‹ BackHN Continuity

Thread

Micron CEO Says Memory Supply Will Be Much Tighter in 2027 and 2028 Than in 2026

394 points · 443 comments · speckx

  1. PowerElectronix · · focus · HN ↗
    It'd be a shame if their two main sources of income (openai and anthropic) were to go bust...
    1. layer8 · · focus · HN ↗
      In that unlikely case, open-weight hosting providers (and Google and Meta and SpaceXAI) would pick up the slack.

      The only chance of memory demand going down would be breakthroughs in model size reduction.

      1. wood_spirit · · focus · HN ↗
        The catchup models are all basically distillations of the sota models. Without the next training cycle things are going to stagnate. And as all those companies you mentioned are all linked to OpenAI and Anthropic and relying on hosting deals and things they are all going to be in the ringer when things to go south.

        So on the one hand you have all the sota model makers doing investor expectation management in saying wet need a slowdown for safety (may or may not be true, but also means they don’t spend on the next training cycle before IPO? Could be making their books look better too?) and on the other we have a sense that the models aren’t yet at the stopping place where we can just not train another cycle and use distillations of the current generation?

        1. lnenad · · focus · HN ↗
          > The catchup models are all basically distillations of the sota models

          Source/citation?

          1. wood_spirit · · focus · HN ↗
            <a href="https:&#x2F;&#x2F;www.dwarkesh.com&#x2F;p&#x2F;john-beren-charlie" rel="nofollow">https:&#x2F;&#x2F;www.dwarkesh.com&#x2F;p&#x2F;john-beren-charlie Is one of many and it describes how the dynamics well and talks about how proxy logs help.
            1. lnenad · · focus · HN ↗
              Thoughts&#x2F;opinions aren&#x27;t source.
              1. wood_spirit · · focus · HN ↗
                The people in that interview stating this are working at the forefront of open source models and research and includes a cofounder of OpenAI. They don’t say it like it is conjecture.
      2. formvoltron · · focus · HN ↗
        Also gpu speedup. If vera Rubin is 7x faster then to serve same number of tokens you need 1&#x2F;7 the memory. Did I get that right?
Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.