AI Energy Crisis Concern

Nick Huber believes AI is overhyped and unsustainable in its current form, not because the technology lacks long-term potential, but because the energy costs required to power it will decrease quality of life for most people. While he sees personal value in AI tools, the business case hasn't materialized, and the infrastructure demands are already causing real economic pain.

Key Points:

• Long-term vs. Short-term Reality:

  • Similar to the internet boom where only 1% of companies survived the bubble
  • Not doubting AI's power to change the world over the next 20 years
  • Current hype, money, and energy being poured in is unsustainable

• Personal Experience with AI Implementation:

  • Added 15 AI tools across his portfolio companies
  • Canceled 11-13 of them—couldn't justify the cost
  • AI is "unbelievable" for personal use: doctor in pocket, recipes, daily tasks
  • Business value is unclear: "where is the long term value?"

• The Energy Crisis Problem:

  • Energy bills in Athens, Georgia went up 60% in the last three years
  • Raised six times due to data center construction
  • 30 data centers came online in Georgia in the last four years
  • 30 more in the pipeline currently being built
  • These centers negotiated "unsustainable energy rates" with local municipalities

• The 99% vs. 1% Divide:

  • Companies may replace employees with AI (benefiting the 1%)
  • But energy costs will devastate everyone else's quality of life
  • Basic necessities like dishwashers, air conditioners, and heaters require massive electricity
  • More expensive energy means less money for things that improve quality of life

• Subsidization Masking True Costs:

  • Current AI tools are subsidized "all the way down the chain"
  • Nvidia sells chips and makes bank
  • Amazon cloud services subsidizes costs
  • VC-backed companies subsidize data usage for the "land grab"
  • It's going to get more expensive as subsidies end

• Diminishing Returns Ahead:

  • Compares to Elon Musk promising self-driving "five years away" for 20 years
  • Will get "harder and harder and harder" to improve AI from here
  • Will require exponentially more energy for incremental improvements

• The Nuclear Solution:

  • Would feel "much different about AI" if we had nuclear power plants everywhere
  • Energy infrastructure is the core issue, not the technology itself

• Current Stance:

  • "Very bearish on AI right now"
  • The promise doesn't match the reality for business use cases
  • The energy cost externality will hurt ordinary people most