Divya van Mahajan

The Right Way to Deal With AI Data Centers

ai-daily-brief-podcast

The Right Way to Deal with AI Data Centers

Overview

This episode of The AI Daily Brief (dated June 23, 2026) addresses the growing public and political backlash against AI data centers, arguing that the conversation on both sides is reductive and that communities stand to gain significantly more from data center negotiations than they currently realize. The host also covers several headline stories: the NSA/Mythos red-team clarification, OpenAI’s cybersecurity model update, quantum computing executive orders, SpaceX’s NeoCloud deals, and Google’s talent departures.

Source video: URL not provided.

Prerequisites

  • Basic familiarity with AI infrastructure concepts (data centers, GPU compute, model training)
  • General awareness of the AI competitive landscape (Anthropic, OpenAI, Google DeepMind, Meta)
  • Understanding of how electricity markets and grid infrastructure work
  • Familiarity with U.S. political discourse around tech regulation and energy policy

Main Points

Headline: NSA/Mythos Red-Team Clarification

  • Reports claiming Mythos (an advanced AI model) had been used to hack the NSA were based on a misunderstanding of testimony given to Senator Mark Warner.
  • The Economist’s Shashank Joshi confirmed via a U.S. official that NSA Director Rudd’s remarks referred to a controlled red-team exercise under Project Glasswing, not a live cyberattack.
  • The NSA’s red teams no longer have access to Mythos because their authorization under Project Glasswing has lapsed.
  • Key takeaway: Mythos does not enable arbitrary intrusion into classified systems, but it does accelerate exploit design and execution once initial access is gained, raising the stakes for defenders.

Headline: OpenAI’s Cybersecurity Model and “Patch the Planet”

  • OpenAI expanded its Daybreak security initiative (launched in May as a counterpart to Anthropic’s Project Glasswing), releasing the full version of GPT-5.5 Cyber, their first cybersecurity-fine-tuned model with reduced guardrails for professional use.
  • OpenAI claims GPT-5.5 Cyber surpasses Mythos on the CyberGym benchmark.
  • A new initiative, Patch the Planet, launched in partnership with Trail of Bits, has already identified hundreds of bugs in open-source libraries and deployed 37 patches; over 30 open-source projects have joined.
  • Trail of Bits observed a fundamental shift: the expensive part of security work has moved from finding bugs to confirming, patching, and coordinating disclosure — areas now being flooded by AI-generated reports.
  • The Five Eyes intelligence agencies issued a rare public bulletin warning that AI-driven cyber risk is evolving faster than organizational assumptions, calling cyber risk “a core business risk and leadership responsibility.”

Headline: Quantum Computing Executive Orders

  • President Trump signed two executive orders directing federal agencies to collaborate with private industry to deploy a quantum computer for scientific research by 2028 and migrate federal systems to quantum-secure cryptography by 2031.
  • A second order focuses on protecting intellectual property within quantum companies and hardening the supply chain against adversarial acquisition.
  • The host notes genuine uncertainty about the current state of the technology and the motivations behind the orders.

Headline: SpaceX NeoCloud Expansion

  • Reflection AI signed a deal to pay $150M/month for compute capacity at SpaceX’s Colossus 2 data center, running through 2029 for a total deal value of $6.3 billion (smaller than Anthropic’s and Google’s ~$1B/month deals).
  • Analyst Swyx argued SpaceX is underestimated as a market force: it has recouped roughly half its investment in Colossus via compute deals, and operates uniquely as both a leading model lab and a NeoCloud provider simultaneously.
  • SpaceX stock nonetheless fell 16% during the same period; Google stock fell as much as 7.2% following the departures of Nobel laureate John Jumper (to Anthropic) and Noam Shazir (to OpenAI) from DeepMind.
  • The host cautions against over-reading personnel moves, noting Google’s unique advantages in TPUs and data.

Main Episode: The Right Way to Deal with AI Data Centers

The Political Landscape of Data Center Opposition

  • Opposition to data centers is becoming a mainstream, bipartisan issue, exemplified by figures ranging from comedian Theo Vaughn to Erin Brockovich to former Tea Party conservatives planning nationwide protests.
  • The two primary public concerns are: (1) water consumption and (2) electricity price increases.

Water Use: Putting the Numbers in Context

  • Amazon’s global data center operations consumed 2.5 billion gallons of water in 2025, down 2% year-over-year despite footprint expansion.
  • Comparative benchmarks:
    • U.S. golf courses: >500 billion gallons/year (a full year of Amazon data center water use ≈ one day of golf course maintenance)
    • California almonds: 1.2–1.8 trillion gallons/year (5–8× all U.S. data center water use)
    • U.S. leaky pipe losses: 3.29 trillion gallons/year (~15× data center use)
  • In Indiana, the Amazon facility cited by critics represents approximately 0.2% of the state’s domestic water use.
  • A widely cited book (Empire of AI by Karen Howe) contained a water consumption claim that was off by a factor of 1,000 due to a unit error; a correction was issued but the original claim remains in circulation.
  • NVIDIA has touted new liquid cooling approaches capable of reducing data center water use to near zero.

Electricity Prices: Short-Term Local vs. Long-Term National Picture

  • The Institute for Energy Research found no statistically significant correlation between data center concentration in a state and either current electricity prices or the rate of price increases.
  • However, a Bloomberg analysis of 25,000 grid nodes found prices rose as much as 276% since 2020 near major data center clusters, with over 70% of nodes recording increases located within 50 miles of significant data center activity.
  • The cause is constrained transmission capacity and generation not yet online — a grid infrastructure problem predating the data center build-out.
  • Policy responses include Oregon’s Power Act (requiring large electricity users to bear costs of infrastructure built for them) and the White House’s Ratepayer Protection Pledge, under which companies committed to:
    • Building, bringing, or buying new power supply
    • Paying for new power delivery infrastructure upgrades
    • Paying whether or not they use the power
    • Investing in local job creation and workforce development
    • Contributing to electric and community resilience

The Middle Path: Negotiating Real Community Benefits

  • The host argues the public debate is trapped in a false binary: either acquiesce to tech companies or oppose data centers entirely.
  • A middle path exists in which communities actively negotiate for significant financial concessions.
  • Example: In rural Richland Parish, Louisiana, hundreds of teachers received $50,000 bonuses funded by a surge in tax receipts tied to Meta’s AI campus construction — quintupling the bonuses mandated under an existing sales tax ordinance.
  • Labor unions are identified as a key actor capable of bridging community concerns and economic opportunity, representing both the affected communities and the skilled trades benefiting from construction demand.
  • The host’s conclusion: the priority should be moving from a “knee-jerk binary phase” to a “how everyone wins phase” through informed negotiation.

Key Concepts

  • Red-team exercise: A controlled, authorized security test in which an internal team attempts to breach systems to identify vulnerabilities — distinct from a live cyberattack.
  • Project Glasswing: Anthropic’s initiative providing AI model access to government cybersecurity partners; also the authorization framework under which NSA red teams accessed Mythos.
  • Daybreak: OpenAI’s cybersecurity initiative providing access to GPT-5.5 Cyber for vetted security professionals, positioned as a more democratized alternative to Glasswing.
  • CyberGym: A benchmark used to evaluate AI model performance on cybersecurity tasks.
  • Patch the Planet: OpenAI and Trail of Bits initiative deploying AI to find and patch vulnerabilities in open-source software libraries.
  • Five Eyes: Intelligence-sharing alliance comprising Australia, Canada, New Zealand, the United Kingdom, and the United States.
  • NeoCloud: A cloud compute provider built around next-generation GPU infrastructure; SpaceX’s Colossus facility is cited as an example.
  • Colossus 2: SpaceX’s large-scale GPU data center from which compute capacity is being leased to AI companies.
  • Oregon’s Power Act: State legislation requiring large electricity consumers (e.g., data centers) to fund grid infrastructure built specifically to serve them.
  • White House Ratepayer Protection Pledge: A voluntary commitment by major tech companies to protect consumers from electricity price increases caused by data center energy demands.
  • Quantum-secure cryptography: Encryption methods designed to remain secure against attacks from quantum computers.

Summary

The central argument of this episode is that the public debate over AI data centers has become unnecessarily polarized, with critics overstating harms and proponents dismissing concerns too quickly. The host methodically addresses the two dominant objections — water consumption and electricity prices — showing through comparative data that water use by data centers is modest relative to agriculture, golf courses, and infrastructure losses, and that electricity price pressure is primarily a symptom of underinvestment in grid infrastructure rather than an inherent feature of data centers. At the same time, the host acknowledges that localized, short-term price spikes are real and that dismissing all concerns is both intellectually dishonest and politically counterproductive. The episode’s constructive proposal is that communities, particularly through labor unions and local government, are in a position of significant negotiating leverage and should use it — pushing data center developers to fund teacher bonuses, local jobs, grid upgrades, and infrastructure rather than choosing between unconditional acceptance and outright opposition.