AI chips - Page 2

TSMC will add a further 3nm fab in Tainan, with volume production in the first half of 2028, making three new 3nm fabs in Taiwan, Arizona and Japan.

TSMC's third-quarter revenue rose 50.9% to NT$1,494.24bn. It guided Q4 to US$52.3-54.3bn and now expects about 44% revenue growth for 2026.

ByteDance is close to a deal to buy custom MediaTek chips for its data centres, Bloomberg reports, citing people familiar. Neither company has confirmed.

A Delaware jury found Arm breached its license with Qualcomm but awarded only $2 in nominal damages. Both companies are claiming victory.

Astera Labs' Orion is a rack reference design for AI inference that works with any GPU or CPU and scales to 160 per pod. Partners sell it from Q1 2027.

AMD has launched its next Ryzen AI Halo developer platform, with the Ryzen AI Max+ PRO 495 and up to 192GB of memory for running big AI models locally.

Chinese tech firms bought over 1mn Nvidia gaming and workstation chips this year, the FT reported, citing industry insiders. Nvidia has not confirmed it.

IBM's Spyre accelerator now compiles through PyTorch's open-source Inductor. IBM says the new compiler code is about 7x smaller than the one it replaced.

Intel names its Crescent Island inference GPU the Intel GPU D-series. The first card, the D270, is an air-cooled PCIe card with up to 480GB of memory.

AMD unveils Helion, an open OCP-based rack blueprint for agentic AI with up to 32,768 EPYC CPU cores. AMD says it is planned for November 2026.