OpenAI’s Plan To Revive Sora

Remember OpenAI’s video model Sora and all the great Sam Altman memes that came from it? Earlier this year, the ChatGPT-maker canned both the model and the TikTok-like app it powered as part of broader culling of server-intensive “side projects” even as rivals xAI and Google forged ahead with their own video models.

Well, there’s a chance that a version of Sora is coming back from the dead.

A stealthy startup developing AI video for filmmakers, which we revealed last month, is in talks with OpenAI to access Sora, tweak it and sell software or other services using it, according to a person with knowledge of the talks. Hollywood mogul Jeffrey Katzenberg, former Sora research lead Bill Peebles, and former Dropbox CFO Sujay Jaswa co-founded the startup in recent months. (We don’t know the startup’s name).

It’s not clear whether the startup will get access to Sora’s weights, the “settings” that determine how the model responds to queries, but if the deal were to go through, the startup would get deep enough access to the model to heavily customize it, the person said. (AI developers like OpenAI tend to guard their model weights very closely, but in this case, sharing weights or other sensitive information wouldn’t be surprising when you consider that Peebles previously led Sora and already knows how the model works and was trained.)

If a deal gets completed, OpenAI would receive equity in the new startup, this person said, though the amount couldn’t be learned.

Working with Peebles makes a lot of sense, given how much money and time OpenAI spent developing Sora. The cost of making and running Sora was so high that the company decided to shift resources to its traditional large language models and ChatGPT, which are more economically viable.

The new deal means OpenAI can salvage some sort of return on its Sora investment.

OpenAI had abandoned Sora as part of a push to cut side projects and remain focused on ChatGPT and fighting Anthropic in the coding and enterprise fronts. In recent months, there’s been signs that OpenAI has backslid a bit on that push. It has revived efforts in robotics, hired a Patreon co-founder to lead a new “creator product” division, and it is also pushing hard on a costly personal-device initiative.

Here’s what else is going on…

Big Number

OpenAI recently told investors that its annualized revenue was approaching $50 billion at the end of September, according to a person with knowledge of the communications. That’s around $20 billion less than revenue figures disclosed in other recent reports, including The Information's.

Overheard

The U.S. government is limiting Microsoft’s ability to sponsor visas for certain immigrant workers, Vice President J.D. Vance said Thursday, citing an “ongoing investigation” into claims that Microsoft laid off Americans while continuing to hire more immigrants.

Top executives at AI companies including Anthropic and OpenAI are gaming out scenarios for public revolt that could follow a catastrophic AI event, Axios reported.

Anthropic said Friday that its AI used the internet to inadvertently fill out government forms and submit a false homicide crime tip to police. In response, Trump administration officials said they are now mandating AI companies to report and correct security incidents.

The three safety researchers whom OpenAI fired last week challenged OpenAI’s decision, saying on Thursday they were punished for raising safety concerns.

The perpetrator of a recent hack of South Korean financial institutions that resulted in data leaks used an open-source Chinese AI agent called Artex and Chinese large language models, according to CrowdStrike.

Deals and Debuts

See The Information’s Generative AI Database for an exclusive list of private companies and their investors.

Firmus, a Nvidia-backed Australian AI data-center and compute company, withdrew its planned initial public offering at a valuation above $30 billion. It is now exploring a $2 billion to $3 billion private funding round with existing investors instead.

TypeSafe AI, which makes an AI model that outputs decisions instead of text, raised $870 million in a Series A funding round at a $7.5 billion valuation led by Andreessen Horowitz. The Information first reported on the round here.

Amazon is in talks to acquire Israeli AI startup Decart, whose software makes it easier for developers to switch between different AI chips, in a deal that could value it at around $7 billion, the Wall Street Journal reported.

Nuvacore, a six-month-old chip startup with no product yet that's designing a new type of central processor for data centers, is raising hundreds of millions of dollars at a valuation of roughly $2.5 billion, Reuters reported.

GlobalFoundries signed a $2 billion, five-year manufacturing agreement with TSMC to produce silicon interposers — a packaging component used in AI chips — at its Malta, New York facility.

Nvidia committed $1 billion over five years to expand the US's capacity for “super intelligence” research and development, funding work in quantum computing, healthcare and energy at national laboratories and universities.

Arena, formerly known as Chatbot Arena, which runs a leaderboard that ranks AI models, raised $200 million in a Series B funding round co-led by Lightspeed Venture Partners and Khosla Ventures.

Veir, which makes superconducting power-delivery systems for AI data centers, raised $110 million in a Series C funding round co-led by Matter Venture Partners and Tyche Partners.

Ultra, a warehouse-robotics startup that rents out robots on a subscription basis, disclosed $62 million in total funding—a $50 million Series A led by Framework Ventures plus an earlier $12 million seed led by Y Combinator—and deepened its partnership with Physical Intelligence to supply AI software for its robots.

Hone, founded by former OpenAI and Cognition staffers to build AI agents that can run entire business operations rather than single tasks, raised $60 million in seed funding led by Benchmark and Index Ventures.

Sabi, which is building a baseball cap lined with roughly 100,000 non-contact neural sensors that converts brain signals into prompts for AI systems, raised $50 million in seed funding led by Khosla Ventures.

Gallatin AI, which makes an AI-native platform for military supply-chain and logistics planning, raised $50 million in a Series A funding round, with participation from 8VC and Silent Ventures.

KONST, a Taipei-based operator of AI data centers and GPU cloud infrastructure across Asia, raised $30 million in a Series B funding round led by ADATA Technology.

Catalyst, which is building AI agents that turn plain-language investing instructions into trading strategies for retail investors, raised $30 million in a seed funding round led by Sequoia Capital.

Ambi, an app that listens to users to provide insights to them, raised $26 million in funding led by Luminous Ventures.

Rein Security, an AI-native runtime-security platform that monitors enterprise AI agents and blocks their harmful actions, raised $25 million in a Series A funding round co-led by Glilot Capital and Sienna Venture Capital.

Vocca, an AI phone assistant that automates patient calls and appointment scheduling for medical practices, raised $20 million in a Series A funding round led by Norrsken VC.

Nvidia is in early talks to acquire, or invest further in, Reflection AI, a maker of “open-weight” AI models already valued at about $25 billion, the Financial Times reported.

Anthropic has quietly signed a multiyear cloud deal with bitcoin miner IREN, The Information reported, part of a broader push that now spans more than 15 gigawatts of data-center leasing commitments across more than 10 cloud providers. Anthropic also launched new programs to support cybersecurity for critical infrastructure and open-source software.

Thank you for reading the AI Agenda Newsletter! I’d love your feedback, ideas and tips: [email protected].

If you think someone else might enjoy this newsletter, please pass it forward or they can sign up here.