←Story detail

Trending story

Old Ma says H100 rental prices are higher than factory prices and the payback period may shrink to 6 months

1 reports1 story updatesUpdated Oct 21 13:30

Understand this story

AI overview

On October 21, 2026, in his latest interview, Old Ma offered a judgment: the current rental price of Nvidia's H100 chip is more expensive than when it was brand new from the factory, and he stated that it is not impossible for the payback period of computing hardware to shorten to 6 months next year. He used this to explain why SpaceX insists on signing only short-term contracts when renting out computing power, the reason being that the useful intelligence produced per watt of electricity rises year by year, and short contracts are renewed at a higher unit price per watt upon expiry. He also mentioned that looking outward for GB300, GB200, Google TPU, Amazon Trainium, and even overseas H100, none are in stock. Related reports also pointed out that he himself is a seller of computing power rentals, and emphasized that the shortened payback period is only an expectation that is not entirely impossible. At present, this judgment appears only in this interview report, and there are no subsequent factual developments.

AI-generated from public reports

Latest development · Report

English translation is being prepared.

Developments

1 confirmed developments

  1. Old Ma claims H100 rental price is higher than factory price and payback period may shrink to 6 months

    Source:Ayi AI Notes

Reports and discussion timeline

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

Oct 21

  1. ayi_ai_notes

    English translation is being prepared.

Related stories

  1. Related storiesMusk says H100 rental prices are higher than factory prices, and payback period may shrink to 6 months next year
  2. Related storiesMeta releases open-weight model Muse Spark Open
  3. Related storiesThe German federal government plans to cooperate with Nvidia
  4. Related storiesDeepSeek open-sources inference framework DeepInfer