OpenAI Shuts Down Sora; AI Begins "Settling the Score"

What happened

OpenAI announced it is shutting down Sora, its consumer-facing AI video generator, removing the app, its API and the planned ChatGPT video integration. The move was abrupt: Sora went from a viral hit—reportedly surpassing one million downloads in ten days and peaking at about 3.33 million—to being taken offline after roughly six months. According to mobile analytics firm Appfigures, the product generated only about $2.1 million in in‑app purchases across its life, while operational costs for video generation proved enormous.

Why Sora was stopped

The calculus was simple: stunning user interest did not translate into sustainable revenue. Video generation is one of the most compute‑intensive AI tasks and, at scale, can produce daily cloud bills in the millions of dollars. Growth, in Sora’s case, amplified losses: the more people used it, the larger the deficit. It has been reported that OpenAI is reallocating the Sora research team to “world simulation” work and refocusing compute and personnel on ChatGPT, enterprise tools and coding assistants. It has also been reported that company leadership—citing competitive and capital pressures—decided to prioritize projects with clearer, faster paths to monetization.

Bigger picture: from capability to profitability

Sora’s fate underscores a broader industry pivot. For the past few years the dominant narrative was “can we build it?”—pushing the frontier of multimodal capabilities. Now, with compute scarce, expensive and geopolitically sensitive, the emphasis is shifting to “can it pay?” It has been reported that OpenAI secured roughly $110 billion in funding at a reported $730 billion pre‑money valuation, with portions (reportedly including a $35 billion Amazon tranche) tied to future milestones such as AGI or IPO; at the same time the company faces large reported losses and pressure to show viable revenue streams. Competitors such as Anthropic have focused more narrowly on enterprise and coding products that scale to revenue faster—an approach that is forcing a reckoning across the sector.

What this means for the market

The industry is entering a “算账时代” — an era of accounting. Flashy demo products that burn compute but lack durable monetization will be pared back. For Western and Chinese firms alike, the lesson is clear: technical novelty alone no longer guarantees continued investment. Geopolitical factors—export controls on advanced chips and cloud concentration among a few providers—also make access to affordable compute a strategic constraint. Will we see more pruning of consumer‑facing, high‑cost generative projects? Likely so. The winners will be those who convert model strength into repeatable, high‑margin business models.