←Story detail

Trending story

Evan Conrad says either the B300 inference price or the market rental price must change

1 reports1 story updatesUpdated Oct 22 20:21

Understand this story

AI overview

On October 22, 2026, Evan Conrad of SF Compute put forward a judgment: a single B300 card running inference can earn 40 dollars per hour, while buying it on the market for one hour costs only 5 to 6 dollars; one of these two numbers must change. He also believes that compute has not yet been truly priced. The related report was published by Saito Hardcore Hacker on the same day, and there are currently no subsequent factual developments.

AI-generated from public reports

Latest development · Report

SF Compute's Evan Conrad: One of the two must change—B300 inference renting at $40/hour versus the market price of $5-6

Developments

1 confirmed developments

  1. Evan Conrad says either the B300 inference price or the market rental price must change

    Source:English text is being prepared.

Reports and discussion timeline

Browse reports and discussion by time. Their evidence levels remain distinct.

Oct 22

  1. Saito Hardcore Hacker

    SF Compute's Evan Conrad: One of the two must change—B300 inference renting at $40/hour versus the market price of $5-6

    SF Compute's Evan Conrad points out that a B300 card running inference can earn $40 per hour, while buying it on the market for an hour costs only $5 to $6. One of these two numbers must change. He believes that compute has still not been truly priced to date.

Related stories

  1. Related storiesDeepSeek open-sources inference framework DeepInfer
  2. Related storiesMeta releases open-weight model Muse Spark Open
  3. Related storiesDeepSeek open-sources inference framework DeepInfer
  4. Related storiesJassy says most AWS capacity is already booked through 2028