TSMC shared its massive manufacturing roadmap at the North American Technology Symposium 2026 on Wednesday. The giant chipmaker laid out its exact factory plans through the year 2029. Executives introduced brand new 1.2-nanometer and 1.3-nanometer fabrication processes, officially named A12 and A13. Furthermore, the company announced an unexpected extension for its older 2-nanometer family called N2U. Surprisingly, the company also confirmed it will absolutely not buy the highly expensive High-NA EUV lithography machines anytime before 2029. Instead, TSMC plans to rely on clever engineering to shrink its chips and save a lot of money.
For many decades, everyday smartphones generated the most money for TSMC. However, artificial intelligence and high-performance computing recently took over the top spot in the financial books. Because of this massive market shift, TSMC decided to completely change the rules of the game. The company will split its manufacturing strategy into 2 paths rather than forcing a one-size-fits-all solution on its buyers. Moving forward, TSMC will release a new, highly efficient chip design for smartphones every single year. Meanwhile, the company will only roll out a new, heavy-duty AI server chip every 2 years.
Kevin Zhang, a senior vice president of global sales at TSMC, explained the new smartphone plan to the eager crowd. Last year, the company announced the A14 technology, which features advanced nanosheet transistors. TSMC plans to start making these A14 chips in 2028. On Wednesday, Zhang introduced the A13 chip for the year 2029. The A13 acts as a slightly smaller, upgraded version of the A14. It shrinks the physical chip dimensions by roughly 3 percent, which creates an overall 6 percent area reduction. This clever optical shrink allows tech companies to use the new chips and gain performance without spending millions of dollars redesigning their entire phone motherboard.
TSMC also wants to offer its customers a more affordable way to upgrade older designs without breaking the bank. The company introduced the N2U node to extend the life of its massive 2-nanometer platform into its 3rd year. When N2U arrives in 2028, it will offer companies a fantastic choice. Hardware makers can squeeze out 3 to 4 percent faster computing speeds using the same power, or use 8 to 10 percent less battery power while maintaining current speeds. N2U also provides a modest 2-3% improvement in logic density. If a gadget maker builds a high-end product in 2027 using the older N2P process, they can easily upgrade that exact design to N2U in 2028, saving massive engineering costs.
On the opposite side of the business, AI data centers demand maximum speed at any cost. TSMC created the A16 and A12 nodes specifically for these power-hungry computing giants. The A16 features a brand new backside power delivery system called Super Power Rail. While TSMC will finish designing the A16 technology later in 2026, actual factories will wait for customer demand and start pumping out high volumes in 2027. Interestingly, the A16 will not replace the N2X, which remains a specialized chip that pushes traditional front-side power delivery to its absolute maximum clock speeds.
After the A16 hits the market, the incredibly powerful A12 node will take the performance crown in 2029. The A12 will shrink both the front and back sides of the silicon to pack even more processing power into a tiny space. The most shocking news from the event involved the actual tools TSMC uses to build all these tiny components. Rival company Intel currently plans to use ultra-expensive High-NA EUV machines starting in 2027 to build its own 14A chips. However, TSMC refuses to touch that expensive equipment before 2029. Zhang proudly told the crowd that his research team amazed him by finding clever ways to advance technology with their current machines. While TSMC might eventually need the massive new machines in 1 day, the company will continue to save billions of dollars right now by simply maximizing its current factory setups.
