NVIDIA CEO Says TSMC May Need to Double Chip Capacity as AI Demand Skyrockets

NVIDIA’s CEO Jensen Huang has been everywhere in the headlines over the past few days. First came a Wall Street Journal report suggesting internal doubts at NVIDIA over a potential $100 billion investment in OpenAI.

Later, Huang clarified that the figure was never a binding commitment, but rather an open-ended invitation to participate in OpenAI’s fundraising round. That clarification set the tone for a busy weekend, as Huang continued meetings across Taiwan and addressed far bigger issues facing the AI industry.

During his Taiwan visit, Huang turned attention to a growing pressure point that goes well beyond any single deal: chip supply. Speaking with reporters after a private dinner with key supply-chain partners, including TSMC CEO C.C. Wei and Foxconn chairman Young Liu, Huang warned that NVIDIA’s demand for advanced wafers is accelerating at a pace that could fundamentally reshape manufacturing capacity over the next decade.

According to comments reported by the South China Morning Post, Huang said NVIDIA alone may require TSMC to more than double its output over the next ten years. He stressed that AI demand this year is already intense, noting that NVIDIA “needs a lot of wafers,” even as he praised TSMC for executing under extraordinary pressure. The message was blunt: keeping up will not be easy.

Those remarks are similar what TSMC itself has been saying. While the company has acknowledged concerns about a potential AI bubble, it has also repeatedly stated that demand remains overwhelming. Last year, Wei admitted advanced-node capacity was still falling well short of AI needs, estimating supply lagged demand by roughly three times.

Huang also addressed speculation around shifting chip production to the United States. He emphasized that Taiwan’s role remains central, explaining that any overseas expansion would come from new fabs rather than relocating existing capacity. TSMC produced the equivalent of 17 million 12-inch wafers in 2024 and is investing heavily to expand further.

That includes $42 billion in planned spending this year and an accelerated timeline for its Arizona Fab 21 site, now targeting mass production in 2027. Taken together, Huang’s comments paint a clear picture: AI demand isn’t cooling, and supply constraints may define the next phase of the industry.

Article feature image source: NVIDIA