Gavin Baker on

ai infrastructure

5 quotes · 10 posts · Aug 2024 – Aug 2026

Saidverbatim, newest first

  1. Baker identifies memory as the single most important factor for increasing token output per compute unit.

    “It's the single most important thing you could do to increase token output per unit of compute, and then that obviously, definitionally, actually lowers costs,”

    32:26 · Invest Like the Best · 4 Aug 2026 · permalink
  2. Baker explains MFU runs at 35-40%, measuring the percentage of theoretical compute actually used for training.

    “MFU, model flops utilization, and that generally runs around 35 to 40%. And that's literally the percentage of compute, theoretical compute flops that you're actually applying to trading.”

    15:44 · listen · Invest Like the Best · 27 Aug 2024 · permalink
  3. Baker explains higher MFU allows 25% faster time to market with same GPU and power spending.

    “You have the same amount of GPUs and the same amount of power presumably. You could choose between faster time to market.”

    16:24 · listen · Invest Like the Best · 27 Aug 2024 · permalink
  4. Baker explains that 50% MFU versus 40% MFU enables 25% faster time to market for AI models.

    “You could choose between faster time to market. If you run a 50% MFU and your competitor's running 40 for an equivalent amount of trading flops, you could be in market 25% faster.”

    16:28 · listen · Invest Like the Best · 27 Aug 2024 · permalink
  5. Baker proposes a new metric decomposing MFU that he developed the morning of the interview.

    “MFU is the most important metric because it gives you all of these advantages and ways to differentiate yourself amongst five people, six people who've trained these GPT core class models.”

    17:53 · listen · Invest Like the Best · 27 Aug 2024 · permalink

Postedtheir own words, on X

  1. post Baker predicts vertically focused AI native companies will accelerate due to routers, open-source models and specialized post-training

  2. post Baker argues game theory suggests memory LTAs are durable and Nvidia may have significant cloud revenue shares

  3. post Baker hypothesizes tomorrow's Ultrafast will be GPUs or Jalapeño with CS-4 or CS-5

  4. post Baker speculates Ultrafast will use GPUs plus CS-4/5, notes Jalapeño not financeable at GPU rates

  5. post Baker suggests NVIDIA accelerators have residual value that distinguishes them from first-party ASICs, responding to concerns about financing OpenAI hardware

  6. post Baker questions whether AFD is compatible with data locality, disputes OpenAI bet against disaggregation

Everything Gavin Baker is on record saying · RSS