US Media Reveals the Inside Story of Sora's Shutdown: Losing Millions Daily, User Base Halved, and Facing Talent Poaching from Competitors

marsbit2026-03-30 tarihinde yayınlandı2026-03-30 tarihinde güncellendi

Özet

OpenAI has abruptly shut down its Sora video-generation tool, a project once hailed as the next consumer AI frontier after ChatGPT. Initially launched with high expectations, including a major partnership with Disney, Sora allowed users to insert themselves into various video scenes. However, it quickly became a significant financial drain, reportedly losing about $1 million daily. Its user base peaked at roughly one million before halving, and it consumed excessive computational resources needed for OpenAI's core priorities, such as its upcoming "Spud" AI model and enterprise productivity tools. Facing stiff competition from rivals like Anthropic and a strategic shift towards AI agents and coding tools ahead of its IPO, OpenAI decided to reallocate resources. Key researchers behind Sora were also targeted in a Silicon Valley talent war, with Meta's CEO personally attempting to recruit them. The shutdown shocked Disney, which had planned to invest $1 billion and use Sora across its operations. The deal has now collapsed, and Disney is exploring other AI partnerships. OpenAI will focus its efforts on robotics and other long-term projects.

Earlier this month, when OpenAI CEO Sam Altman arrived in Los Angeles to attend Vanity Fair's post-Oscars party, his company was just weeks away from licensing the Sora video generation tool to major Hollywood studios.

Following the massive success of ChatGPT, Sora was hyped as the next consumer-facing AI frontier. This simple app allowed users to place themselves and their friends into various video scenes, whether dribbling like the Harlem Globetrotters or engaging in a lightsaber duel with the Star Wars character Darth Vader.

Then-Disney CEO Bob Iger was also optimistic about its prospects, agreeing to invest $1 billion in OpenAI and allowing characters from the studio's Marvel, Pixar Animation, and other properties to appear in Sora-generated videos. Equally important, amid widespread industry concerns about AI's threat to creative work, he lent the emerging technology the valuable seal of approval from Disney.

Computational Burden

However, OpenAI suddenly decided to shut down Sora.

Disney executives were shocked, many of them learning of the decision less than an hour before it was announced. What they didn't know was that Sora had quietly become a major burden for OpenAI in the months following its launch, especially as the startup was tightening its business focus ahead of its initial public offering (IPO).

At the time, OpenAI was just weeks away from completing a new AI model codenamed "Spud" and needed to free up more computational resources to support the coding tools and enterprise products running on that model. AI chips are the most precious resource for any top research lab, and at OpenAI, Sora was consuming too many of them.

Moreover, Sora was not yet profitable, and every user who "stitched" themselves into a WWII newsreel or a Hollywood chase scene was consuming finite computing resources.

Strategic Miscalculation

Today, Sora looks like a costly strategic miscalculation, and the key employees who led the project are at the center of the AI talent war sweeping Silicon Valley.

Altman described the decision to shut down Sora as a difficult but necessary sacrifice for the company to achieve its larger goals. In a memo to employees, he wrote that he was encouraged to see everyone willing to make "tough trade-offs" for the company's benefit.

Altman

The decision marked a stunning conclusion to a project Altman had once pinned high hopes on. He had dreamed of using Sora to position OpenAI as the creative pioneer of the AI era and bring in substantial new revenue streams for the company.

Two years ago, OpenAI first showcased Sora to the world, presenting dreamlike landscapes generated by the technology, reminiscent of Hayao Miyazaki's fantasy worlds or the surrealist paintings of Salvador Dalí. When OpenAI launched the standalone Sora app for consumers last September, Altman compared it to the company's initial release of ChatGPT.

However, the app never became as popular as its developers envisioned. It was more akin to low-quality AI-generated content than an AI creative marvel. By the end of last year, its usage had plateaued.

Shortly after its launch, the global user base peaked at around 1 million but never reached that level again. According to data from AI research firm Similarweb, user numbers dropped below 500,000 in the following months.

A person familiar with the matter revealed that Sora was losing approximately $1 million per day. As OpenAI tightened its purse strings ahead of the IPO, company executives began scrutinizing Sora more strictly, and the results were not to their satisfaction.

At the time, OpenAI's research team was about to begin training a new model designed to power video generation features for ChatGPT. Unlike language models that learn from text, video models need to understand the entire dynamic world, making their creation significantly more expensive. After carefully calculating the required costs, OpenAI decided to cancel Sora.

OpenAI plans to shift its focus to a new type of "super app" it is building, which will integrate so-called AI agent tools capable of autonomously performing tasks for users, such as writing software, analyzing data, and booking travel. These productivity-focused products are gradually gaining widespread adoption in the labor market, and OpenAI currently lags behind competitor Anthropic in the race for this market, posing a threat to its leading position in the AI competition.

Altman told employees that the Sora team would next focus on longer-term strategic initiatives like robotics.

An OpenAI spokesperson said the company is rigorously prioritizing computational resources based on the areas that can deliver the greatest long-term economic value. She stated: "This careful focus on where we deploy compute allows us to achieve growth, accelerate innovation, and serve businesses and developers more efficiently."

Zuckerberg's Talent Poaching

Sora was the brainchild of Tim Brooks and Bill Peebles. The two researchers became close friends while pursuing their PhDs at UC Berkeley. They joined OpenAI in early 2023 with the goal of building models capable of simulating the physical world by generating high-quality video from text.

Sora was part of OpenAI's World Simulations team, led by Aditya Ramesh. This department operated independently from OpenAI's core research team, which builds the large language models powering ChatGPT.

Last spring, Meta CEO Mark Zuckerberg launched a full-scale talent war against OpenAI, personally contacting dozens of OpenAI's top researchers with lucrative compensation packages to lure them to his new AI lab. One of his targets was Peebles, who received an offer and briefly considered joining Meta.

According to people familiar with the matter, OpenAI managed to retain Peebles by giving him a raise. Shortly thereafter, his responsibilities on the Sora project expanded. Peebles oversaw the training of a new version of the video generation model and the development of the Sora consumer app.

Disney's AI Dream

Despite Sora's daily losses of $1 million, OpenAI had tried to find a viable path for it. Last December, it announced a multi-year agreement with Disney, securing licensing rights to use over 200 characters from the entertainment giant's film library. As part of the deal, Disney agreed to become a major customer of OpenAI and invest $1 billion in it.

Iger, in an interview with CNBC, said the deal gave Disney an opportunity to participate in the rapid development of AI and new forms of media entertainment. Altman said he hoped the collaboration would offer users a new way to engage in creative expression with AI.

Then-Disney CEO Iger

For Disney, the deal demonstrated a viable business model for AI licensing around its intellectual property. Just a day before announcing the agreement with OpenAI, Disney had sent a cease-and-desist letter to Google, accusing the tech giant of "massive copyright infringement of Disney's works."

Cloud Service Solution

In February, Iger said on an earnings call that short videos generated by Sora would soon appear on the Disney+ streaming platform, which was preparing to launch a vertical video feed. According to people familiar with the matter, Disney was also in talks with OpenAI about using ChatGPT company-wide.

In recent weeks, OpenAI began piloting an enterprise version of Sora, allowing companies like Disney to use the tool with security guarantees. Disney had expected to roll out this tool as early as this spring. It would have allowed select Disney executives to use Sora for everything from marketing campaign design to special effects production, without OpenAI having access to their work.

However, OpenAI was already considering how to withdraw from the Sora project. The gap with competitor Anthropic widened further. Anthropic's recent technical progress reignited concerns that AI could replace traditional software and services, briefly causing a dip in software stocks. OpenAI realized it needed to devote more resources to building these so-called productivity tools and began deprioritizing certain areas.

After initially planning to continue offering video generation features through ChatGPT, the company ultimately decided to shut down Sora completely.

Disney's planned $1 billion investment in OpenAI ultimately did not materialize, and the relationship between the two is effectively stalled.

According to a person familiar with the matter, under new CEO Josh D'Amaro, Disney is actively discussing the application of other AI tools with more than a dozen partners.

Real-time Information Push

Disney said in a statement: "As the emerging AI field develops rapidly, we respect OpenAI's decision to exit the video generation business and realign its priorities. We appreciate the constructive collaboration between our teams and the lessons learned from it."

On Sora's X account, its team posted a message reminiscent of a digital obituary: "To everyone who ever created, shared, and built a community around Sora: thank you."

İlgili Sorular

QWhat were the main reasons behind OpenAI's decision to shut down Sora?

AOpenAI shut down Sora primarily due to its excessive consumption of limited and expensive AI computing resources (chips), its failure to become profitable (reportedly losing about $1 million daily), and a strategic shift in company priorities. The resources were needed for the upcoming 'Spud' AI model, coding tools, and enterprise products, especially as OpenAI focused on its IPO preparation and fell behind competitors like Anthropic in productivity-focused AI agent tools.

QHow did the user base of the Sora consumer app perform after its launch?

AThe Sora consumer app initially reached a peak of around 1 million users shortly after its launch. However, its user base subsequently declined and never reached that peak again, dropping to under 500,000 users in the following months, according to data from AI research firm Similarweb.

QWhat was the significance of the deal between OpenAI and Disney regarding Sora?

AThe deal was significant as it involved Disney investment of $1 billion in OpenAI and granted OpenAI a license to use over 200 characters from Disney's film library (including Marvel and Pixar) for Sora-generated videos. It was seen as a major endorsement from a creative industry leader and an attempt to establish a viable business model for AI-generated content. However, the investment ultimately did not materialize after Sora was shut down.

QWho were the key researchers behind the Sora project, and what external challenge did they face?

AThe Sora project was the brainchild of researchers Tim Brooks and Bill Peebles. A key external challenge was a talent war from Meta CEO Mark Zuckerberg, who personally contacted dozens of OpenAI's top researchers, including Peebles, offering lucrative compensation to join Meta's AI lab. OpenAI managed to retain Peebles by giving him a raise.

QWhat is OpenAI's new strategic focus after shutting down the Sora project?

AOpenAI is shifting its strategic focus towards developing a new 'super app' that integrates AI agent tools capable of autonomously performing tasks for users, such as writing code, analyzing data, and booking travel. The company is also prioritizing resources for its upcoming 'Spud' AI model and enterprise products, areas where it feels it is lagging behind competitors and which offer greater long-term economic value.

İlgili Okumalar

Morning Post | Trump Media Group Releases Q1 Financial Report; Top Three DeFi Applications Return Nearly $100 Million in Revenue to Token Holders in 30 Days; Michael Saylor Shares Bitcoin Tracker Info Again

**Title: Daily Briefing | Trump Media Group Releases Q1 Report; Top 3 DeFi Apps Return Nearly $100M to Token Holders; Michael Saylor Signals Potential Bitcoin Buy** **Summary:** Key developments in the past 24 hours include: * **Economic Outlook:** Goldman Sachs has pushed back its forecast for the next two Federal Reserve interest rate cuts to December 2026 and March 2027, citing persistent inflationary pressures from energy costs. This delayed timeline is expected to tighten liquidity flow into risk assets, including cryptocurrencies. * **DeFi & Revenue:** Data from DefiLlama shows that three leading DeFi applications—Hyperliquid, Pump.fun, and EdgeX—collectively distributed $96.3 million in revenue to their token holders over the last 30 days. This trend highlights a shift in the crypto community's focus towards real protocol earnings and sustainable economic models. * **Corporate Bitcoin Moves:** Michael Saylor, founder of MicroStrategy (note: referred to as 'Strategy' in the text, likely a typographical error), has signaled potential upcoming Bitcoin purchases by posting a "Bitcoin Tracker" update, following a pattern that typically precedes the company's official disclosure of new acquisitions. * **Market Integrity:** Prediction market platform Polymarket announced updates to address platform issues, including identifying and banning clusters of accounts involved in "ghost-fill" activities and implementing measures to prevent bulk account creation. * **Regulation:** The Bank of England Governor warned that stablecoin regulation could lead to tensions between US and international regulators. In South Korea, the National Tax Service has launched a pilot program to entrust seized virtual assets to private custody firms for management. * **Meme Token Trends:** GMGN data lists the top trending meme tokens on Ethereum (e.g., HEX, SHIB), Solana (e.g., FWOG, TROLL), and Base (e.g., SKITTEN, PEPE) over the past day. **Financial Note:** Trump Media & Technology Group reported a Q1 loss of approximately $4 billion, primarily attributed to unrealized losses on its Bitcoin and other digital asset holdings.

链捕手19 dk önce

Morning Post | Trump Media Group Releases Q1 Financial Report; Top Three DeFi Applications Return Nearly $100 Million in Revenue to Token Holders in 30 Days; Michael Saylor Shares Bitcoin Tracker Info Again

链捕手19 dk önce

Telegram Takes Direct Control of TON, Social Traffic Rewrites the Public Chain Narrative

Telegram founder Pavel Durov announced that Telegram will replace the TON Foundation as the core driver and largest validator of The Open Network (TON). Key initiatives include a sixfold reduction in transaction fees, performance upgrades, and improved developer tools within the next few weeks. This marks a strategic shift from Telegram merely providing user access to deeply integrating TON into its platform's core infrastructure. The goal is to transform Telegram's massive social traffic into sustainable on-chain activity. While viral mini-apps like Notcoin have demonstrated Telegram's ability to drive user adoption, TON aims to support frequent, low-value transactions inherent to social platforms—such as tipping, in-app payments, and game rewards. Ultra-low fees and sub-second finality (0.6 seconds) are crucial to making blockchain interactions seamless and nearly invisible within the Telegram user experience. However, Telegram's increased central role raises questions about network decentralization. Durov argues that Telegram's participation will attract more large validators, thereby enhancing decentralization. TON also offers high annual staking rewards (18.8%), aiming to retain capital within its ecosystem. The fundamental challenge for TON is no longer leveraging Telegram's user base, but becoming an indispensable, seamless infrastructure layer for Telegram's everyday applications—moving from an adjacent chain to an embedded utility.

marsbit20 dk önce

Telegram Takes Direct Control of TON, Social Traffic Rewrites the Public Chain Narrative

marsbit20 dk önce

Telegram Takes Direct Control of TON, Social Traffic Reshapes Public Chain Narrative

Telegram's founder, Pavel Durov, has announced a major shift in the development of The Open Network (TON). Telegram will now become the core driver of TON, replacing the TON Foundation and becoming its largest validator. The focus will be on technical upgrades over the next few weeks, including slashing network fees by six times to near-zero and improving finality time to 0.6 seconds. This move signifies a deeper integration between Telegram and TON, moving beyond just providing a user base. The goal is to transform Telegram's vast social traffic and built-in features—like Mini Apps, payments, and bots—into sustainable, on-chain usage scenarios. The reduced fees and faster speeds are crucial for enabling the small, frequent transactions typical of social interactions. While this promises stronger execution and product alignment, it raises questions about centralization. Durov argues Telegram's involvement will attract more validators, enhancing decentralization, but the outcome remains to be seen. Additionally, TON's high annual staking reward of 18.8% aims to retain capital within the ecosystem. The key challenge for TON is no longer just leveraging Telegram's entry point, but becoming an invisible, seamless infrastructure layer within Telegram's daily use. Its success hinges on converting viral attention into lasting, embedded utility.

Odaily星球日报30 dk önce

Telegram Takes Direct Control of TON, Social Traffic Reshapes Public Chain Narrative

Odaily星球日报30 dk önce

OpenAI Post-Training Engineer Weng Jiayi Proposes a New Paradigm Hypothesis for Agentic AI

OpenAI engineer Weng Jiayi's "Heuristic Learning" experiments propose a new paradigm for Agentic AI, suggesting that intelligent agents can improve not just by training neural networks, but also by autonomously writing and refining code based on environmental feedback. In the experiment, a coding agent (powered by Codex) was tasked with developing and maintaining a programmatic strategy for the Atari game Breakout. Starting from a basic prompt, the agent iteratively wrote code, ran the game, analyzed logs and video replays to identify failures, and then modified the code. Through this engineering loop of "code-run-debug-update," it evolved a pure Python heuristic strategy that achieved a perfect score of 864 in Breakout and performed competitively with deep reinforcement learning (RL) algorithms in MuJoCo control tasks like Ant and HalfCheetah. This approach, termed Heuristic Learning (HL), contrasts with Deep RL. In HL, experience is captured in readable, modifiable code, tests, logs, and configurations—a software system—rather than being encoded solely into opaque neural network weights. This offers potential advantages in explainability, auditability for safety-critical applications, easier integration of regression tests to combat catastrophic forgetting, and more efficient sample use in early learning stages, as demonstrated in broader tests on 57 Atari games. However, the blog acknowledges clear limitations. Programmatic strategies struggle with tasks requiring long-horizon planning or complex perception (e.g., Montezuma's Revenge), areas where neural networks excel. The future vision is a hybrid architecture: specialized neural networks for fast perception (System 1), HL systems for rules, safety, and local recovery (also System 1), and LLM agents providing high-level feedback and learning from the HL system's data (System 2). The core proposition is that in the era of capable coding agents, a significant portion of an AI's learned experience could be maintained as an auditable, evolving software system.

marsbit1 saat önce

OpenAI Post-Training Engineer Weng Jiayi Proposes a New Paradigm Hypothesis for Agentic AI

marsbit1 saat önce

Your Claude Will Dream Tonight, Don't Disturb It

This article explores the recent phenomenon of AI companies increasingly using anthropomorphic language—like "thinking," "memory," "hallucination," and now "dreaming"—to describe machine learning processes. Focusing on Anthropic's newly announced "Dreaming" feature for its Claude Agent platform, the piece explains that this function is essentially an automated, offline batch processing of an agent's operational logs. It analyzes past task sessions to identify patterns, optimize future actions, and consolidate learnings into a persistent memory system, akin to a form of reinforcement learning and self-correction. The article draws parallels to similar features in other AI agent systems like Hermes Agent and OpenClaw, which also implement mechanisms for reviewing historical data, extracting reusable "skills," and strengthening long-term memory. It notes a key difference from human dreaming: these AI "dreams" still consume computational resources and user tokens. Further context is provided by discussing the technical challenges of managing AI "memory" or context, highlighting the computational expense of large context windows and innovations like Subquadratic's new model claiming drastically longer contexts. The core critique argues that this strategic use of human-centric vocabulary does more than market products; it subtly reshapes user perception. By framing algorithms with terms associated with consciousness, companies blur the line between tool and autonomous entity. This linguistic shift can influence user expectations, tolerance for errors, and even perceptions of responsibility when systems fail, potentially diverting scrutiny from the companies and engineers behind the technology. The article concludes by speculating that terms like "daydreaming" for predictive task simulation might be next, continuing this trend of embedding the idea of an "inner life" into computational processes.

marsbit1 saat önce

Your Claude Will Dream Tonight, Don't Disturb It

marsbit1 saat önce

İşlemler

Spot
Futures
活动图片