & Insights

1. Historical Context & The Unprecedented Growth of AI Power Demand

• AI as a Power Consumer
  • Dr. Sivaram draws an analogy between today’s surge in AI compute demand and the demand booms from air conditioning post-WWII in the 1950s/60s but highlights how AI data centers are far more concentrated and intensive.
  • Notable quote:

    “AI is likely to use more energy than everything else combined by orders of magnitude... a fundamentally different energy user than any other technology that's come before.”
    — Dr. Varun Sivaram [04:01]

• Unique AI Properties
  • AI workloads, especially training, can be geographically flexible—unlike historic loads such as air conditioning or electric vehicles.

2. Varun Sivaram’s Background & Origin of Emerald AI

• Dr. Sivaram’s transition from building large-scale energy infrastructure (Renew Power, Ørsted) to focusing on demand-side flexibility to unlock power for AI, seeing “just building our way out” as insufficient.
• The genesis of Emerald AI came from the realization that demand-side solutions are critical to both powering AI and keeping energy affordable.
• Notable quote:

  “I realized that just building our way out of this wouldn’t work...Emerald AI was born out of this shift.”
  — Dr. Varun Sivaram [07:19]

3. The Product: ‘Conductor’ and How It Works

• Primary Customer: Data center operators, but product involves both the data center and utility/system operator (e.g., for monitoring, dispatching, and optimization).
• Role: A software “brain” connecting the power grid and data centers, enabling both to be co-optimized.
  — “We are that software layer that unites these two multitrillion dollar networks; the multitrillion-dollar power grid and the multitrillion-dollar network of emerging AI data centers.”
  — Dr. Varun Sivaram [09:57]
• Analogy: Like traditional cloud schedulers, but focused on power availability and grid stress—not just latency and cost.

4. The Twin Problems Emerald Addresses

1. Speed to Power: How quickly data centers can be connected to the grid—often delayed by grid limitations.
2. Affordability: Preventing household power bills from rising due to new data center connections.

• The Solution: By making data centers “flexible loads,” they can be granted quicker interconnection and spread grid costs more efficiently, benefiting all users.

5. Three Pillars of Data Center Flexibility

• A. Temporal (Time) Flexibility: Shifting or pausing compute to relieve grid stress at peak times.

  • Phoenix Demonstration: During a peak event, Emerald’s system enabled a 25% power reduction for 3 hours without hurting essential workloads.
  • Notable quote:

    “The net effect was that we kept the jobs running according to the customer specifications... and we were able to prove it all to both sides, to the compute customer and to the grid.”
    — Dr. Varun Sivaram [22:13]

  • Flexibility options for customers: Emerald can recommend actions or directly orchestrate workloads—the level of “autonomy” is up to the customer.

• B. Spatial (Geographic) Flexibility: Moving workloads between different data center locations in response to grid conditions.

  • Example: Shifting inference loads from Virginia to Chicago during a local Virginia grid event.
  • Notable quote:

    “AI data centers... have this unique capability to move their economic activity around the country or the world on fiber optic networks at the speed of light. This is really rare and awesome.”
    — Dr. Varun Sivaram [24:45]

• C. Resource Flexibility: Coordinating with on-site energy assets (e.g., batteries, fuel cells), using local resources to buffer grid demand.

  • Notable quote:

    “If we have some batteries on site, we should take advantage of those... and also rely on the compute to provide the rest of the flex.”
    — Dr. Varun Sivaram [29:33]

  • True optimization requires integrating resource control, not just treating AI as a “black box.”

6. On Premises Power vs. Grid Integration

• Many new data centers consider operating off-grid (with their own on-site or even space-based power).
• Sivaram’s view: The US grid already has enough excess capacity—except for occasional peaks. Abandoning the grid would reinforce inefficiencies, raise long-term costs, and trigger major policy/spending issues.
  • Notable quote:

    “My exhortation to the field is let's solve this. We're better together... Let's take advantage of [the grid and fiber optic network] to strengthen our existing grid and not abandon it altogether.”
    — Dr. Varun Sivaram [36:48]

7. Policy and Future Outlook

• Total disconnection from the grid is unlikely to dominate; regulatory and public backlash will keep most data centers connected.
• Prediction: 90% of new AI compute will remain grid-connected over the coming decade.
• Emerald’s mission is to enable this integration, maximizing both AI growth and grid efficiency.

Notable Quotes & Memorable Moments

| Timestamp | Speaker | Quote | |-----------|---------|-------| | 04:01 | Dr. Varun Sivaram | “AI is likely to use more energy than everything else combined by orders of magnitude...” | | 07:19 | Dr. Varun Sivaram | “I realized that just building our way out of this wouldn’t work...Emerald AI was born out of this shift.” | | 09:57 | Dr. Varun Sivaram | “We are that software layer that unites these two multitrillion dollar networks; the multitrillion-dollar power grid and the multitrillion-dollar network of emerging AI data centers...” | | 22:13 | Dr. Varun Sivaram | “The net effect was that we kept the jobs running according to the customer specifications... and we were able to prove it all to both sides, to the compute customer and to the grid.” | | 24:45 | Dr. Varun Sivaram | “AI data centers... have this unique capability to move their economic activity around the country or the world on fiber optic networks at the speed of light. This is really rare and awesome.” | | 29:33 | Dr. Varun Sivaram | “If we have some batteries on site, we should take advantage of those batteries to provide some of the flex and also rely on the compute to provide the rest of the flex.” | | 36:48 | Dr. Varun Sivaram | “My exhortation to the field is let's solve this. We're better together... Let's take advantage of [the grid and fiber optic network] to strengthen our existing grid and not abandon it altogether.” |

Timestamps for Key Segments

• [01:36] — Emerald AI’s vision: Power-flexible AI "factories"
• [04:01] — Why AI’s energy surge is different from history
• [06:41] — Varun Sivaram’s background; shift from supply to flexible demand
• [09:47] — Customers and “Conductor” as the bridge layer
• [12:13] — Comparing traditional and power-aware scheduling
• [14:45] — Data center flexibility: Impact on power bills and grid connection
• [19:04] — The three flexibility levers (temporal, spatial, resource)
• [22:13] — Phoenix demo: 25% grid stress reduction
• [24:45] — Spatial/geographic shifting in practice
• [28:20] — Resource flexibility and on-site assets explained
• [31:26] — On-grid vs. off-grid data center models
• [34:34] — Space-based power generation; maximizing earth and space capacity
• [35:38] — Policy/political reality and future compute distribution
• [37:12] — Emerald is hiring; call for talent and closing

Concluding Call to Action

Dr. Sivaram closes the episode encouraging listeners with relevant skills—tech, AI, business, operations—to reach out to Emerald AI, underscoring their rapid growth and ambitious mission.

Overall Tone:
Visionary, candid, and optimistic with a pragmatic engineering focus. Dr. Sivaram is both technical and strategic, making complex grid and compute topics accessible while urging the industry to embrace flexibility and collective progress.