OpenAI sharpens its focus as Sora falls and funding web tightens
OpenAI’s decision to shut down its Sora video tool and unwind its Disney partnership in the space of 24 hours is far from an isolated product call.
It comes at a moment of rapid growth for the firm, which is handling heavy infrastructure demands alongside increasing scrutiny over how the AI boom it sits at the centre of is being financed.
Sora was one of the most high-profile launches in generative AI.
Introduced back in 2024, it allowed its users to render realistic video clips from short text prompts and rapidly became a viral demonstration of the tech’s capabilities.
A standalone app shortly followed, pushing Sam Altman’s AI juggernaut into the same short-form video arena dominated by the likes of Tiktok, Youtube, and Instagram.
It’s reception was mixed, and any technical praise came plagued with copyright and deepfake concerns.
OpenAI tightened safeguards accordingly, but the tool remained entangled in a wider debate over how generative AI should interact with creative industries.
The conversation slightly shifted at the tail end of 2025, when Disney signed a three-year deal to license over 200 of its characters for use within Sora.
The handshake was seen as a potential bridge between the geographically close, but digitally distant Silicon Valley and Hollywood.
It has now been abandoned, with Disney saying it will continue exploring AI partnerships elsewhere.
OpenAI said it is discontinuing its Sora platform to focus its resources on other priorities, putting its energy instead into research and robotics. The company has also pointed to growing constraints on computing capacity.
Chief financial officer Sarah Friar announced that OpenAI is having to make “hard choices” about where that capacity should be deployed.
Video generation is among the most power-hungry uses of the technology, particularly at high quality, and sits beside competing demands from core products like ChatGPT.
Pressure to prioritise
Sora’s short life-span comes as OpenAI’s business continues to scale at dizzying rates. The company generated around $13.1bn in revenue last year, with roughly 60 per cent from consumer products, and 40 per cent from enterprise customers.
That split seems to be involving, with enterprise demand growing faster as every sector turns their attention, and capital, into AI.
It has bagged clients like Jetblue and Estée Lauder, and its model is increasingly based on subscriptions and usage, as supposed to consumer engagement alone.
But at the same time, OpenAI has raised more than $120bn in funding from a mix of venture capital, private equity and major tech titans.
Executives have reportedly said they are preparing the business for a public market entry, despite no timeline having been set.
The scale of that funding, coupled with rising infrastructure costs, have compounded OpenAI’s strategy decisions.
Training and running advanced AI models requires huge amounts of power, and OpenAI hasn’t shied away from its capacity limits at any one time.
Product decisions are beginning to shift due to those constraints, with the more expensive of them facing greater scrutiny than those tied more directly to enterprise demand.
Sora, while technically impressive, sat in a category that combined high cost with regulatory complexity, and less certain commercial returns.
Its closure is sign of narrower priorities as the company concentrated on more lucrative areas.
The circular deals behind the AI boom
It also sits within a broader financial structure that is drawing increasing attention.
OpenAI sits at the centre of an expanding network of deals spanning chipmakers and infrastructure firms.
Agreements with the likes of Nvidia, AMD, Oracle and Coreweave have secured access to the computing power needed to train and run AI models at scale.
Many of these arrangements involve a mix of long-term purchasing commitments, equity stakes and investment flows between just a handful of companies.
Nvidia, for example, has committed tens of billions to support AI infrastructure, while also supplying chips that power it. OpenAI, in turn, commits to buying large volumes of that hardware.
Elsewhere, Oracle is building out data centres to meet OpenAI demand while purchasing hardware from chipmakers.
Cloud providers are investing heavily in capacity that AI firms then agree to use. And in some cases, companies take equity stakes in one another alongside these supply agreements.
This structure has prompted comparisons to earlier tech cycles. Money can appear to circulate within a closed loop, raising concerns about whether headline growth actually reflects underlying demand or is being amplified by interlinked financing.
On the other hand, AI firms require specialised hardware at a scale few industries have ever seen, and suppliers with strong balance sheets are using that position to support customers that will ultimately drive demand.
In that sense, the arrangements look more like established forms of vendor financing, rather than empty revenue estimates.
What remains clear is that the ecosystem is becoming smaller. Chipmakers depend on AI developers, who depend on cloud and infrastructure, and so on.
The result is an interdependent system where growth assumptions are shared across the same group of firms, raising the stakes for the sector.
As if AI demand continues to soar, the model will support that rapid expansion. But if it slows, the impact would be felt across the entire chain.
OpenAI claims that the tech it developed through Sora will still be used in various other areas, including “world simulation” research.
While Sora’s rise and fall was rapid, it has shown that as the AI boom matures, decisions will become increasingly shaped by cost, compute and the commercial priorities of Big Tech, as much as technical possibility.