The Oslo Patient.

Key Takeaways

• AI models keep improving, closing US‑China performance gap
• US hosts 5,427 AI data centers, consuming more power than any nation
• TSMC fabricates nearly all leading AI chips, creating a single‑point supply risk
• AI boosts productivity while entry‑level jobs decline across sectors
• Mathematicians solve research problems with AI, surpassing human speed

Pulse Analysis

The 2026 AI Index paints a picture of relentless acceleration in artificial intelligence. Large language models and multimodal systems are not only getting larger but also more capable, erasing the performance gap that once separated U.S. and Chinese research labs. Revenue streams for AI firms are expanding faster than any prior technology wave, yet the metrics and regulatory frameworks designed to monitor progress lag behind, leaving investors and policymakers scrambling for reliable signals.

Infrastructure demands are the hidden cost of this sprint. U.S. data centers now draw an estimated 29.6 gigawatts—enough to power New York State at peak—while OpenAI’s GPT‑4o alone may consume water equivalent to the daily needs of 12 million people. The concentration of chip manufacturing in Taiwan’s TSMC creates a geopolitical choke point, and the fragility of that supply chain could bottleneck future breakthroughs. Energy‑intensive workloads are prompting governments to weigh AI’s economic upside against climate and resource constraints.

Beyond commercial applications, AI is reshaping scientific discovery. Mathematicians have begun using generative models to generate proofs, cutting weeks of work into days. The First Proof competition demonstrated that AI can tackle novel, research‑level questions with a success rate above 50 percent, signaling a shift from experimental to mainstream academic tools. As AI sovereignty becomes a policy priority, nations that can balance hardware security, talent pipelines, and responsible innovation will set the agenda for the next decade of knowledge creation.