Nvidia's Huang says AI computing demand is up 'substantially' in the last 6 months (www.cnbc.com)

Nvidia CEO Jensen Huang told CNBC that AI computing demand has risen “substantially” over the past six months as models move from simple Q&A to complex reasoning. He argued that reasoning-capable models drive exponentially higher compute needs and, because they work so well, exponentially higher user demand — “two exponentials happening at the same time.” Huang singled out Nvidia’s next‑gen Blackwell chips as being in very high demand and characterized the moment as the start of a major infrastructure buildout, likening it to a new industrial revolution. His comments coincided with a premarket uptick in Nvidia shares. For the AI/ML community this signals a shift from model innovation to large‑scale deployment and infrastructure pressure: training and inference for reasoning models require far more FLOPs, memory bandwidth, and specialized accelerators, stressing data‑center capacity, supply chains, power/thermal design, and cloud economics. The implied technical implications are clear — accelerated rollout of high‑performance GPUs (e.g., Blackwell), expanded datacenter investment, tighter competition for silicon and racks, and rising importance of system‑level optimizations (model parallelism, quantization, memory hierarchies) to control cost and energy. Developers and ops teams should anticipate heavier resource requirements and prioritize efficiency and scaling strategies.