Topic
Explore English articles about Semiconductor Manufacturing.
Intel is scaling its upcoming Diamond Rapids Xeon processor to as many as 256 cores, rebuilding it around 16 core chiplets and two fabric hubs. Peak bandwidth reaches roughly 1.6TB/s, though real-world performance figures remain unpublished.
Cerebras' CS-4 keeps the same 5nm wafer but redesigns power delivery, cooling, and I/O to double peak compute per wafer. Here's what the performance claims actually cover and how production-ready the system really is.
Rapidus has published a roadmap to build an 8-reticle (6,640 mm²) interposer on a 600×600mm glass panel by 2030. Claiming more than 12 times the yield efficiency of a 300mm wafer, the company is pursuing a manufacturing approach distinct from TSMC's wafer-based route to scale up AI chips.
Micron announced a $10 billion, 10-year research hub in Boise, drawing congratulatory remarks from Jensen Huang and Tim Cook. Around the same time, Western Digital launched an AI research center with Tohoku University. As HBM-first production reallocation drives memory prices higher, we examine the brain-power race unfolding across the US and Japan.
China's semiconductor equipment makers are growing sales, but their reported profit increases of over 400% mix in fair-value gains and swings from losses to profits. This article examines, through their financial results, what it would take to turn the current global equipment shortage into a lasting market share gain.
Researchers at the Hebrew University of Jerusalem used millisecond light pulses to instantly heat a bismuth oxide thin film to about 2,000°C while keeping the underlying glass below 100°C, locking in a metastable beta phase at room temperature. Photocurrent rose by up to 50 times. The reversible phase-switching this achieves on transparent conductive glass was previously unreachable with furnace heating.
Reuters reports that new SF4 orders from Samsung rose 10-15% in the US and China, and 5-10% in Taiwan. While demand from China exceeds supply capacity, prioritization of US customers and capacity reserved for in-house use are constraining the production allocation.
A team from the Max Planck Institute for the Science of Light and Harvard University used inverse-design algorithms to shrink wavelength demultiplexers, mode demultiplexers, and mirrors on silicon nitride photonic chips to as little as 1/500th the area of conventional designs, publishing the results in Nature Communications. To the researchers' knowledge, this is the first such demonstration on thick-film silicon nitride, opening a path toward integrated quantum photonics.
ATLANT 3D says an unnamed AI hyperscaler has ordered its atomic layer deposition tool. The device uses a micro-nozzle to directly pattern thin-film samples under varying conditions, aiming to speed up physical validation of AI-generated materials and the generation of training data.
AI-driven HBM demand has absorbed DRAM production capacity, sending DDR5 memory prices soaring to 4-5x year-over-year levels worldwide. Micron warns the supply shortage will persist until 2028, as structural changes in the consumer memory market become increasingly irreversible.
