TSMC may double output in next decade, says Nvidia CEO

Taiwan Semiconductor Manufacturing Co. (TSMC) could increase its production capacity by over 100% in the next decade, according to Nvidia CEO Jensen Huang.

Huang, speaking in Taipei after a dinner with Taiwanese suppliers, said Nvidia’s demand for wafers will more than double in the coming years due to rising AI needs.

He noted that TSMC is actively ramping up production of Nvidia’s latest AI chips, including the Blackwell platform and Vera Rubin series, which involve six advanced designs.

Huang described the scale of expansion as potentially the largest infrastructure build-out in history.

He also highlighted ongoing challenges in memory supply, which is growing at about 100% annually, but demand is increasing faster.

The dinner, attended by TSMC’s chairman C.C. Wei and leaders of major Taiwanese tech firms, was dubbed a “trillion-dollar banquet” for the combined market value of the participants.

Huang is scheduled to leave Taiwan on February 2.

🔗 Source: Focus Taiwan

🧠 Food for thought

Implications, context, and why it matters.

Huang’s wafer demand is only part of the story. Packaging and memory create bottlenecks

Wafer capacity alone does not explain the AI supply crunch.
The tightest spots sit in advanced packaging (the step that combines and connects multiple chips into a single high-performance unit) and High Bandwidth Memory (HBM), a fast type of memory used in top AI chips. Both are needed to turn wafers into working AI chips 1.
Capacity for advanced packaging, including TSMC’s CoWoS (Chip-on-Wafer-on-Substrate, an advanced packaging method used for many AI accelerators), remains a widely cited constraint through at least 2026 1.
Separate estimates put demand for advanced wafers well above supply, though the size of the gap changes by source 1.

Nvidia’s dominance concentrates risk and reshapes the entire tech supply chain

Nvidia’s scale shifts leverage across the supply chain.
Some estimates say Nvidia could lock up a very large share of advanced packaging capacity for 2026. That could widen an “AI haves and have-nots” split where smaller chip designers struggle to secure output 2.
This squeeze also nudges other big tech firms, including cloud providers like Google and Amazon, to design custom AI chips. Cost, control, and performance matter too, beyond place in line at TSMC 3.
Access to advanced manufacturing and packaging now acts like a chokepoint that can reshape competition 1.