Taiwanese tech giant TSMC has begun mass production of its advanced 2-nanometer (nm) semiconductors, opening the door to higher performance for artificial intelligence, the company announced Wednesday.
“TSMC’s 2nm (N2) technology began serial production in the fourth quarter of 2025, as planned,” the group reported in a statement published on its website.
LOOK: South Korea joins the global artificial intelligence race: this is the AX K1 model
The chips will be the “most advanced technology in the semiconductor industry in terms of density and energy efficiency,” he added.
They will be produced at TSMC’s “Fab 22” plant in the southern port city of Kaohsiung.
TSMC is the world’s largest maker of semiconductors used in everything from smartphones to missiles, and counts Nvidia and Apple among its clients, which have spent billions of dollars purchasing semiconductors, servers and data centers.
More than half of the world’s semiconductors and almost all of the most advanced ones used in artificial intelligence technology are manufactured in Taiwan.
Meanwhile, investment in AI has skyrocketed in the world and should reach nearly 1.5 trillion dollars by 2025 and 2 trillion by 2026, almost 2% of global GDP, according to the American research firm Gartner.
LOOK: China raises tension with Taiwan with war test: “It is one of the latent conflicts that may worry the most in 2026”
Advanced 2nm semiconductors have better performance and are structured to contain more key components called transistors.
This new technology will help speed up laptop computers, reduce the carbon footprint of data centers and allow autonomous vehicles to detect objects more quickly, according to American computing giant IBM.
For artificial intelligence, “this benefits both consumer devices – enabling faster and more efficient artificial intelligence on devices – and AI chips in data centers, which can operate large models more efficiently,” said Jan Frederik Slijkerman of the Dutch bank ING.
Slijkerman told AFP that the production of 2nm chips is “extremely difficult and expensive”, and few companies can manufacture them, such as TSMC, the South Korean Samsung and the American Intel.