Silicon, foundries, GPUs — the substrate of everything.
105 stories · page 5 of 5
A new multi-year agreement between Nvidia and SK hynix shifts high-bandwidth memory from a commodity supply model to deep, joint silicon co-design.
New add-in cards utilizing AMD's Promontory 21 chipset offer a low-cost alternative to expensive PCIe switches, bypassing motherboard bifurcation limits for dense I/O.
Gigabyte's new 1U high-density server packs 40 Intel Lunar Lake processors, demonstrating how client-grade on-package memory architectures are redefining edge compute density.
AMD's upcoming Helios platform, designed to challenge Nvidia's rack-scale dominance, will initially rely on UALink-over-Ethernet, introducing latency and overhead trade-offs for scale-up AI clusters.
Nvidia is making a strategic push into the CPU market, partnering with major PC vendors to integrate AI agents directly onto devices. This move signals a significant expansion of its silicon footprint beyond GPUs, leveraging its AI expertise to redefine the personal computing lan
AMD's decision to re-release its iconic 3D V-Cache processor exposed an overlooked challenge in advanced packaging: the rapid obsolescence of early silicon-on-silicon bonding processes.
Nvidia is bringing its Grace Blackwell architecture to the client market with the RTX Spark platform, pairing Arm cores with Blackwell graphics and tackling the Windows-on-Arm compatibility bottleneck.
A new cryptojacking campaign leverages sophisticated social engineering, including SEO poisoning and AI chatbot recommendations, to infect high-end PCs and exploit their GPUs for illicit mining operations. This threat underscores evolving attack vectors against deployed silicon.
Both vendors confirmed in May that export licenses cleared for the H200 and MI308 — the de-facto China-tier successors to Hopper and MI300. The policy whiplash matters less than the volume implication: a market that lost $15B+ in pent-up demand is reopening.