NVIDIA’s Jensen Huang on AI Token Use: “It's Fine to Waste a Little Money — Just Don't Waste Time”

Huang’s message

NVIDIA CEO Jensen Huang reportedly used a closed‑door dinner in Taipei to push a blunt point about AI development: it has been reported that he told partners “it's fine to waste a little money — just don't waste time,” a remark interpreted as urging teams to prioritize speed of deployment over squeezing every penny out of token consumption. The comment — short, provocative and aimed at application builders and cloud operators — underscores NVIDIA’s broader sales pitch that faster model iteration and real‑world application matter more than marginal compute cost savings.

The dinner and who showed up

The remark came during the fifth “兆元宴” banquet in Taipei, a high‑level gathering hosted while Huang was visiting the island. Attendees reportedly included leaders from key links in Taiwan’s chip and server ecosystem: TSMC (台积电) chairman Wei Zhejia (魏哲家), Hon Hai/Foxconn (鸿海) chairman Liu Yang‑wei (刘扬伟), Quanta Computer (广达) chairman Lin Bai‑li (林百里), Wistron (纬创) chairman Lin Xian‑ming (林宪铭), Wiwynn (纬颖) chairwoman Hong Li‑ning (洪丽寗), Inventec (英业达) chairman Yeh Li‑cheng (叶力诚), Compal (仁宝) chairman Chen Rui‑cong (陈瑞聪), and Pegatron (和硕) chairman Tung Tzu‑hsien (童子贤). It has been reported that the event was informal — no speeches, no signing ceremonies — and conversations centered on wafer capacity, order delivery and AI trends; an anecdote that went viral saw Huang shrug off his jacket saying “it’s hot,” and Wei joking, “that’s because data centers are hot.”

Why the line matters

For Western readers, the scene is a reminder of two dynamics: Taiwan sits at the heart of the global AI hardware supply chain, and NVIDIA sits at the center of demand for those chips. Huang’s quip about tolerating some cost inefficiency in favor of time is a practical pitch to chipmakers, OEMs and cloud partners to prioritize rapid deployment of AI services — even as compute costs and power draw climb. It also arrives against a fraught geopolitical backdrop: U.S. export controls and broader tech tensions with China complicate how advanced chips and AI systems move around the world. Can close commercial ties — and a shared impatience to build useful AI — outpace policy constraints? For now, NVIDIA and Taiwan’s suppliers appear intent on finding out.