At the World Economic Forum in Davos, OpenAI CEO Sam Altman issued a frank assessment on the path to superintelligent AI, emphasizing that without significant breakthroughs in energy technology, progress will stall despite ongoing improvements in AI capabilities.
- AI data centers projected to consume 945 TWh annually by 2030
- Energy shortages seen as a major bottleneck in AI scalability
- Fusion energy breakthroughs may be critical for future AI growth
What happened
OpenAI CEO Sam Altman spoke candidly about the challenges facing AI development during a Bloomberg-hosted session at the 2026 World Economic Forum in Davos. He pointed out that while AI models have improved significantly, reaching superintelligent AI requires breakthroughs beyond incremental advances, especially in energy production.
Altman highlighted the extraordinary energy consumption associated with expanding AI infrastructure, referencing estimates from the International Energy Agency that AI data centers could use nearly 3% of global electricity by 2030. The current lack of sufficient energy infrastructure poses a significant barrier to scaling AI systems further.
Why it matters
AI development has historically been constrained by technological and resource limitations, with energy availability now emerging as a leading bottleneck. As models grow larger and more complex, the demand for electricity intensifies, threatening to outpace supply and slow progress.
Without breakthroughs in renewable or advanced energy solutions like fusion, the AI sector may struggle to sustain growth or improve model capabilities meaningfully. This energy challenge impacts AI companies' ability to build new data centers and operate them efficiently, potentially delaying next-generation innovations.
What to watch next
Observers should track advancements in fusion energy and other clean power technologies as potential enablers for future AI scalability. These breakthroughs could supply the vast and reliable energy needed to support ever-larger AI models and more widespread deployment.
Additionally, improvements in data center efficiency, including better interconnects, optical components, and memory technology, will play a crucial role in mitigating energy constraints. Developments in these areas will determine how quickly and sustainably the path to superintelligent AI can progress.