TSMC Unveils Advanced Technology for Assembling Larger, Faster Chips

Taiwan Semiconductor Manufacturing Co. unveiled technology on Wednesday for producing faster chips and assembling them into dinner-plate-sized packages, which will improve performance for artificial intelligence applications.

It stated that its A14 manufacturing technology will be available in 2028 and will be capable of producing processors that are 15% faster at the same power consumption as its N2 chips, which are set to go into production this year, or that use 30% less power at the same speed.

The world's largest contract manufacturer, which counts Nvidia and Advanced Micro Devices as clients, added that its forthcoming "System on Wafer-X" will be able to weave together at least 16 large computing chips, memory chips, and fast optical interconnections, as well as new technology to deliver thousands of watts of power to the chips. 

By comparison, Nvidia's current flagship graphics processing units are made up of two large chips stitched together, while its "Rubin Ultra" GPUs, set to be released in 2027, will be made up of four.

TSMC intends to build two factories to complete the work near its chip plants in Arizona.

Intel, which is working to establish a contract manufacturing business to compete with TSMC, plans to announce new manufacturing technologies next week. Last year, it claimed to outperform TSMC in terms of chip speed.

Demand for massive AI chips packaged together has shifted the battleground between the two companies from simply manufacturing fast chips to integrating them, a complex task that necessitates close collaboration with customers.

"They're neck-and-neck. "You're not going to choose one over the other because they have a technological advantage," said Dan Hutcheson, vice chair of analyst firm TechInsights.

"You're going to pick one over the other for different reasons."

Customer service, pricing, and wafer allocation are all factors that will likely influence a company's decision on which chip manufacturer to use.