The off-grid data center fantasy

Episode Summary

Executive Summary: The episode debates how data centers should be powered amid explosive load growth: off-grid gas, behind-the-meter generation, or grid-connected flexibility. Tim Haid and Jigger Shah argue that fully islanded data centers are rare, hard to build, and politically counterproductive, while batteries, load flexibility, and grid-connected solutions can scale faster and better serve ratepayers and the energy transition.

Main Topics: Off-grid data centers are technically and operationally difficult (Priority: 5/5): Tim and Jigger argue that fully islanded data centers require building a microgrid-like system with scarce expertise, long lead times, and complex power electronics, making them more of a specialty or science experiment than a scalable model. Behind-the-meter gas versus batteries (Priority: 5/5): The discussion contrasts a likely near-term buildout of behind-the-meter gas generation with the case for batteries, especially CNI batteries, which Tim says can scale through AI-reduced transaction costs and Jigger says are better positioned financially and operationally. Load flexibility as the fastest scalable path (Priority: 5/5): Both speakers frame load flexibility—using software, market design, and existing grid infrastructure—as the main way to accommodate new data center demand in the next three to five years. Critique of SemiAnalysis forecast (Priority: 4/5): They sharply dispute the forecast of 40 GW of behind-the-meter data center power by 2028, arguing it ignores build-time realities, workforce constraints, supply chain bottlenecks, and flexibility. Rhodium’s transition framework and the climate debate (Priority: 4/5): The guests discuss Rhodium’s scoring of data center power choices, agreeing that grid-connected options are better but criticizing the framework for being outdated, too simplistic, and not useful for current policymaking. Politics, ratepayers, and moratoriums (Priority: 4/5): The conversation turns to how data centers are becoming politically toxic as governors face ratepayer anger, utility blame-shifting, and pressure to impose moratoriums while still attracting investment. Batteries as community and system assets (Priority: 3/5): Beyond industrial sites, Jigger emphasizes deploying batteries in homes, churches, food banks, schools, and emergency centers to build trust and deliver resiliency benefits more broadly.

Key Arguments: Fully islanded data centers are not a scalable answer because building them is equivalent to creating a new electric grid, requiring rare talent, long lead times, and specialized power electronics. The next three to five years will be dominated by using the grid already built more efficiently through load flexibility rather than constructing entirely new standalone infrastructure. SemiAnalysis’ 40 GW behind-the-meter projection is unrealistic because it assumes buildouts that cannot happen in 18 months and ignores 200-week lead times for key equipment. Utilities and regulators are increasingly aligned around flexibility solutions because they need near-term capacity and cannot wait for long-duration new-build generation. CNI batteries have been under-deployed despite representing a huge share of grid demand; AI can reduce transaction costs and help unlock this market. Batteries offer multiple revenue streams beyond capacity value, including retail peak reduction, arbitrage, ancillary services, and distribution deferral, making them more attractive than gas in many cases. Off-grid gas can still be useful as a bridge asset, but only if it eventually becomes grid-connected and operates as flexible peaking capacity rather than a permanent island. Politically, governors need solutions that visibly protect voters from rate hikes, which makes flexibility and community-benefit models more viable than purely private off-grid systems.

Data Points: Flex Summit timing and location: October in Austin, Texas - Latitude Media conference promoted at the top of the episode Discount code: PODS10 for 10% off - Listener promotion for Flex Summit registration Expected load growth: ~100 GW of net new demand over the next 3–5 years - Tim and Stephen discuss the scale of growth from data centers and electrification Battery deployment mix: ~90% utility-scale, ~10% residential, ~0% commercial and industrial - Tim explains the current U.S. battery deployment landscape Commercial/industrial share of grid demand: ~60% - Used to show the mismatch between demand and battery deployment CNI transaction costs: ~25% of a typical project - Tim says development, permitting, interconnection, modeling, and related soft costs are a major barrier Large data center threshold: 500 MW+ - Tim defines what counts as a large data center in the discussion Off-grid large data centers today: 0 on the planet Earth today - Tim argues there are no truly large, long-duration islanded data centers operating today Only known off-grid large example: Colossus - Jigger and Tim reference Elon Musk’s project in Mississippi/Memphis as a singular case Build time for off-grid projects: 36–48 months best case - Tim estimates the minimum practical schedule for these complex projects Lead times for key components: up to 200 weeks - Tim cites transformer/switchgear/conduit supply constraints Utility-scale battery interconnection wait: ~6 years in PJM - Tim cites average queue times for new utility-scale batteries Utility-scale battery interconnection wait: ~9 years in CAISO - Tim describes very long waits for new batteries to connect Utility-scale battery interconnection wait: north of 4 years in ERCOT - Tim says even Texas has meaningful delays SemiAnalysis forecast: 40 GW of behind-the-meter data center power by 2028 - The report being critiqued throughout the episode SemiAnalysis equipment demand forecast: 50 GW per year by 2029 - A second projection from the report Grid headroom claim in SemiAnalysis: negative by 2027 - They say spare capacity after peak demand will turn negative nationwide Alternative headroom estimate cited: 300 GW - Jigger references Amit Narayan/GridCare as arguing there is significant grid headroom Battery cost trend: down 27% last year - Jigger contrasts battery costs with gas costs Natural gas cost trend: up 27% last year - Jigger argues gas solutions are becoming more expensive Battery revenue streams: 5 - Jigger says batteries can earn value from multiple markets beyond capacity Utility-scale battery system value: 90% of the market today - Jigger notes most batteries are still utility-scale Data center project example: 2.7 GW - Microsoft’s Project Kilby in Texas, discussed as a possible model Batteries manufacturing support: through 2034 - Jigger references extended battery tax credits under OBBA Community capacity found by software: 650 MW - Jigger cites National Grid New York finding underused capacity via GridCare Governors under pressure: 36 governors running for election - Used to explain political sensitivity around data centers and rates Conference date mention: September 18 - Dervos conference referenced as part of fall event season

Pivotal Quotes: "the notion that you're going to do this en masse across the country is a deeply selfish decision" — Jigger Shah: Jigger criticizes the idea of large-scale permanent off-grid data centers for worsening supply chain constraints and not benefiting the broader system "the next three to five years is about load flexibility" — Tim Haid: Tim argues that the immediate path to serving new load is better use of the existing grid through flexibility, not massive new isolated infrastructure "you can reduce your peak load that your building is showing... you can get paid for that" — Jigger Shah: Jigger explains how batteries can monetize multiple grid services beyond simple capacity value

Implications: The industry’s near-term winners will likely be grid-connected, flexibility-enabled projects that reduce strain and support communities. Permanent off-grid data centers look unlikely to scale, while batteries, software, and market reform become the main tools for adding load without breaking the grid.