• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Musk Says 15GW of 2027 AI Compute Cannot Activate

    Musk highlights infrastructure challenges beyond power for activating new AI systems.

    EM
    KA
    T(
    5 Sources, 32d ago, first seen 32d ago

    TLDR

    In a post on X, Elon Musk stated that a consensus estimate shows roughly 15GW of AI compute produced in 2027 cannot be turned on that year. He explained activation requires building transformers, wiring, liquid-cooling, massive chillers and complex networking in addition to securing power. Research engineer Yacine retweeted the message. A linked Grok conversation examines the probable 2027 AI power shortfall relative to compute production.

    Combined views

    6.1M

    5 Sources, first seen 32d ago

    Combined views

    6.1M

    5 Sources, first seen 32d ago

    19.2K likes
    19.2K likes
    2K comments
    3.5K saves
    4.3K reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    2K comments
    3.5K saves
    4.3K reposts
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    5 Sources

    @elonmuskConsensus estimate is that ~15GW of AI compute produced in 2027 cannot be turned on in 2027. This is harder than just finding power, as you also need to build out all the transformers, wiring, liquid-cooling, (massive) chillers & complex networking. https://grok.com/share/bGVnYWN5_4f06d401-90e9-4ed4-bad8-89e3b4fadde2
    @teortaxesTexSo Elon/Gork expect ≈20 GW of compute coming online in the US in 2027, vs ≈5 GW for China. Given chip quality disparity, that's about a 20x gap, maybe up to 50x if Rubins, Jalapenos etc still have a lot of optimization headroom.
    @DanielleFongis that a challenge? that sounds like a challenge
    @yacineMTBRT @elonmusk: Consensus estimate is that ~15GW of AI compute produced in 2027 cannot be turned on in 2027. This is harder than just findi…
    @FundaAIWhen Elon was doing research using Grok, the first source cited was FundaAI.

    5 Sources

    @elonmuskConsensus estimate is that ~15GW of AI compute produced in 2027 cannot be turned on in 2027. This is harder than just finding power, as you also need to build out all the transformers, wiring, liquid-cooling, (massive) chillers & complex networking. https://grok.com/share/bGVnYWN5_4f06d401-90e9-4ed4-bad8-89e3b4fadde2
    @teortaxesTexSo Elon/Gork expect ≈20 GW of compute coming online in the US in 2027, vs ≈5 GW for China. Given chip quality disparity, that's about a 20x gap, maybe up to 50x if Rubins, Jalapenos etc still have a lot of optimization headroom.
    @DanielleFongis that a challenge? that sounds like a challenge
    @yacineMTBRT @elonmusk: Consensus estimate is that ~15GW of AI compute produced in 2027 cannot be turned on in 2027. This is harder than just findi…
    @FundaAIWhen Elon was doing research using Grok, the first source cited was FundaAI.