Tenstorrent unveils Open Chiplet Atlas ecosystem to standardize chiplet interoperability
Tenstorrent introduced the Open Chiplet Atlas ecosystem in San Francisco, aiming to make chip design more modular and accessible. The effort focuses on plug-and-play interoperability for heterogeneous chiplets and already counts more than 50 partners.
AMD lands U.S. Energy Department pact for Lux and Discovery Artificial Intelligence supercomputers
AMD secured a U.S. Department of Energy partnership to build Lux and Discovery, two next-generation supercomputers designed for scientific and national security workloads. Lux arrives within six months, with Discovery due by the end of the decade, advancing U.S. leadership in high-performance computing and Artificial Intelligence.
Server dram prices jump up to 50 percent as allocations drop to 70 percent
Samsung and SK Hynix have retroactively raised rdimm contract prices by 40 to 50 percent and cut allocations by 30 percent, leaving tier-1 cloud buyers at roughly 70 percent fulfillment. Smaller oems face 35 to 40 percent fulfillment into early 2026 as ddr4 supply tightens and lead times lengthen.
Harbour BioMed launches fully human generative Artificial Intelligence HCAb model to speed biologics discovery
Harbour BioMed unveiled a generative Artificial Intelligence model for fully human heavy chain only antibodies, promising faster, higher-quality biologics discovery. The company also launched a new alliance to advance Artificial Intelligence-driven drug R&D.
Nvidia introduces NVQLink to connect quantum processors with GPU computing
Nvidia unveiled NVQLink, an open system architecture that tightly couples GPU computing with quantum processors to build accelerated quantum supercomputers. The effort is guided by researchers at major U.S. labs and supports 17 QPU builders, five controller builders and nine national laboratories.
Onexplayer unveils water-cooled Onexfly Apex with AMD Ryzen Artificial Intelligence Max+ 395
Onexplayer’s Onexfly Apex pairs AMD’s Ryzen Artificial Intelligence Max+ 395 with a detachable liquid cooling tower and an 85 Wh external battery, targeting unprecedented power in a Windows handheld. The company claims up to 120 W APU draw with the cooling dock attached.
Nvidia launches BlueField-4 DPUs with 800 Gb/s throughput for Artificial Intelligence data centers
Nvidia unveiled its BlueField-4 data processing unit at GTC Washington, D.C., targeting gigascale Artificial Intelligence infrastructure with 800 Gb/s throughput and software-defined acceleration across storage, networking and security.
Nvidia IGX Thor brings real-time physical Artificial Intelligence to the industrial and medical edge
Nvidia unveiled IGX Thor, an industrial-grade platform that brings real-time physical Artificial Intelligence to the edge with high-speed sensor processing, enterprise reliability and functional safety. The compact module promises up to 8x the compute performance of IGX Orin and is drawing interest from robotics, industrial and healthcare leaders.
Supermicro expands Nvidia collaboration with U.S.-made Artificial Intelligence systems for federal use
Supermicro is showcasing new U.S.-manufactured, TAA-compliant Artificial Intelligence systems at Nvidia GTC in Washington, D.C., and plans to deliver next-generation Vera Rubin platforms in 2026. The lineup includes high-density HGX B300 configurations, a Super AI Station based on GB300, and rack-scale GB200 NVL4 HPC solutions.
Nvidia and Google Cloud expand enterprise artificial intelligence with G4 VMs and Omniverse
Google Cloud made G4 virtual machines generally available, powered by Nvidia RTX Pro 6000 Blackwell Server Edition GPUs, and listed Omniverse and Isaac Sim images on its marketplace. The releases target visual computing, agentic and physical artificial intelligence, and industrial digitalization.