Less than a year after a promising launch, Sora, the video social network exclusively powered by OpenAI’s artificial intelligence, is taking its bow. This sudden shutdown, announced in March 2026, surprises an industry accustomed to spectacular Tech successes. However, behind this decision lie financial, technical, and strategic issues far more complex than a simple commercial mishap. How could an application that had attracted nearly a million users collapse so quickly? OpenAI has finally lifted the veil on the real reasons that motivated the removal of Sora, demystifying several misconceptions fueled by controversy and speculation.
This detailed analysis sheds light on all the challenges and critical choices OpenAI faced, offering valuable insight into the current dynamics of generative AI technology, the emergence of video content, and the titanic battles waged by the giants of Silicon Valley and beyond. From managing exponentially growing computational costs to the impact of legal challenges, and the price war triggered by competition, the dismantling of the proven reasons highlights a new key stage in the evolution of artificial intelligence applications dedicated to the general public and businesses.
This revelation also invites reflection on the limits of technological innovation if it is not supported by a viable economic model and appropriate regulation. This is the crossroads where OpenAI stands today, between disruptive ambitions and ruthless realities.
- 1 The Sora project: a revolutionary but short-lived AI video experience
- 2 Sky-high costs: a major barrier to Sora’s sustainability revealed by OpenAI
- 3 Paltry revenues in the face of towering expenses
- 4 The legal controversy: between copyrights and uncontrolled regulation
- 5 The price war and the rise of competition
- 6 Impact on the video artificial intelligence industry: toward a new phase
- 7 Strong alternatives to Sora: exploring the new champions of AI-generated video
- 8 Legacy and strategic lesson for OpenAI after Sora’s disappearance
The Sora project: a revolutionary but short-lived AI video experience
Launched with high hopes thanks to the power of its Sora 2 video model, the social network embodied a unique promise: to merge audiovisual creativity with automatic generation, at the heart of a social platform both playful and innovative. From its beginnings, the application managed to capture attention, with nearly a million downloads and a strong presence in the rankings of the most popular apps on the US App Store.
On paper, Sora combined all the ingredients for success. Its interface resembled that of giants like TikTok or Instagram, with a continuous feed of short videos, each clip automatically generated by artificial intelligence. However, as users explored this content, the line between fascination and discomfort became palpable. Indeed, the videos, often very colorful, with sometimes jerky or offbeat movements, fell into what is called the “uncanny valley,” where imitation seems almost human but not entirely convincing. This technical discomfort zone could deter viewers in the long term.
Beyond this strange appearance, Sora also offered important social functions. It was possible to like, comment, and even remix videos to modify elements, thus changing characters and atmosphere. This participatory dimension enhanced appeal but especially opened the door to exponential virality and a contagion phenomenon of content, positive as well as problematic.
Sky-high costs: a major barrier to Sora’s sustainability revealed by OpenAI
The main technical reason explaining OpenAI’s shutdown of Sora is undoubtedly the astronomical operating cost. Creating videos via artificial intelligence requires gargantuan computing power, especially when ensuring a smooth experience for hundreds of thousands of simultaneous users.
According to data from recent surveys by Forbes and financial analyses such as Deepak Mathivanan’s at Cantor Fitzgerald, the platform’s daily bill approached nearly 15 million dollars, amounting to an annual projection exceeding 5 billion. This staggering sum mainly results from GPU consumption, each roughly ten-second clip costing about 1.30 dollars in computing resources alone.
Bill Peebles, project manager, himself explained as early as October 2025 that the economics around Sora were “totally unsustainable.” This remark clearly illustrates the growing gap between the deployed technological power and the actual ability to monetize the application effectively at such a scale. Even with a million users, the financial burden remained out of reach.
Detailed cost analysis per video
The Sora 2 model, meant to deliver maximum quality, commands a fairly high unit price. To illustrate:
- Typical duration of a generated video: 10 seconds.
- GPU resource cost per video: about 1.30 dollars.
- Daily user volume sometimes exceeding several hundred thousand.
Each viewing incurs these computing costs, the real-time nature of which prevents many economies of scale. This situation rapidly amplified the financial impact until making the project economically untenable.
Paltry revenues in the face of towering expenses
OpenAI did not just invest in cutting-edge technology: the company had also planned a monetization strategy for Sora. This mainly relied on in-app purchases, allowing users access to certain premium features or advanced video customization.
However, these commercial efforts remained dramatically insufficient. Early 2026 revenues amounted to only 367,000 dollars per month, sharply down from 540,000 dollars recorded in December 2025. This imbalance maintained a revenue/cost ratio below 0.08%.
At the same time, the app’s traffic quickly declined, with a 45% drop in downloads recorded in the first quarter of 2026 according to TechCrunch. Sora thus went from high popularity to an active base below half a million, further weakening financial viability.
Why did Sora’s monetization fail?
Several factors explain these weak results:
- The lack of a strong subscription model: unlike other video platforms, Sora never succeeded in retaining users behind a recurring paid offer.
- The difficulty in valuing generated content: videos being automatically generated, their perceived value and users’ attachment remained limited.
- Competitive offering: the rise of more affordable and easily accessible alternatives dispersed consumer attention.
The legal controversy: between copyrights and uncontrolled regulation
Beyond financial aspects, one of the main woes contributing to Sora’s removal lies in the legal uncertainty surrounding the platform. The massive creation of videos inspired by protected works triggered a wave of disputes from rights holders in several countries.
OpenAI anticipated these tensions by forming a major partnership with Disney in December 2025, for 1 billion dollars, authorizing the use of over 200 licensed characters. However, this agreement remained an isolated exception. Other heavyweights, particularly in Japan, fiercely resisted uncontrolled use of their intellectual property.
The CODA organization, representing studios such as Studio Ghibli and Bandai Namco, launched a growing number of lawsuits, while Hollywood also mobilized, placing Sora at the center of a thorny legal conflict. This context became a serious obstacle, especially ahead of a planned OpenAI IPO. The threat of a cascade of litigations played a decisive role in the decision to abruptly halt the project.
The issues of watermarking AI-generated content
To limit abuses, Sora displayed a clearly visible watermark on each video indicating it had been created by artificial intelligence. This initiative aimed to protect users against misinformation and preserve trust in visual media.
However, the system was soon circumvented by determined internet users, who succeeded in removing this watermark, making these videos indistinguishable from real footage. This loophole greatly amplified the risks of manipulation and misinformation, weakening the platform’s credibility and fueling controversy.
The price war and the rise of competition
In this difficult context, the emergence of strong competition contributed to accelerating Sora’s downfall. In April 2026, Google responded to OpenAI’s announced closure by launching Veo 3.1 Lite, a generative video model accessible at significantly lower costs: about 0.05 dollars per second at 720p, a fraction of Sora’s unit cost.
Google thus reshuffled the cards in this segment, increasing price pressure and imposing an unprecedented development pace. Moreover, Google also revised the pricing of its Veo 3.1 Fast model, making video generation even more economical. This strategic offensive, supported by solid resources, marginalized Sora on the pricing front.
At the same time, other players such as Gemini with its Nano Banana 2 model helped hasten this transition by offering solutions that were both more reliable, with better technical control, and more financially competitive. The combination of these factors made maintaining Sora not only economically impossible but also strategically perilous.
Comparison of the main video generation AI pricing in 2026
| Product | Price per second | Resolution / key features | Advantages and disadvantages |
|---|---|---|---|
| OpenAI Sora 2 | ~$1.30 / 10 s (~$0.13/s) | 10 seconds HD video | Advanced quality, very high cost, limited duration |
| Google Veo 3.1 Lite | $0.05 / s (720p) | Eco HD video | Very low cost, good quality, limited duration |
| Gemini Nano Banana 2 | $0.08 / s (1080p) | 4K, good technical control | Reliable, competitive price |
| ByteDance Seedance 2.0 | $0.13 / s | 15 s, multimodal 1080p | Possible monthly subscription, parental control |
Impact on the video artificial intelligence industry: toward a new phase
The withdrawal of Sora resonated as a real shock in the sector, confirmed by the swift departure of Bill Peebles, project leader. This mishap marks the end of an era where technological prowess alone was enough to guarantee success. From now on, profitability, cost control, and respect for legal frameworks take precedence.
Several experts see in this decision the signal of a paradigm shift: generative AI, after a phase of exploration and market introduction, is entering a phase of maturity. Stakeholders must now develop sustainable industrial uses, with increased attention to economic viability and social and legal responsibility.
This turning point also invites reflection on the red lines companies must set to avoid comparable crises, with Sora’s failure painting a picture of risks and difficulties in this promising yet complex sector.
Strong alternatives to Sora: exploring the new champions of AI-generated video
With Sora’s disappearance, other solutions are emerging and currently dominate the market. Each option offers a particular balance between quality, price, and technical constraints, thus providing a comprehensive range for creators, businesses, and developers.
Among them, ByteDance’s Seedance 2.0 stands out with an advanced multimodal approach, capable of integrating text, images, and audio for creating clips up to 15 seconds in 1080p. This offer relies on a clear and competitive economic model, with a rate of €0.12 per second and a monthly subscription system, something Sora never managed to achieve.
Google Veo 3.1, with its ability to output 4K at 60 frames per second, remains a reference for high-quality productions. Its duration limit of 8 seconds however requires external editing for longer creations, an important compromise for some users.
Runway Gen-4.5, first in the Artificial Analysis 2026 benchmark, seduces with its cinematic rendering and high degree of creative control. However, its performance is marred by continuity errors, such as missing objects or unsynchronized actions, still revealing the current limits of AI video generation.
Kling AI 3.0 plays on length and native multilingual audio quality, combined with a reasonable price. Its strict content moderation has however earned some criticism regarding users’ creative freedom. This strict moderation limits certain cultural or sensitive uses, a still lively issue in AI discussions.
| Solution | Key features | Advantages | Limitations |
|---|---|---|---|
| Seedance 2.0 (ByteDance) | 15 s video, audio, multimodal, 1080p | Structured economic model, subscription, quality | Geographically limited, vigilance on copyright |
| Google Veo 3.1 | 4K 60 fps, audio/video synchronization | Exceptional quality, character coherence | Limited to 8 s, external editing needed |
| Runway Gen-4.5 | Cinematic, advanced creative control | Very realistic visual rendering | Logical causality issues, high cost |
| Kling AI 3.0 | 15 s 4K, native multilingual audio | Competitive duration and price | Strict moderation, imperfect transitions |
Legacy and strategic lesson for OpenAI after Sora’s disappearance
Beyond the project’s termination, OpenAI emerges transformed from this experience. The removal of Sora embodies a pivotal moment revealing that raw innovation is no longer enough. The proof: advanced video generation technology could not compensate for a deficient economic model and insufficient risk management, notably legal and social.
The project leader’s departure and the ensuing strategic pivot illustrate a strong awareness at OpenAI: the future of artificial intelligence now lies in pragmatic solutions adapted to real needs, particularly for businesses and developers, rather than in public experiments with uncertain effectiveness.
This evolution also fits into a broader context, where the generative AI sector is structuring, regulating, and seeking sustainability through innovations better framed financially, ethically, and legally.
