Altman announces OpenAI's "small goal": invest one trillion in infrastructure per year and achieve fully autonomous "AI researchers" by 2028.

Through a new agreement with Microsoft, OpenAI has completed its controversial profit-oriented restructuring. In a livestream on Tuesday the 28th, CEO Sam Altman announced the completion of the restructuring and revealed that the company plans to greatly expand its infrastructure investment scale, with the ultimate goal of investing $1 trillion annually in infrastructure. Altman disclosed that OpenAI has so far committed about $1.4 trillion for infrastructure construction, equivalent to about 30 gigawatts (GW) of data center capacity. This total includes previously announced deals with partners such as AMD, Broadcom, Nvidia, Oracle, and others. Altman said this is only the starting point. In the long term, OpenAI hopes to establish a technological and financial system capable of adding new capacity at a rate of 1 GW per week, with each GW costing about $20 billion. Altman admitted that this means OpenAI will eventually need to achieve annual revenues of hundreds of billions of dollars. Altman said that OpenAI is on a steep growth curve towards this goal, still facing technical and financial obstacles. Prior to Altman’s remarks, OpenAI had just finalized a major agreement with Microsoft, pushing forward its restructuring into a profit-oriented organization. Under the new structure, OpenAI’s profit-oriented division is now a Public Benefit Corporation called OpenAI Group PBC, and the nonprofit holding the organization’s shares has been renamed the OpenAI Foundation, which holds about 26% equity in the PBC, currently valued at about $130 billion. Infrastructure Investment Plan In Tuesday’s livestream, Altman elaborated on OpenAI’s infrastructure expansion plan. He said the company’s committed $1.4 trillion infrastructure investment will be deployed over the next few years, a scale that clarifies many of the company’s previous announcements regarding chip, data center, and financing partnerships. However, Altman admitted there are challenges in achieving the goal of adding 1 GW of capacity per week. “This is the target we want to reach, and over the coming months we’ll be doing a ton of work to see if it’s possible,” he said. To support such huge capital expenditures, OpenAI needs to significantly increase its revenue scale. In the Q&A session, Altman said enterprise customers will be an important source of revenue, but he also sees ways to get funding from the consumer side, not limited to monthly paid subscriptions. “Ultimately we need to reach annual revenue in the hundreds of billions of dollars, and we’re on a steep growth curve towards that goal,” Altman added. As for whether OpenAI plans to go public, Altman said an IPO will very likely happen eventually, but there are currently no specific plans or timetable. “Given our capital needs and the size of the company, this is probably our most likely path,” he said. AI Research Capabilities Timeline In the same livestream, Altman also disclosed OpenAI’s aggressive timeline for AI research capabilities. According to internal progress tracking, OpenAI is expected to achieve intern-level research assistants by September 2026 and fully automated “real AI researchers” by 2028. OpenAI Chief Scientist Jakub Pachocki described this AI researcher as “a system capable of independently completing large-scale research projects.” Pachocki added, “We believe that deep learning systems may be less than a decade away from superintelligence.” He defined superintelligence as systems that surpass humans in a wide array of key actions. To achieve these goals, OpenAI is betting on two key strategies: continued algorithm innovations and a dramatic expansion of ‘test-time compute’—the amount of time a model spends thinking on a problem. Pachocki said current models can handle tasks spanning about five hours and achieve top human performer levels in competitions like the International Math Olympiad. But he believes this time span will rapidly lengthen, partly by allowing the model to devote more compute resources to thinking through complex problems. For major scientific breakthroughs, he said it’s worth dedicating an entire data center’s compute capacity to a single problem. Restructuring Completion and New Microsoft Agreement OpenAI’s profit-oriented restructuring is finally complete after over a year of negotiations. The company’s profit-oriented division is now a Public Benefit Corporation called OpenAI Group PBC; the nonprofit branch is renamed the OpenAI Foundation, which holds equity in the profit-oriented division currently valued around $130 billion. The nonprofit foundation will start operating with a $25 billion focus on healthcare and disease projects and an “AI resilience” project. When OpenAI’s profit-oriented division reaches unspecified valuation milestones, the nonprofit will receive “additional ownership.” This news comes some time after OpenAI received approval for the restructuring from the California and Delaware Attorneys General. Without these regulatory approvals, OpenAI would have been unable to move forward. The restructuring also endured lengthy legal disputes with OpenAI co-founder and Tesla CEO Elon Musk, who has been trying to block OpenAI’s transition through litigation. As part of the restructuring, Microsoft and OpenAI reached a brand new agreement. Under the new agreement, Microsoft’s stake appears to have declined. Previously, Microsoft held 32.5% of the profit-oriented organization’s conversion basis equity in OpenAI. Under the new deal, including all owners’ conversion-diluted equity, Microsoft will hold about 27% equity, valued at approximately $135 billion. Commentators say the new agreement resolves one of the biggest sticking points in the relationship between OpenAI and Microsoft—Microsoft would lose its rights to OpenAI’s technology after the realization of AGI (Artificial General Intelligence). This so-called 'AGI clause' was previously rather vague, but the new agreement clarifies the situation. The new agreement between the two companies also covers OpenAI’s upcoming consumer hardware plans, and arrangements for what happens if AGI is ultimately achieved. Risk Disclaimer and Limitation Clauses The market involves risk and investment should be undertaken cautiously. This article does not constitute personal investment advice, nor does it consider the individual investment objectives, financial situation, or needs of any specific user. Users should consider whether any opinions, views, or conclusions in the article are suitable for their particular situation. Investing based on this information is at your own risk.