NVIDIA stock climbed 56.4% from mid-2025 to mid-2026 as demand shifted from model training to reasoning AI, which can require 100x more compute per task

In February 2025, NVIDIA stated that long-thinking reasoning AI can require 100x more compute per task than simple one-shot answers. By May 2025, the company noted these new AI agents required hundreds to thousands more tokens per task, with Microsoft seeing a five-fold increase in tokens processed in a single quarter.

Blackwell architecture deployments were earmarked for inference (the step where an AI produces an answer) rather than training—a first for a new chip generation. This reflected two compounding forces: companies shipping AI to millions of users, and each query burning far more compute due to reasoning workloads.

Despite a $8 billion revenue loss from new China export controls announced in May 2025, the underlying inference-driven demand was strong enough that the business continued growing, yet options market implied volatility had eased to the 19th percentile of its one-year range, signaling trader complacency.