Commentary

Sora 2 launches as OpenAI bets on AI video as a content platform

Sep 30, 2025

Key Points

  • OpenAI's Sora 2 matches or exceeds Google's video generation quality, but the speed of improvement from Sora 1 raises unanswered questions about whether competitors can replicate the jump.
  • OpenAI is positioning Sora as a consumption platform competing with Meta's Vibes, but users default to posting AI video on established networks where audiences already exist.
  • OpenAI's separate agentic commerce launch adds a second revenue lever with take rates potentially scaling from Amazon's 8-15% to Meta's 30% advertising model.

Summary

OpenAI released Sora 2, a video generation app that matches or exceeds Google's Vids in quality. The launch video shows 12-second clips with improved lighting, motion physics, and visual consistency polished enough to serve as standalone content rather than demos.

Sora 1 lagged behind Google's offering for months. The speed of this catch-up raises questions about OpenAI's approach. Whether they licensed new training data, discovered superior algorithmic techniques, or benefited from partnerships matters because it determines whether competitors like Anthropic can replicate the jump.

OpenAI is positioning Sora as a content platform, not just a tool. The company is chasing what Meta attempted with Vibes—a consumption network where users create, post, and browse AI-generated video natively. The economics would be enormous if it worked. If ChatGPT captures $300 billion in e-commerce at a 20% take rate, that's $60 billion in revenue.

But the platform play faces structural headwinds, according to Justine Moore at Andreessen Horowitz. Users default to posting AI-generated content on established social networks—X, Reddit, Instagram, TikTok—where the audience already exists. Building a new consumption platform requires critical mass that's hard to achieve when creators fragment across multiple outlets. A feed of only Sora-generated video also risks aesthetic monotony. Meta's Vibes leaned heavily on Midjourney's visual signature. OpenAI is trying to signal style diversity in promotional materials, but whether that translates to actual feed quality remains untested.

The practical test is live: what can an average user actually generate in the app, and does it match the launch video? Meta has the music advantage—access to Instagram's full catalog of licensed tracks, which lends polish to short-form video. Sora 2 will rely on AI-generated audio and sound, which is newer territory.

OpenAI's separate agentic commerce launch yesterday adds another revenue lever. The protocol currently has limits—single-product purchases only, physical goods only—but those constraints are temporary. If take rates follow Amazon's precedent of 8–15% and scale toward Meta's current advertising model approaching 30%, OpenAI could build a second major cash engine alongside content consumption.