With a $1.4 trillion infrastructure plan and new corporate structure, OpenAI sets its sights on building a machine capable of making scientific breakthroughs.
From Assistant to Autonomous: The Road to AI Research
OpenAI CEO Sam Altman believes the world could see its first “legitimate AI researcher” — a fully autonomous system capable of conducting complex scientific inquiry — by 2028.
Speaking on a livestream, Altman and Jakub Pachocki, OpenAI’s chief scientist, outlined a bold roadmap:
- Intern-level AI research assistant by September 2026
- Fully capable AI researcher by 2028
- Superintelligence — systems smarter than humans across key tasks — potentially within the decade
“We believe this is possible,” Pachocki said, “with deep learning systems evolving faster than expected.”
A New Corporate Structure for a New AI Era
The announcement coincided with OpenAI’s formal transition to a Public Benefit Corporation (PBC), marking a move away from its non-profit origins.
Key structural updates:
- OpenAI Foundation (non-profit) now owns 26% of the for-profit entity
- The foundation holds a $25B commitment to use AI for solving societal challenges like curing disease
- The PBC model enables larger capital raises to fund infrastructure and R&D while maintaining governance focused on scientific advancement
Altman said the new structure balances aggressive innovation with responsible oversight.
The Technical Push: Smarter Algorithms, More Compute
According to Pachocki, OpenAI’s models can already:
- Match top human performers in academic competitions
- Handle problems with a five-hour reasoning window
But to reach higher levels of intelligence, OpenAI is focused on two levers:
- Algorithmic innovation — new methods to scale intelligence efficiently
- Test-time compute scaling — dedicating far more computational time and energy to complex problems
“For meaningful breakthroughs,” Pachocki noted, “we’re willing to use entire data centers on a single task.”
This shift could stretch an AI’s reasoning horizon from hours to days — or even longer — unlocking capabilities currently inaccessible to humans.
Scientific Ambitions Beyond Human Limits
The long-term goal is to use AI systems not only as tools but as independent researchers that:
- Generate and test new hypotheses
- Explore novel solutions in fields like medicine, physics, energy, and materials science
- Accelerate technological development far beyond human timelines
“We want AI to make scientific discoveries faster than humans ever could,” Altman emphasized.
This reflects OpenAI’s central thesis: AI can be a force multiplier for science, not just software.
Infrastructure at Superintelligent Scale
Supporting these ambitions will require massive infrastructure investment. Altman revealed that OpenAI is committing to:
- 30 gigawatts of compute infrastructure
- A total build-out cost of $1.4 trillion in the coming years
This scale dwarfs previous AI efforts and highlights the magnitude of OpenAI’s vision — AI not just as a product, but as a global utility for knowledge and progress.
The Stakes: Progress and Prudence
Even as OpenAI moves aggressively, Altman emphasized that responsible development remains key.
- The OpenAI Foundation will guide research alignment and safety
- Long-term governance structures aim to prioritize humanity’s interests over profits
- Altman reiterated that breakthroughs in AI must be shared widely, not monopolized
Still, the timeline is ambitious and unprecedented, pushing the frontier of what’s technically — and ethically — possible.
OpenAI CEO Sam Altman says the company is on track to develop a fully autonomous AI researcher by 2028. Backed by a new corporate structure, a $1.4 trillion infrastructure plan, and major advances in compute and model design, the company aims to accelerate scientific discovery through artificial intelligence.
