Why Nvidia’s CEO Is Personally Pushing TSMC for More AI Chip Production

Major cloud providers such as Microsoft, Amazon, and other hyperscalers are spending heavily on data centers designed to train and run large AI models. These deployments require thousands of Nvidia GPUs per cluster, dramatically increasing demand for the company’s chips.

Nvidia Is Now TSMC’s Largest Customer

The shift toward AI computing has also transformed TSMC’s customer landscape.

By 2025, Nvidia had overtaken Apple to become the largest customer of TSMC, accounting for roughly 19% of the foundry’s revenue.

That represents a major industry shift. For years, smartphone chips dominated the advanced manufacturing pipeline. Today, AI processors—many designed by Nvidia—have become one of the biggest drivers of leading‑edge semiconductor demand.

This relationship gives Nvidia enormous influence but also deep dependence. If TSMC cannot produce enough wafers or advanced packaging capacity, Nvidia’s shipments and revenue growth can stall even when demand remains extremely strong.

Why Huang Is Pushing for More Capacity

Huang has openly warned that AI demand could stretch the semiconductor supply chain for years.

He has said that Nvidia alone may require far more wafers in the coming years, and that TSMC may need to more than double its manufacturing capacity over the next decade to keep up with demand.

These discussions are particularly important as Nvidia prepares new generations of AI hardware such as its Blackwell and Rubin architectures, which require cutting‑edge manufacturing processes and advanced packaging.

Securing that capacity early matters because many competitors—AMD, hyperscale chip designers, and AI startups—are also competing for the same limited production resources.

The Trillion‑Dollar AI Infrastructure Buildout

Behind Nvidia’s urgency is a much larger economic trend: the global race to build AI infrastructure.

Huang has repeatedly argued that the industry is only at the beginning of a massive investment cycle. According to Nvidia’s projections, global spending on AI infrastructure could reach roughly $3 trillion to $4 trillion over time.

In this view, data centers are evolving into “AI factories” that convert computing power into tokens, predictions, and automated decisions. If that vision holds true, the demand for high‑performance GPUs could remain strong for many years.

The Strategic Stakes

All of this means Nvidia’s future growth depends on two different races happening simultaneously:

• The race to design the most powerful AI chips
• The race to secure enough manufacturing capacity to build them

Nvidia has clearly won the first race so far. But the second—access to scarce semiconductor fabrication and advanced packaging—may determine how quickly the company can convert booming demand into real shipments and revenue.

That’s why Jensen Huang’s visits to Taiwan matter. In the AI era, the most valuable technology company in the world still depends on something very physical: enough factories to make the chips.