Sora 2 launches: AI video strategy breakdown — who is really winning between OpenAI, Meta, and Google?

Summary

OpenAI's Sora 2 launch exposes sharp strategic divergence among the three major AI video platforms. Over three weeks, Google, Meta, and OpenAI shipped products targeting different user tiers. Google built tools for YouTube creators. Meta's Vibes app positioned itself as a consumption-first, dreamlike feed. OpenAI optimized for ease of use and virality among general consumers.

Sora is winning on output quality and memetic power. Videos like Sam Altman stealing GPUs from a CVS have gone viral on X, something neither Google's Vimeo V3 nor Meta's Vibes content has achieved. Sora's strength is generating short, inherently shareable scenes. When creators have the tool, they produce content that stops users mid-scroll. Video creation is friction-free enough that ordinary people can generate something entertaining in seconds.

The actual constraint is consumption. Scrolling the Sora feed is unpleasant. Jarring cuts, uncanny valley artifacts, and a general sense of wrongness make users describe the experience as making their skin crawl. Meta's Vibes leans into illustrated aesthetics and dreamlike imagery that users tolerate better at length, even if the content rarely travels beyond the app.

The core tension is whether creation ease translates to durable consumption habits. ChatGPT worked because each user generated customized text for their own use, fusing consumption and creation. With video, the two decouple. You might enjoy making a video of yourself as a golden retriever, but strangers may not want to watch it. OpenAI's cameo strategy adds novelty by featuring recognizable people or characters, but it remains unclear whether that novelty sustains engagement or burns out quickly.

OpenAI appears to be betting on network effects through creator economics. The company built infrastructure for monetization. If a creator uploads their likeness as a cameo and someone generates views using that cameo, revenue flows back. This mirrors YouTube's Content ID system, where copyright holders claim ad revenue. Building a self-sustaining creator economy requires more than plumbing. Platforms like TikTok and YouTube succeeded because they created real economic incentives: meaningful ad splits, creator funds, and audience-building mechanics. Sora currently has supply but no durable demand loop.

The copyright strategy powering Sora's training remains unresolved. OpenAI gave IP holders 24 hours to opt out of the training set, a permissive approach similar to how YouTube initially operated. Some IP holders negotiated partial restrictions, blocking certain songs or styles, while others appear to have opted out entirely. The pattern mirrors YouTube's own trajectory: move fast, train on everything, then settle copyright disputes later. Google paid roughly $1 billion to resolve YouTube copyright claims after acquisition. Whether OpenAI has already secured revenue-sharing deals with major rights holders or is deferring that reckoning is unclear.

For OpenAI's core business, Sora is a modest supplement at best. Even if Sora reaches Snapchat scale at 460 million daily active users, that represents a rounding error on OpenAI's valuation. The real strategic value is iterative. Millions of users creating and sharing video daily generates enormous volumes of feedback data that OpenAI can use to train better models faster. The app itself is the flywheel for model improvement.

The open question is whether OpenAI can make the feed itself compelling before users churn. Meta's Vibes suggests there is an audience for AI-generated visual content, but only if it does not feel uncanny or chaotic. Sora's creation tools are world-class. Its consumption experience remains a liability.