OpenAI is discontinuing Sora, its AI video generation platform, according to the Wall Street Journal. The product, which launched to enormous fanfare in early 2025 after months of demos that dominated social media, is being wound down after failing to gain meaningful traction in a crowded and fast-moving market.
This is a significant retreat. Sora was positioned as OpenAI's flagship expansion beyond text and images — proof that the company could own the entire generative media stack. The initial demos were genuinely impressive, generating coherent video clips that made competitors look years behind. OpenAI leaned hard into the hype, drip-feeding increasingly polished examples for months before launch.
But hype and product-market fit are different things. Sora launched into a market where Runway, Pika, Kling, and Google's Veo were all iterating fast. The quality gap that looked insurmountable in demo form narrowed quickly once competitors had a target to aim at. Meanwhile, Sora faced the same problem every video AI tool faces: professional creators need controllability, not just quality. Generating a cool 10-second clip is a party trick. Generating the specific shot a director needs, with consistent characters, camera angles, and lighting across a sequence — that's a product. Sora never convincingly solved this.
The 517-point Hacker News discussion surfaces a recurring theme: OpenAI may be spreading itself too thin. The company is simultaneously running ChatGPT, the API platform, enterprise sales, a hardware effort, and various media generation tools. Each of these is a company-sized problem. Killing Sora suggests someone inside OpenAI is finally prioritizing.
For developers building on generative video APIs, the lesson is familiar but worth repeating: don't build on a platform that isn't the company's core business. OpenAI's core business is language models and the ChatGPT ecosystem. Everything else is an experiment, and experiments get cut. If you're integrating video generation, Runway's API is a safer bet — video is their entire company, not a side project.
The broader signal here is that the 'do everything' era of AI labs may be ending. Training frontier models is expensive enough that even the best-funded companies have to pick their battles. OpenAI choosing to focus is probably the right call. The question is whether this is a one-time correction or the beginning of a wider contraction in scope across the industry.
I think they have started seeing scratches in data center build up.Sora was a perfect example of using a lot of compute to generate the video -> we need a lot of GPUs -> a lot of RAMs -> energy and landI am predicting in the next 6 months RAM shortage will soften, not too much, because war
I had so much fun making videos with my mom when it came out. During the first two weeks, we made over 100 cameo videos together - we were constantly running up against the upload limit. It unleashed tons of genuine creativity, joy, and laughter from us.After those first two weeks though, we just… d
As someone who generally liked the products that OpenAI puts out, I think Sora was their first product that I really didn't like. I liked GPT primarily because I felt like it respected me: I never felt like it was trying to distract me from my work or get me to waste time doomscrolling. It'
Notably, this primer on Sora safeguards was published only yesterday: https://openai.com/index/creating-with-sora-safely/Not a great look that either the teams responsible for Sora didn't know this was coming or the decision was so brash that things changed overnight.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
Good riddence to bad trash. To me, this idea represents the absolute worst of the AI wave (out of a lot to choose from): a corporate controlled endless stream of the feelies to keep people plugged in and scrolling for nobody’s benefit except those in control of the output. If “entertainment” can be