Sam Altman, CEO of OpenAI, recently made two moves that say more about the company’s future than any product launch ever could. On March 23, he resigned as president of Helion Energy, the nuclear fusion startup he supported for more than a decade, while OpenAI explores a large-scale energy deal with the company. A day later, OpenAI retired its popular video generation model, Sora, saying it would reallocate resources to its next big language model, codenamed “Spud,” as well as a growing suite of programming agents and enterprise tools.
Together, the decisions signal the CEO removing distractions, tightening oversight and preparing for the upcoming IPO. OpenAI has fueled this shift by strengthening its financial leadership, including hiring Cynthia Gaylor, former CFO of DocuSign, to oversee investor relations.
According to the Information, Altman also gave up direct oversight of OpenAI’s safety and security teams to focus on “capital, supply chains, and building data centers at an unprecedented scale.” The company has consolidated its safety team into the research organization led by chief research officer Mark Chen, while moving security within an “expansion” division run by co-founder and president Greg Brockman. Meanwhile, OpenAI has renamed its product organization — led by Fidji Simo, CEO of Applications — to “AGI Deployment.”
Altman has also reportedly narrowed reporting, reducing the number of executives reporting directly to him, marking another step toward operational clarity ahead of a potential IPO. Reuters first reported in October 2025 that OpenAI could submit the application as early as the second half of 2026, It targets a valuation of up to $1 trillion. In early-stage discussions, the company has offered at least $60 billion in the offering, depending on market conditions and revenue growth.
For much of the past year, Altman has described AI infrastructure in sweeping terms, citing numbers approaching $1 trillion for access to artificial general intelligence (AGI). That rhetoric is now giving way to something more realistic: nearly $600 billion in computing spending by 2030.
Sora, which launched in December 2024, was one of OpenAI’s most compelling demonstrations of how far AI has come. But it also revealed a fundamental obstacle. Forbes estimates that a 10-second Sora video is being created Costs range between $1 and $5 per accountwith higher costs for premium versions. On a large scale, this could exceed $5 billion annually, or approximately $15 million daily.
OpenAI is now shifting resources toward more profitable enterprise products, especially programming agents. Internal communications and recent reports from CNBC and The Information indicate that the company is “aggressively moving” toward enterprise use cases, with ChatGPT evolving from a general-purpose chatbot into a productivity layer — a front-end for software development, business processes, and knowledge work automation.
As OpenAI leans toward enterprise systems, it also faces another constraint: power. Training and running advanced AI models at scale requires huge amounts of electricity, making energy one of the most important bottlenecks in the industry. Altman has long argued that solving AI at scale requires solving the energy problem at scale, and OpenAI’s interest in Helion reflects this reality.
His departure from Helion removes a potential conflict as OpenAI explores a partnership with the merger startup. Reports indicate that the deal could give OpenAI access to up to 12.5% of Helion’s expected production – approximately 5 GW by 2030 and up to 50 GW by 2035.
Helion has not yet begun generating power commercially, but is moving quickly from prototypes to grid-tied fusion plants. The company is currently building Orion, its first commercial facility in Washington state, with development starting in July 2025 with a goal of delivering an initial 50 MW to Microsoft by 2028. If that timeline continues, Orion could serve as a model for expansion to multi-gigawatt fleets over the next decade.
Altman built his reputation on ambitious, system-level thinking, often speaking in abstract terms about the future of intelligence. He now serves as CEO preparing the company for the public markets, where ambition must match discipline and vision must translate into returns. Altman must convince investors that OpenAI can turn unprecedented capital spending into sustainable returns while continuing to pursue the long-term goal of artificial general intelligence (AGI).
