[Insights] Sam Altman’s Next Move: AGI, Infrastructure, and the Era of Human-Centered AI

From a Tech Company to an Infrastructure for Humanity

It has been two years since ChatGPT changed how we think about intelligence and communication. Yet Sam Altman is already focused on what comes next. In a recent interview, the OpenAI CEO described his company not as a product maker but as a multi-layered infrastructure enterprise designed to support a future where everyone will have a personal AI.

OpenAI now operates across four dimensions: consumer services like ChatGPT, a massive computing infrastructure, a research lab, and emerging ventures in hardware and commerce. These layers are converging toward a single vision: enabling each individual to own an AI companion that understands their context, learns from their behavior, and becomes genuinely useful in daily life.

To realize that vision, OpenAI is becoming increasingly vertically integrated. Altman admitted that he once believed in specialization, but now he insists that building such an ecosystem requires direct control over infrastructure, training, and deployment. “To achieve our mission, we have to do more ourselves,” he said.

The Road to AGI: Sora and the Next Interface

To many, Sora might look like just a text-to-video model. Altman sees something deeper. He believes it represents the foundation of a world model, a system that understands the physical and emotional logic of the real world, not just words on a screen.

AGI will require this kind of multimodal understanding. Video data, Altman argues, contains the subtle dynamics of cause, emotion, and consequence that text alone cannot capture. By training on such data, AI learns not only to generate visuals but to reason about the world.

Just as ChatGPT helped society learn how to communicate with AI, Sora helps society confront and understand the coming age of synthetic media. Because video carries far more emotional resonance than text, people need to experience it firsthand to build intuition and ethics around it.

Beyond video, Altman foresees a new era of interfaces. Conversation is no longer enough. The next generation will feature real-time rendered environments and ambiently aware devices that anticipate users’ needs instead of interrupting them.

He is especially excited about the emergence of the AI Scientist, models that can autonomously perform experiments, analyze data, and make small scientific discoveries. Early signs are already appearing in GPT-5, and he expects major breakthroughs within two years.

Technology, Personalization, and Leadership

Altman admits that AI’s progress has far exceeded public perception. What he calls “capability overhang” is the widening gap between what current models can actually do and what most people think they can do. This gap is shaping both opportunity and unease across society.

He also acknowledges that people do not want the same AI personality. “It was naive to think everyone would enjoy talking to the same assistant,” he said. The future lies in personalization, AIs that adapt to each person’s tone, preferences, and emotional patterns. Some users will define these traits themselves, while others will let the model learn and evolve organically.

As a leader, Altman has also evolved. Once a venture investor focused on deals and vision, he now sees himself as an operator who must execute at scale. Partnering with Microsoft, Nvidia, AMD, and Oracle, he is making bold infrastructure investments to ensure OpenAI’s long-term capacity to train frontier models. This shift reflects a deeper lesson: without operational discipline, even the most ambitious vision collapses.

AI and Society: Safety, Copyright, and the Future of Creation

Altman does not downplay the risks. He believes some really bad things will inevitably happen as AI becomes more powerful, just as fire once caused destruction before it transformed civilization. The key is not avoidance but adaptation, developing societal norms and guardrails that evolve alongside the technology.

On copyright, he expects AI training to be treated as fair use, similar to how humans read and learn from existing works. Still, he foresees the rise of new models of ownership, especially when AI reproduces specific artistic styles or intellectual properties.

The bigger concern, he says, is incentive collapse. If AI systems can summarize everything without sending users to the original sources, who will keep creating? His answer is that creation will thrive as long as there remains a fair system of recognition and reward through attention, credit, or revenue. “If AI makes creation easier and fairer,” he argues, “we will see more content, not less.”

Energy and the Economics of Trust

AI and energy, Altman believes, are converging into a single challenge. The world’s computational demand is rising so fast that the future of AI depends directly on breakthroughs in power generation and storage.

In the short term, he expects natural gas to remain the main baseload in the United States. But over the long run, two technologies will dominate: solar plus storage and nuclear energy. If nuclear power becomes cheap enough, he predicts, political resistance will eventually yield to economic logic.

When it comes to monetization, OpenAI is still exploring. Sora’s high computational cost might require a per-generation pricing model instead of a flat subscription. Advertising, however, is unlikely. “ChatGPT is built on trust,” Altman said. “If it ever recommends something because it was paid to, that trust disappears.”

For him, the most valuable currency in the AI economy is not data or code, it is authenticity. In a world where anything can be generated, credibility itself becomes the scarcest resource.

Beyond Technology: AI as a Relationship

Altman’s interview was not about machines. It was about humanity. OpenAI is evolving from a lab that builds models into a company building the infrastructure for human-machine coexistence.

AGI, in his view, is not designed to replace people but to grow alongside them, to amplify curiosity, creativity, and understanding of the world.

“The future of AI,” he suggests, “is not about intelligence alone. It is about relationship.”

*references: a16z