Tiny tech, big AI power: what are 2-nanometre chips?

Taiwan Semiconductor Manufacturing Company (TSMC), a leader in microchip technology, is set to commence mass production of its cutting-edge 2-nanometre chips by the end of this year. This breakthrough promises to significantly enhance computing power, offering a glimpse into the future of technology. The evolution of chip technology has been remarkable, with manufacturers continuously cramming more microscopic components into these crucial devices. The introduction of 2nm chips marks a leap forward, providing greater energy efficiency and performance compared to their predecessors. These advanced chips are designed to incorporate more transistors, thereby accelerating the processing capabilities of laptops, reducing carbon footprints in data centers, and enabling self-driving cars to detect objects more swiftly, as highlighted by IBM. According to Jan Frederik Slijkerman, a senior sector strategist at ING, this innovation will facilitate faster and more capable AI on consumer devices, as well as enhance the efficiency of AI models in data centers. However, the production of these state-of-the-art chips is no small feat. It requires sophisticated lithography machinery, extensive knowledge of manufacturing processes, and substantial investment. Only a select few companies, including TSMC, Samsung, and Intel, have the capability to produce 2nm chips. TSMC currently holds a competitive edge, while its rivals are still working on improving their production yields. Furthermore, Japanese chipmaker Rapidus is in the process of establishing a plant in northern Japan, with plans to start mass production of 2nm chips in 2027. The journey to mass production has not been without challenges. Recently, three individuals were charged in Taiwan for allegedly stealing trade secrets related to these chips to assist Tokyo Electron, a Japanese firm supplying TSMC. This incident underscores the importance of national technology security. Additionally, geopolitical tensions impact TSMC's operations, particularly in relation to U.S. restrictions. Reports indicate that the company will avoid using Chinese chipmaking equipment for its 2nm production to mitigate potential disruptions. TSMC is also aiming to boost its production capacity in the U.S., with plans set for the end of the decade. When discussing the size of 2nm, it’s important to clarify that this measurement does not directly reflect the physical size of the chips themselves but serves as a marketing reference. The smaller the number, the greater the density of transistors that can be packed into a chip. IBM notes that 2nm technology can accommodate up to 50 billion transistors on a chip the size of a fingernail. This is achieved through a meticulous process of etching silicon slices and layering them with various materials to create the transistors, ultimately leading to increased processing power. As for the future, advancements are already underway, with TSMC developing 1.4-nanometre technology anticipated to enter mass production by 2028, keeping pace with competitors Samsung and Intel. TSMC has also recently commenced high-volume production of 3nm chips in 2023 and is reportedly constructing a facility for 1.4nm chips in Taichung, Taiwan. Rapidus has expressed that 2nm chips are particularly suited for AI servers and are poised to become foundational to the next generation of digital infrastructure, despite the complexities and costs involved in their production.