StarZero All shows Try the podcast clipper →

Apr 9, 2026 · 44m · catalyst

The rise of flexible data centers

Varun Sivaram · 24m spoken Shayle Kann · 12m spoken
0:00 / 0:00

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 and Emerald AI founder Dr. Varun Sivaram discuss how orchestrating AI compute workloads and behind-the-meter energy assets transforms data centers into flexible grid resources. They explore how computational flexibility resolves power interconnection bottlenecks, prevents ratepayer cost spikes, and enables sustainable digital infrastructure growth.

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 30.5% of the talking time here. How this is scored →

Shayle as informed peer 6.1 Guest teaching 4.0 Guest disagreement 1.6 Shayle pushing back 2.4
05100:0015:0030:003:53–8:22 · Shayle as informed peer 6/10 AI Growth Projections and Rising Electricity Affordability Concerns The conversation is highly collaborative as both host and guest are longtime colleagues and current investment partners. Shayle steers the discussion directly toward affordability as the primary driver of current grid tension, which Varun validates with recent industry load growth projections.8:22–13:43 · Shayle as informed peer 6/10 The Disconnect Between Compute Flexibility and Utility Service Tiers Varun outlines the disconnect between flexible compute tiers and rigid utility tariffs. Shayle pushes back by pointing out that traditional demand response already offers a version of non-firm service, prompting Varun to explain why traditional DR fails to provide the true prize: faster interconnection queues.13:44–20:39 · Shayle as informed peer 7/10 Redefining Economic Incentives: Interconnection Speed Versus Curtailment Costs Shayle highlights the bit-watt spread argument, asserting that curtailing compute purely for power bill savings makes little economic sense compared to interconnection acceleration. Varun politely pushes back by explaining how token prices asymptotically approach energy costs and how shifting along the power-performance curve preserves revenue.20:40–25:53 · Shayle as informed peer 6/10 Demonstrating Workload Flexibility: Field Pilots and Spatial Load Shifting Varun walks through experimental pilot data demonstrating spatial load shifting between Virginia and Chicago alongside workload throttling during grid spikes. Shayle probes on how workloads differentiate between mission-critical inference and flexible batch processes.25:57–31:24 · Shayle as informed peer 6/10 Sponsor Break: Fast Power Infrastructure and Grid-Edge Management Following the mid-roll ad read, Shayle analyzes the complex multi-stakeholder challenge spanning data center real estate owners, cloud hyperscalers, and end users. Varun explains Emerald's modular software architecture designed to interface across every layer of the operating stack.31:26–41:11 · Shayle as informed peer 8/10 Behind-the-Meter Generation, Storage, and the Data Center Dispatch Curve Shayle introduces an elegant economic framework framing data center flexibility as a micro dispatch curve where software flex is the zero-marginal-cost tier ahead of behind-the-meter generation. Varun enthusiastically adopts the concept while expanding it into dynamic token-watt arbitrage.41:12–43:38 · Shayle as informed peer 4/10 The Four Strategic Benefits and the Path to Commercial Scale Varun delivers his closing synthesis outlining four macro benefits of flexible data centers and previews Emerald's upcoming 100 MW pilot facility. Shayle concurs and wraps up the episode on an agreeable note.3:53–8:22 · Guest teaching 3/10 AI Growth Projections and Rising Electricity Affordability Concerns The conversation is highly collaborative as both host and guest are longtime colleagues and current investment partners. Shayle steers the discussion directly toward affordability as the primary driver of current grid tension, which Varun validates with recent industry load growth projections.8:22–13:43 · Guest teaching 4/10 The Disconnect Between Compute Flexibility and Utility Service Tiers Varun outlines the disconnect between flexible compute tiers and rigid utility tariffs. Shayle pushes back by pointing out that traditional demand response already offers a version of non-firm service, prompting Varun to explain why traditional DR fails to provide the true prize: faster interconnection queues.13:44–20:39 · Guest teaching 5/10 Redefining Economic Incentives: Interconnection Speed Versus Curtailment Costs Shayle highlights the bit-watt spread argument, asserting that curtailing compute purely for power bill savings makes little economic sense compared to interconnection acceleration. Varun politely pushes back by explaining how token prices asymptotically approach energy costs and how shifting along the power-performance curve preserves revenue.20:40–25:53 · Guest teaching 5/10 Demonstrating Workload Flexibility: Field Pilots and Spatial Load Shifting Varun walks through experimental pilot data demonstrating spatial load shifting between Virginia and Chicago alongside workload throttling during grid spikes. Shayle probes on how workloads differentiate between mission-critical inference and flexible batch processes.25:57–31:24 · Guest teaching 4/10 Sponsor Break: Fast Power Infrastructure and Grid-Edge Management Following the mid-roll ad read, Shayle analyzes the complex multi-stakeholder challenge spanning data center real estate owners, cloud hyperscalers, and end users. Varun explains Emerald's modular software architecture designed to interface across every layer of the operating stack.31:26–41:11 · Guest teaching 4/10 Behind-the-Meter Generation, Storage, and the Data Center Dispatch Curve Shayle introduces an elegant economic framework framing data center flexibility as a micro dispatch curve where software flex is the zero-marginal-cost tier ahead of behind-the-meter generation. Varun enthusiastically adopts the concept while expanding it into dynamic token-watt arbitrage.41:12–43:38 · Guest teaching 3/10 The Four Strategic Benefits and the Path to Commercial Scale Varun delivers his closing synthesis outlining four macro benefits of flexible data centers and previews Emerald's upcoming 100 MW pilot facility. Shayle concurs and wraps up the episode on an agreeable note.3:53–8:22 · Guest disagreement 1/10 AI Growth Projections and Rising Electricity Affordability Concerns The conversation is highly collaborative as both host and guest are longtime colleagues and current investment partners. Shayle steers the discussion directly toward affordability as the primary driver of current grid tension, which Varun validates with recent industry load growth projections.8:22–13:43 · Guest disagreement 2/10 The Disconnect Between Compute Flexibility and Utility Service Tiers Varun outlines the disconnect between flexible compute tiers and rigid utility tariffs. Shayle pushes back by pointing out that traditional demand response already offers a version of non-firm service, prompting Varun to explain why traditional DR fails to provide the true prize: faster interconnection queues.13:44–20:39 · Guest disagreement 3/10 Redefining Economic Incentives: Interconnection Speed Versus Curtailment Costs Shayle highlights the bit-watt spread argument, asserting that curtailing compute purely for power bill savings makes little economic sense compared to interconnection acceleration. Varun politely pushes back by explaining how token prices asymptotically approach energy costs and how shifting along the power-performance curve preserves revenue.20:40–25:53 · Guest disagreement 1/10 Demonstrating Workload Flexibility: Field Pilots and Spatial Load Shifting Varun walks through experimental pilot data demonstrating spatial load shifting between Virginia and Chicago alongside workload throttling during grid spikes. Shayle probes on how workloads differentiate between mission-critical inference and flexible batch processes.25:57–31:24 · Guest disagreement 1/10 Sponsor Break: Fast Power Infrastructure and Grid-Edge Management Following the mid-roll ad read, Shayle analyzes the complex multi-stakeholder challenge spanning data center real estate owners, cloud hyperscalers, and end users. Varun explains Emerald's modular software architecture designed to interface across every layer of the operating stack.31:26–41:11 · Guest disagreement 2/10 Behind-the-Meter Generation, Storage, and the Data Center Dispatch Curve Shayle introduces an elegant economic framework framing data center flexibility as a micro dispatch curve where software flex is the zero-marginal-cost tier ahead of behind-the-meter generation. Varun enthusiastically adopts the concept while expanding it into dynamic token-watt arbitrage.41:12–43:38 · Guest disagreement 1/10 The Four Strategic Benefits and the Path to Commercial Scale Varun delivers his closing synthesis outlining four macro benefits of flexible data centers and previews Emerald's upcoming 100 MW pilot facility. Shayle concurs and wraps up the episode on an agreeable note.3:53–8:22 · Shayle pushing back 2/10 AI Growth Projections and Rising Electricity Affordability Concerns The conversation is highly collaborative as both host and guest are longtime colleagues and current investment partners. Shayle steers the discussion directly toward affordability as the primary driver of current grid tension, which Varun validates with recent industry load growth projections.8:22–13:43 · Shayle pushing back 3/10 The Disconnect Between Compute Flexibility and Utility Service Tiers Varun outlines the disconnect between flexible compute tiers and rigid utility tariffs. Shayle pushes back by pointing out that traditional demand response already offers a version of non-firm service, prompting Varun to explain why traditional DR fails to provide the true prize: faster interconnection queues.13:44–20:39 · Shayle pushing back 4/10 Redefining Economic Incentives: Interconnection Speed Versus Curtailment Costs Shayle highlights the bit-watt spread argument, asserting that curtailing compute purely for power bill savings makes little economic sense compared to interconnection acceleration. Varun politely pushes back by explaining how token prices asymptotically approach energy costs and how shifting along the power-performance curve preserves revenue.20:40–25:53 · Shayle pushing back 2/10 Demonstrating Workload Flexibility: Field Pilots and Spatial Load Shifting Varun walks through experimental pilot data demonstrating spatial load shifting between Virginia and Chicago alongside workload throttling during grid spikes. Shayle probes on how workloads differentiate between mission-critical inference and flexible batch processes.25:57–31:24 · Shayle pushing back 2/10 Sponsor Break: Fast Power Infrastructure and Grid-Edge Management Following the mid-roll ad read, Shayle analyzes the complex multi-stakeholder challenge spanning data center real estate owners, cloud hyperscalers, and end users. Varun explains Emerald's modular software architecture designed to interface across every layer of the operating stack.31:26–41:11 · Shayle pushing back 3/10 Behind-the-Meter Generation, Storage, and the Data Center Dispatch Curve Shayle introduces an elegant economic framework framing data center flexibility as a micro dispatch curve where software flex is the zero-marginal-cost tier ahead of behind-the-meter generation. Varun enthusiastically adopts the concept while expanding it into dynamic token-watt arbitrage.41:12–43:38 · Shayle pushing back 1/10 The Four Strategic Benefits and the Path to Commercial Scale Varun delivers his closing synthesis outlining four macro benefits of flexible data centers and previews Emerald's upcoming 100 MW pilot facility. Shayle concurs and wraps up the episode on an agreeable note.

speaking balance: gold is Shayle, purple is the guest (3 minute bins)

0:00 · Shayle 71.2% · guest 28.8%0:00 · Shayle 71.2% · guest 28.8%3:00 · Shayle 33.9% · guest 66.1%3:00 · Shayle 33.9% · guest 66.1%6:00 · Shayle 34.5% · guest 65.5%6:00 · Shayle 34.5% · guest 65.5%9:00 · Shayle 8.7% · guest 91.3%9:00 · Shayle 8.7% · guest 91.3%12:00 · Shayle 50% · guest 50%12:00 · Shayle 50% · guest 50%15:00 · Shayle 22.8% · guest 77.2%15:00 · Shayle 22.8% · guest 77.2%18:00 · Shayle 37.8% · guest 62.2%18:00 · Shayle 37.8% · guest 62.2%21:00 · Shayle 7.3% · guest 92.7%21:00 · Shayle 7.3% · guest 92.7%24:00 · Shayle 0% · guest 100%24:00 · Shayle 0% · guest 100%27:00 · Shayle 29.4% · guest 70.6%27:00 · Shayle 29.4% · guest 70.6%30:00 · Shayle 43.7% · guest 56.3%30:00 · Shayle 43.7% · guest 56.3%33:00 · Shayle 0% · guest 100%33:00 · Shayle 0% · guest 100%36:00 · Shayle 72.9% · guest 27.1%36:00 · Shayle 72.9% · guest 27.1%39:00 · Shayle 24.8% · guest 75.2%39:00 · Shayle 24.8% · guest 75.2%42:00 · Shayle 19.7% · guest 80.3%42:00 · Shayle 19.7% · guest 80.3%
Sharpest disagreement ▶ 18:27 Varun challenges premise on compute curtailment economics

Varun directly splits with Shayle's assertion that curtailing compute purely for power bill OPEX savings is economically irrational, countering that dropping token margins will soon force operational power efficiency.

 Hardest push from Shayle ▶ 11:44 Shayle challenges claim regarding lack of utility flexibility tiers

Shayle interrupts Varun's framing by pointing out that demand response programs have historically provided interruptible electric service tiers to large industrial loads for decades.

 Biggest teaching moment ▶ 12:14 Varun clarifies why traditional DR fails data centers

Varun educates the listener and host on why traditional DR bill credits are inadequate for gigawatt-scale data centers, which require structural non-firm interconnection tariffs that bypass queue delays.

 Shayle holds their own ▶ 36:05 Shayle constructs the data center micro dispatch framework

Shayle demonstrates deep power market expertise by modeling on-site flexibility as a merit-order dispatch curve, establishing zero-marginal-cost workload throttling at the base ahead of physical fuel and battery dispatch.
the scores for every segment, with the reasoning behind each
ChapterTopicShayle as informed peerGuest teachingGuest disagreementShayle pushing backWhy
AI Growth Projections and Rising Electricity Affordability Concerns 6312 The conversation is highly collaborative as both host and guest are longtime colleagues and current investment partners. Shayle steers the discussion directly toward affordability as the primary driver of current grid tension, which Varun validates with recent industry load growth projections.
The Disconnect Between Compute Flexibility and Utility Service Tiers 6423 Varun outlines the disconnect between flexible compute tiers and rigid utility tariffs. Shayle pushes back by pointing out that traditional demand response already offers a version of non-firm service, prompting Varun to explain why traditional DR fails to provide the true prize: faster interconnection queues.
Redefining Economic Incentives: Interconnection Speed Versus Curtailment Costs 7534 Shayle highlights the bit-watt spread argument, asserting that curtailing compute purely for power bill savings makes little economic sense compared to interconnection acceleration. Varun politely pushes back by explaining how token prices asymptotically approach energy costs and how shifting along the power-performance curve preserves revenue.
Demonstrating Workload Flexibility: Field Pilots and Spatial Load Shifting 6512 Varun walks through experimental pilot data demonstrating spatial load shifting between Virginia and Chicago alongside workload throttling during grid spikes. Shayle probes on how workloads differentiate between mission-critical inference and flexible batch processes.
Sponsor Break: Fast Power Infrastructure and Grid-Edge Management 6412 Following the mid-roll ad read, Shayle analyzes the complex multi-stakeholder challenge spanning data center real estate owners, cloud hyperscalers, and end users. Varun explains Emerald's modular software architecture designed to interface across every layer of the operating stack.
Behind-the-Meter Generation, Storage, and the Data Center Dispatch Curve 8423 Shayle introduces an elegant economic framework framing data center flexibility as a micro dispatch curve where software flex is the zero-marginal-cost tier ahead of behind-the-meter generation. Varun enthusiastically adopts the concept while expanding it into dynamic token-watt arbitrage.
The Four Strategic Benefits and the Path to Commercial Scale 4311 Varun delivers his closing synthesis outlining four macro benefits of flexible data centers and previews Emerald's upcoming 100 MW pilot facility. Shayle concurs and wraps up the episode on an agreeable note.

Statements from this episode (10)

Assertion Supported
Data centers drive 94% of PJM load growth, 17% of US power
“The latest numbers show that in January of this year, NERC forecasts a summer peak increase of 224 gigawatts, almost all of which will come from data centers. Data centers now account for 94% of PJM's projected peak load growth, and by 20:30 EPRI forecast that…”
Varun Sivaram Apr 9, 2026 ▶ 5:47
Assertion Not checkable as stated
Power rates historically rose slower in regions with faster data center growth
“And, you know, some data shows that in areas where data centers grew more quickly, rates actually increased more slowly.”
Varun Sivaram Apr 9, 2026 ▶ 7:42
Assertion Supported
Sivaram: Over 100GW of stranded grid capacity could power flexible data centers
“The grid is utilized at roughly 50% or less during most of the year, and therefore, we have all of this hundred plus gigawatts of stranded power capacity that could power AI factories connecting to the grid today if during those rare peak moments AI factories …”
Varun Sivaram Apr 9, 2026 ▶ 9:48
Assertion Supported
Google reached 1 GW of flexible data center capacity across five utilities
“Google now, ah, has reached a gigawatt of contracted flexible capacity across, I think, five different utility territories, at least some of whom, ah, are willing to provide some of these benefits.”
Varun Sivaram Apr 9, 2026 ▶ 15:14
Prediction Not checkable as stated
Data centers will accept power curtailment tiers for faster grid interconnection
“I believe that if electric power utilities and grid operators offered these, a range of different service tiers, just like on the compute side, the cloud operators offer a range of different service tiers to their end AI customers, innovation would solve this …”
Varun Sivaram Apr 9, 2026 ▶ 16:44
Prediction Not checkable as stated
AI token costs will asymptotically approach the fundamental cost of power
“I think the answer will increasingly become yes in the future as the cost of inference, the cost of token generation asymptotically approaches the cost of power, which is the only real operational expense input into the Into the cost of intelligence generation…”
Varun Sivaram Apr 9, 2026 ▶ 19:00
Assertion Supported
Sivaram: Study shows AI workloads offer 18% to 55% power flexibility
“There's a March, 20, 26 paper that was published actually by Emerald's chief scientist, Professor Aisha Koskun at Boston University with Two others, both at BU and at Emerald AI, that shows a 18 to 55% power flexibility opportunity across a range of different …”
Varun Sivaram Apr 9, 2026 ▶ 21:24
Assertion Supported
Emerald AI and Oracle migrated live inference queries to avoid winter peak
“So with Oracle, we showcased migrating AI workloads from one location, Virginia, to another, Chicago. And this was during the data winter, the Dominion winter peak period, the inference queries got rerouted in such a way that you were able to precisely meet th…”
Varun Sivaram Apr 9, 2026 ▶ 23:32
Prediction Not checkable as stated
Data centers will install large-scale behind-the-meter battery storage systems
“We're going to put a lot of batteries behind the meter at data centers. Like, I'm pretty convinced that that's going to happen.”
Shayle Kann Apr 9, 2026 ▶ 40:28
Prediction Open · timeframe Dec 2026
Emerald and NVIDIA to deploy 100MW power-flexible AI factory in 2026
“I think later this year, as you know, Emerald and NVIDIA and some other partners, Digital BLT, Epri, Dominion, and PJM, we will put together the world's first, ah, A hundred megawatt commercial scale AI factory that is truly power flexible. It's custom designe…”
Varun Sivaram Apr 9, 2026 ▶ 42:49
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 200 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more