I don’t buy that OpenAI abruptly decided to shutdown Sora because they have something better coming soon, but I have to guesses as to why:

  • They found out it’s being used for something really bad and they might be held liable

  • They’re trying to get ahead of growing anti-AI backlash and are trying to present themselves as less “sloppy” since a lot more of the backlash is targeted at generated images and video.

  • The AI bubble is bursting and video generation is one of the more expensive parts of their business and they’re trying to cut costs.

  • @bridget

    4. They're freeing up compute for the DoD. OpenAI signed a contract with the Pentagon on February 28th to deploy models on classified networks. This is the same contract Anthropic got designated a 'supply chain risk' for walking away from because the DoD wouldn't guarantee against mass surveillance and autonomous weapons. Sora was burning $15M a day in compute and made $2.1M in its entire lifetime. A defense contract is worth $200M+ with guaranteed government money behind it. They didn't kill Sora because they have something better coming. They killed it because every GPU running Sora is a GPU not running classified inference for the military. OpenAI said it themselves, the shutdown was about 'compute reallocation.' The timeline speaks for itself.