OpenAI "abandons" Sora and turns to AI agents and new models, Altman focuses on "fundraising, supply chain and large-scale construction of data centers"

He Hao, Dong Jing

Source: Wall Street CN

OpenAI has announced it is shutting down its AI video platform Sora, putting an end to a planned $1 billion investment from Disney. Sam Altman will step back from direct oversight of the company’s safety team to focus on fundraising and large-scale data center construction. Meanwhile, pre-training on its next flagship model “Spud” has been completed, with a launch expected “within weeks” — billed as a model that “could truly accelerate economic growth.” Analysts say this strategic pullback represents a targeted reprioritization as OpenAI gears up for a potential IPO.

OpenAI is undergoing a deep restructuring of its product lineup and management structure. The company has announced the shutdown of Sora, its AI video generation platform, to reallocate computing resources to its next-generation AI models and agent-based products. At the same time, Sam Altman has stepped down from direct oversight of the safety team, shifting his focus to fundraising, supply chain management, and “building data centers at an unprecedented scale.”

In an internal memo to staff on Tuesday, Altman announced that the company had finished pre-training its next flagship AI model, “Spud,” with a launch expected “within weeks,” describing the model as one that “could truly accelerate economic development.” The shutdown of Sora has directly led Disney to abandon its planned $1 billion investment, a deal that never formally closed.

The reshuffle is seen as a key signal that OpenAI is narrowing its focus and doubling down on core commercial value ahead of a potential IPO as early as the fourth quarter of this year. Valued at $730 billion, the company is locked in fierce competition with Anthropic for enterprise clients and investor attention.

### Shuttering Sora: Computing Reallocated, Disney Deal Collapses

On Tuesday, OpenAI’s Sora team released a statement saying it would be “saying goodbye to Sora.”

The news sparked intense discussion on X, with some users questioning why anyone should trust or invest in OpenAI’s consumer products if they could be discontinued just months after launch.

Others remained calm, noting that strong alternatives such as Seedance 2.0 and Kling 3.0 are already available.

Sora officially launched as a standalone app last September and quickly topped the Apple App Store charts upon release, but its ranking steadily declined afterward.

According to tech media outlet The Information, Sora was long viewed internally as a drain on computing resources — an especially pressing issue amid intensifying competition with Anthropic and Google.

In his memo, Altman announced that OpenAI would fully discontinue the entire Sora video product line, including the consumer-facing standalone app, developer APIs, and video features originally planned for integration into ChatGPT.

An OpenAI spokesperson stated:

“As we focus on our core priorities amid growing compute demand, the Sora research team will concentrate on world simulation research to advance robotics and help people solve physical tasks in the real world.”

Notably, the shutdown of Sora immediately derailed a partnership with Disney. Last December, Disney announced a planned $1 billion investment in OpenAI alongside a three-year licensing agreement that would have allowed users to generate video content using more than 200 classic Disney characters within Sora.

Citing people familiar with the matter, The Information reported the deal was never actually finalized, and the investment will not proceed now that OpenAI is exiting the video generation business. A Disney spokesperson said:

“As the early-stage AI industry evolves rapidly, we respect OpenAI’s decision to exit video generation and refocus on other priorities.”

Bill Peebles, head of the Sora product team, wrote in an internal Slack message that Sora’s research direction would shift toward “systems that deeply understand the world by learning to simulate arbitrary environments with high fidelity” — so-called “world models” — with the ultimate goal of “automating the physical economy.”

Altman similarly noted in his memo that the Sora research team would “prioritize long-term world simulation research, particularly in areas related to robotics.”

### Altman Steps Back from Safety Oversight, Focuses on Fundraising and Infrastructure

One of the most strategically significant changes in the restructuring is Altman’s decision to relinquish direct oversight of the safety and security teams.

According to The Information, Altman clearly stated in the memo that he would no longer directly supervise the company’s safety and security teams, allowing him to concentrate on “fundraising, supply chains, and building data centers at an unprecedented scale.”

Under the organizational overhaul, the safety team will report to the research division led by Chief Research Officer Mark Chen, while the security team will move to the “scaling” division headed by co-founder and President Greg Brockman.

Bloomberg reported that Altman also noted in the memo that the move was intended to better integrate safety and security work into the development process while freeing up more time for capital raising and infrastructure projects.

Altman did not elaborate on what “supply chain” efforts referred to. The Information previously reported that OpenAI has been developing custom AI server chips and negotiating to build its own data centers and secure power sources, though it has recently shifted focus to leasing more cloud server capacity.

Additionally, Altman announced that the product division led by executive Fidji Simo would be renamed “AGI Deployment,” aligning it directly with the long-term goal of artificial general intelligence.

### New Model Spud and “Super App”: Betting on Agent Capabilities

Even as it winds down Sora, OpenAI is channeling freed-up computing resources into its new model Spud.

In his memo, Altman said the company had completed pre-training on Spud and planned to launch it “within weeks,” calling it a “very powerful model” that the team believes “can truly accelerate economic growth.”

“Things are moving faster than many of us expected,” he added.

On the product side, OpenAI announced last week that it would merge the ChatGPT desktop app, its Codex programming tool, and the Atlas browser into a single “super app,” aiming to unify employees around a cohesive product vision.

Bloomberg reported that this integration aligns closely with the company’s strategy to strengthen “agentic capabilities.”

Agentic systems refer to AI software that can operate autonomously on users’ devices to perform tasks including coding and data analysis.

At an all-hands meeting earlier this month, Fidji Simo, head of apps, made clear that staff should no longer be distracted by “side projects” and identified building agent capabilities as the top priority.

### Behind the Strategic Pullback: Prioritization Before IPO

The overhaul comes as OpenAI systematically reviews product priorities ahead of a potential IPO.

Internally, according to The Information, shutting down Sora is widely viewed as the first step in eliminating “side projects.”

Since its launch, Sora has faced multiple criticisms. It initially lacked copyright protection mechanisms, sparking brief disputes over intellectual property before controls were added. Some insiders expressed surprise at the large amount of computing power poured into the product amid uncertain demand.

The shutdown is also seen as a correction to OpenAI’s earlier product expansion strategy, in which it pursued multiple lines simultaneously, leading to a complex organizational structure and competing resource priorities.

OpenAI now faces direct competitive pressure from Anthropic.

Anthropic has gained significant ground in the enterprise market with AI agents built for programming and white-collar work, and Altman has publicly stated that OpenAI is accelerating the launch of similar tools. With Spud nearing release, the next phase of the rivalry is set to begin.

### Nonprofit Arm: Plans to Spend $10 Billion in 2026

Alongside the strategic shifts, OpenAI announced that its nonprofit arm, the OpenAI Foundation, plans to spend $10 billion in 2026 on AI-related initiatives, a sharp increase from previous levels.

Spending will focus on two main areas: helping society guard against potential AI risks such as biological threats, and funding projects that use AI to advance life sciences.

In personnel moves, Wojciech Zaremba will lead the AI resilience program, focusing on biosecurity and general safety issues, while Jacob Trefethen will head life science efforts, with an initial focus on using AI to accelerate research into diseases including Alzheimer’s.

As part of a late-2025 restructuring, OpenAI’s nonprofit entity was granted a 26% equity stake in its for-profit arm, worth roughly $130 billion at the time.

The massive expansion in spending also responds to concerns raised by the attorneys general of Delaware and California, both of whom have sought assurances that OpenAI’s vast resources will be used to benefit the public.

---

### Risk Warning and Disclaimer

The market involves risks, and investments require caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are appropriate to their particular circumstances. Any investment made based on this article is at one’s own risk.