Episode Summary
Executive Summary: The episode examines the mounting backlash against data centers and AI, arguing it stems from two distinct forces: fixable local impacts like rates, water, and transparency, and a deeper values fight over AI itself. Brian Janice says good developers can address the former with better siting, tariffs, and community benefits, but the latter requires the AI industry to better explain why the technology matters. The conversation also explores emerging responses such as behind-the-meter generation, off-grid/edge builds, and why these are likely partial, imperfect, and costly solutions.
Main Topics: Two kinds of backlash: fixable local impacts vs. broader AI opposition (Priority: 5/5): Shayle and Brian separate community objections into negotiable concerns (power bills, water, secrecy, emissions) and a more fundamental rejection of AI and big tech. The latter is harder to solve because it is rooted in values, not project design. How data center development shifted from welcome to resistance (Priority: 5/5): Brian describes a fast change from communities rolling out the 'red carpet' to 'pitchforks.' The jump is driven by much larger project scale, public misunderstanding of resource use, and a sudden rise in attention on data centers as one of the few large-scale things being built in the U.S. Transparency, NDAs, and responsible development (Priority: 4/5): The discussion covers secrecy in data center siting and Brian’s view that NDAs with officials are poor practice. Transparency helps, but it does not fully overcome emotional opposition or distrust of AI and hyperscalers. Power prices, inflation, and community benefits (Priority: 5/5): They unpack the paradox that new data centers can lower local rates while contributing to broader system-level inflation through supply-chain strain and equipment shortages. Brian argues the industry should target community benefits and bill credits to offset real short-term pain. Behind-the-meter and bring-your-own-capacity responses (Priority: 4/5): States are pushing data centers to bring their own generation/capacity. Brian is skeptical of true behind-the-meter solutions, saying they can worsen system economics and may be more politically appealing than technically useful; he prefers utility-integrated, fully funded capacity additions. Future siting models: off-grid, underwater, space, and edge (Priority: 4/5): The conversation considers whether constrained grid access could push data centers offshore, underwater, into space, or toward edge computing. Brian thinks some niche use cases will emerge, but most compute will remain in large terrestrial data centers, with edge and alternative siting remaining hard businesses.
Key Arguments: Community backlash is partly solvable with better project design, fair cost allocation, and real local benefits, but a broad anti-AI sentiment is not something a single developer can fix. The industry underestimated the emotional and political dimensions of public reaction by focusing too much on facts and not enough on why people should want AI. Project scale matters: building in months what took decades in places like Quincy, Washington, creates a much bigger target for scrutiny. Transparency is necessary but insufficient; even disclosed projects can be rejected if the underlying issue is distrust of AI or big tech. Data centers may lower local rates at the project level, but their aggregate buildout can raise costs across the system through supply-chain and equipment inflation. Behind-the-meter generation is not a clean solution because it can require overbuilding, create local emissions concerns, and fail to reduce broader grid pressure. The most viable near-term response is to make data centers financially responsible for new capacity and to structure community benefits around specific local pain points. Alternative models like underwater, offshore, space, and edge computing may work in niche cases, but most compute will still likely be built in large terrestrial facilities.
Data Points: First data center project year for Brian Janice: 2006 - Brian says he worked on his first data center project in 2006, illustrating nearly 20 years in the sector. Quincy, Washington data center buildout period: ~20 years - Example of Microsoft scaling to near-gigawatt capacity gradually over two decades. Quincy unemployment rate change: 29% to 6% - Used to show long-term community benefits from data center investment. Quincy aquatic center investment: $15 million - Community benefit tied to long-term data center presence. Quincy school investment: $150 million - Another example of downstream local benefits. Energy Hub device count: 2.5 million customer devices - Promotional segment describing VPP aggregation resources. Energy Hub dispatchable capacity: 3.4 gigawatts - Promotional segment comparing VPP output to traditional generation. Equivalent grid capacity: More than three nuclear reactors - Promotional comparison for VPP flexibility. Data center property tax trend in Loudoun County: Down every year for the last decade - Example Brian cites to show local fiscal benefits despite backlash. North Dakota facility scale referenced: Applied Digital facility - Cited as an example of a positive local story around jobs and tax revenue. Grid build example in Oklahoma: 100% sourcing of wind, solar, storage, and other generation - Brian describes a project where all capacity additions are procured and packaged for the utility. Gigawatt data center extra capacity requirement: 2.6 gigawatts - Brian says one gigawatt-scale project required 2.6 GW of generation and batteries to support it. Large data center threshold in Brian’s framing: 100 MW or more - His definition of large-scale facilities likely to dominate future compute. Lower bound he hears from customers: 50 MW - Shows that even desperate buyers still need substantial scale. Underwater prototype scale: Kilowatt scale - Microsoft’s initial underwater data center was an experiment, not a commercial deployment. Second underwater prototype scale: Megawatt scale - Microsoft later built a much larger underwater unit in the North Sea.
Pivotal Quotes: "From red carpet to pitchforks" — Brian Janice: Describing how community sentiment toward data centers changed over a short period. "You can't redesign your way out of a values fight." — Shayle Khan: Summarizing the idea that technical fixes cannot solve opposition rooted in beliefs about AI. "The industry has tried to attack this with facts and really miss the emotional resonance of like people just aren't happy about this stuff right now." — Brian Janice: On why data center and AI messaging has failed to shift public sentiment.
Implications: Data center developers will likely need cleaner siting, transparent practices, and targeted community compensation, but AI’s political and cultural backlash may still constrain growth. Expect more experimentation with behind-the-meter, edge, and alternative siting models, though large grid-connected facilities will likely remain dominant.