TSMC Needs to Double Production Over the Next Decade Just to Meet NVIDIA’s Demand, Says Jensen Huang, Highlighting the Future Scale of the AI Boom
TSMC’s production lines are already under severe pressure, and based on what NVIDIA’s CEO is saying, the Taiwan chip giant would need to expand production by more than 100% just to meet NVIDIA’s demand alone.
TSMC’s Aggressive CapEx Spending & Fab Buildout Is Dedicated Towards the “World’s Largest Infrastructure” Buildout
NVIDIA’s AI bandwagon is moving at a rapid pace, and with that, the company seems confident about how demand for AI infrastructure will evolve in the coming years. During an interview with local media, Jensen disclosed that TSMC plans to expand production by more than 100% over the next decade. Interestingly, this expansion alone will meet NVIDIA’s demand, suggesting Jensen expects the AI boom to be much longer. TSMC’s upcoming fab projects and CapEX hikes over the quarters are examples of the demand the industry expects.
TSMC’s production capacity may grow by more than 100% in the next ten years, which is a very significant scale expansion, the largest infrastructure investment in human history, and it will have to double just to meet NVIDIA’s demand.
– NVIDIA’s Jensen Huang via UDN
TSMC’s fab expansion plans have grown significantly in the past few quarters, as the company has factored in geopolitical concerns, leading to massive investments in regions like the EU, Japan, and the US. More specifically, TSMC’s plans to build a supply chain in America will involve a $250 billion mega-buildout, including advanced packaging, semiconductors, and R&D centers. TSMC’s Arizona plants are currently shifting production to 3nm, and the Taiwan giant plans to move to A16 moving ahead, while being mindful of the “N-2” policies.
Image Credits: NVIDIA
The more important point here is that, with Grace Blackwell and Vera Rubin alone, NVIDIA takes a large share of TSMC’s production lines, which is why Team Green has become the chip giant’s largest customer, overtaking Apple in just a few years. And, since TSMC has opened up the option of capacity ‘prepayments’, there is no doubt that future production lines coming online will ultimately have a large share allocated towards NVIDIA and other HPC customers, which is another indicator of how aggressive the AI infrastructure buildout currently is.
NVIDIA beats its competitors at scale, which is why rivalry from ASIC manufacturers or AMD isn’t much of a factor for the company. And the second major factor is, of course, access to production lines earlier than others, given Jensen’s relations with Taiwanese suppliers and others around the world.
Follow Wccftech on Google to get more of our news coverage in your feeds.
