NVIDIA has dethroned Apple as TSMC’s largest customer, according to CEO Jensen Huang. This revelation, along with his call to invest “trillions” of dollars in new AI infrastructure, illustrates how the center of gravity of technological hardware It is rapidly moving from mobile devices and PCs to artificial intelligence systems.
Huang made this comment about TSMC during a podcast interview «A Bit Personal with Jodi Shelton». When the host mentioned that Morris Chang, founder of TSMC, remembered that the young Huang promised to become one of the foundry’s major customers, Huang laughed and said: “Morris will be happy to know that NVIDIA is now TSMC’s largest customer.”
It’s a symbolic twist. Apple took over that position more than a decade ago, after TSMC became the exclusive manufacturer of its custom processors for the iPhone and iPad. The world’s largest contract manufacturer of semiconductors for third parties and which has the most advanced production nodes on the planet is a good thermometer to assess changes in global technology.
And this change reflects the explosive AI economy, at least in infrastructure. All the major cloud providers and companies are competing to get NVIDIA accelerators, which has fueled its record revenues, rise in the stock market to become the most valuable company in the world and capitalizing like no one else on the era of AI.
NVIDIA and TSMC, keys in chips for AI
Huang expanded his perspective at the World Economic Forum in Davos, where he participated in a fireside chat with Larry Fink, CEO of BlackRock, about the economic footprint of AI. There, he described the global AI system as a “five layer cake”: energy, at the base, semiconductors, cloud infrastructure, AI models and finally the application layer that generates revenue across sectors.
He described the current moment as the initial stage of “the largest infrastructure construction in human history”a transformation that already involves hundreds of billions of dollars and ultimately requires trillions more.
All of this is forcing contract chip manufacturers to reallocate capacity towards hardware for AIwhich threatens to cause the decline of the PC market and the rest of the industry in general. TSMC plans to build 20 new manufacturing plants around the world, while manufacturers such as Foxconn, Quanta and Inventec are developing dozens of new “AI factories” to assemble systems based on GPU accelerators.
Large semiconductor manufacturers such as Micron, SK Hynix and Samsung are also investing tens to hundreds of billions of dollars to meet growing demand for memory.
