OpenAI unveils Spud model and $50 billion Amazon alliance while axing Sora video tool

An internal OpenAI memo recently leaked to the press has sent shockwaves through the artificial intelligence industry, detailing a bold strategic pivot and a new core model codenamed Spud that the company believes will make its entire product suite significantly more capable.[1][2][3] The four-page document, authored by Chief Revenue Officer Denise Dresser, outlines five critical priorities for OpenAI's enterprise business as it moves into what leadership describes as a more mature phase of the AI market. This roadmap signals a definitive shift from providing simple chatbot interfaces toward building a comprehensive platform for autonomous agents, while also revealing deep-seated competitive tensions between OpenAI and its primary rival, Anthropic.[1]

At the center of this new strategy is Spud, a next-generation model that represents the culmination of nearly two years of research breakthroughs.[2] According to the memo, Spud is intended to serve as the intelligence foundation for the next generation of digital work.[1] Early feedback from select customers indicates that the model provides a substantial leap in reasoning capabilities, a deeper understanding of complex human intentions, and a more reliable execution of multi-step dependencies compared to current frontier models. The internal document suggests that Spud will not just be another incremental update but a foundational shift that will eventually underpin a unified super app for desktop users. This app is envisioned as a single entry point that merges the capabilities of ChatGPT, the coding tool Codex, and a proprietary browser, effectively attempting to replace the fractured landscape of individual productivity tools with a cohesive AI-driven operating environment.

To facilitate the massive compute requirements for Spud and its associated enterprise products, OpenAI has reportedly made the difficult decision to discontinue its video generation tool, Sora.[4] Despite the significant public attention and early partnerships associated with Sora, including a highly publicized potential deal with Disney, the memo indicates that the project was too resource-intensive to maintain alongside the company's more critical productivity goals. Internal estimates suggested that Sora was consuming hundreds of millions of dollars in GPU costs while generating comparatively little revenue.[3] By shuttering the project, OpenAI is redirecting its vast compute resources toward Spud and a newly renamed product group called AGI Deployment.[5][3] This group, formerly led by head of applications Fidji Simo, is now tasked with ensuring that OpenAI’s most advanced reasoning models are integrated directly into the core workflows of the world’s largest corporations.

The transition from a product-focused company to a platform-centric one is further solidified by the announcement of Frontier, a new agentic platform detailed in the Dresser memo.[1] OpenAI leadership believes the market has moved beyond simple prompting and now demands autonomous systems that can independently use software tools and operate reliably within professional environments.[1] Frontier is positioned as the default platform for enterprise agents, offering the necessary orchestration, security, and governance layers that businesses require to deploy AI at scale.[1] The memo stresses that OpenAI should no longer be viewed as a collection of separate product lines but as a flywheel where better models drive increased usage, and increased usage leads to deeper integration.[1] This deep integration is intended to make OpenAI’s ecosystem essentially irreplaceable for enterprise clients who are increasingly looking for a full-stack solution that includes the model, the agentic framework, and the deployment runtime.

The document also sheds light on a massive new alliance with Amazon that appears to be straining OpenAI’s long-standing exclusive relationship with Microsoft. According to the leak, OpenAI has secured a fifty-billion-dollar investment from Amazon, which includes a commitment of two gigawatts of power specifically for Trainium-based computing.[6] This move is presented as a necessary step to overcome the capacity bottlenecks that have limited the company's growth. While Microsoft remains a key partner, the memo suggests that the demand for OpenAI’s models via Amazon’s Bedrock platform has been staggering.[6] This shift indicates a growing reliance on diversified infrastructure as the company races to maintain its lead in raw intelligence. Dresser notes in the memo that capacity, rather than customer demand, has become the primary bottleneck for the enterprise business, with nine-figure, multi-year deals becoming increasingly common.[1]

One of the most aggressive sections of the leaked memo involves a direct and blunt attack on the financial reporting practices of Anthropic.[1] Dresser accuses the competitor of significantly overstating its commercial success, claiming that Anthropic’s reported thirty-billion-dollar annualized run rate is inflated by approximately eight billion dollars.[1] The core of the accusation lies in how the two companies account for revenue-sharing arrangements with cloud partners like Amazon and Google.[7][8][1] OpenAI claims that it reports its revenue from the Microsoft partnership on a net basis—counting only its share of the income—while alleging that Anthropic grosses up its cloud reseller revenue to make its top-line numbers appear larger than they are.[7][8] If these claims are accurate, they would suggest that Anthropic’s actual comparable run rate is closer to twenty-two billion dollars, placing it behind OpenAI’s stated twenty-five-billion-dollar performance.[7]

Beyond financial discrepancies, the memo characterizes the rivalry with Anthropic as a fundamental clash of philosophies. Dresser writes that Anthropic has built its brand on a narrative of fear and restriction, suggesting that a small elite should control the trajectory of AI development.[1] In contrast, she positions OpenAI’s strategy as a positive, deployment-focused mission aimed at accelerating the global economy. The memo further claims that Anthropic made a strategic error by failing to secure enough compute power early on, a mistake that is allegedly starting to manifest in its products through higher rates of throttling and less reliable availability for enterprise users.[1] These criticisms come at a time when Anthropic is readying its own major breakthrough, the Claude Mythos model, which internal reports suggest is a step-change in coding and cybersecurity capabilities.[3]

The internal correspondence paints a picture of a company in a state of intense urgency, described by some staffers as a Code Red that began late last year. As competitors like Google and Anthropic have narrowed the gap in model performance, OpenAI is betting everything on the Spud model and the Frontier platform to reclaim its definitive lead. The strategy involves owning the entire stack of human-computer interaction, from the underlying foundation models to the agents that execute tasks and the interfaces through which users interact with the internet. By focusing on a super app that can mediate all digital work, OpenAI hopes to transition from a service provider to an indispensable layer of the modern economy.

Ultimately, the leaked memo underscores the high stakes of the current AI arms race. The decision to sacrifice creative tools like Sora in favor of the reasoning-heavy Spud model reflects a calculated gamble that the future of the industry lies in productivity and agentic autonomy rather than media generation. As OpenAI deepens its ties with Amazon and prepares to launch its most ambitious model yet, the company is signaling to investors and competitors alike that it is no longer content with being the creator of a popular chatbot. It is instead aiming to become the primary platform for the age of artificial general intelligence, using its massive compute advantage and a aggressive new business strategy to consolidate its position at the top of the technology hierarchy. While the validity of the revenue accusations against Anthropic remains unverified by third parties, the mere existence of such blunt internal rhetoric highlights how competitive and volatile the landscape has become as the leading labs vie for dominance in a market that is rapidly maturing.