technology
-
d-Matrix is integrating its Raptor inference XPUs with NVIDIA NVLink Fusion and MGX, targeting 144 accelerators and roughly 2.3 TB of memory per rack.
-
Cellular automata researchers have found a new digital spaceship moving at nine tenths of the maximum one cell per generation speed in an isotropic non totalistic rule. The September 17 discovery by Jiahao Yu, known in the ConwayLife community as yujh, fills a speed researchers had been actively trying to reach only days earlier. The…
-
Astera Labs expands its Leo Smart Memory Controller family with fabric-attached AI memory, PCIe 6 and CXL 3.2, claiming up to 62% faster time to first token.
-
Tower Semiconductor is turning its Japanese manufacturing footprint into what is planned to become its largest optical-connectivity semiconductor hub, with new reporting putting the program at roughly $4 billion and projecting the equivalent of 45,000 300 mm wafers per month by 2029. New details published September 24 say the expansion could lift Tower’s Japanese capacity…
-
Qualcomm is moving deeper into physical AI with an agreement to acquire PickNik Robotics, the longtime steward of MoveIt, one of the most widely used open-source frameworks for robot manipulation and motion planning. In a move centered on the software layer that turns computing into physical motion, Qualcomm says MoveIt is expected to integrate more…
-
QumulusAI says it has activated 616 NVIDIA RTX PRO 6000 Blackwell GPUs for Runpod across 77 GPU nodes, putting the full reserved capacity online ahead of its original September 1 target. The company said the deployment reached full activation in August and is being served from its existing U.S. data-center footprint. Separate coverage confirms the…
-
NEURA Robotics and SECO are joining forces to industrialize European physical AI hardware, including compute modules intended to help move NEURA’s 4NE1 humanoid robot toward series production. The partnership, announced September 7, combines Germany-based NEURA’s cognitive robotics architecture with Italian technology company SECO’s embedded engineering and manufacturing capabilities. The companies say they will jointly develop…
-
Leviton is committing $100 million over five years to expand manufacturing capacity for the physical networking systems that increasingly sit underneath AI clusters, data centers and smart buildings. The company announced the program on September 24, saying the capital will go into its Network Solutions factories worldwide, with the largest investments at key U.S. factories.…
-
A BlackRock-backed AI infrastructure consortium and IFM are reportedly in exclusive talks for STACK Infrastructure’s Asia-Pacific data-center portfolio, potentially valuing the assets at $20–25 billion.
-
Enphase is building 4 kW IQ Solid-State Transformer modules in Texas for modular AI data-center power racks scaling to 5 MW and converting medium-voltage AC directly to 800 VDC.
-
Agility Robotics has unveiled Digit 5, a redesigned industrial humanoid intended to work closer to people while handling a wider range of manufacturing and warehouse jobs. The fifth-generation Digit can repeatedly lift up to 50 pounds, about 40% more payload than Digit 4, according to Agility. A redesigned battery provides a stated 90-minute runtime and…
-
SiPearl has delivered the first samples of its Rhea1 processor to Bull, moving the European-designed server CPU from laboratory validation into integration for the JUPITER exascale supercomputer. Bull says it has begun integrating Rhea1 into the BullSequana XH3000 platform used for JUPITER’s CPU-based Cluster Module at the Jülich Supercomputing Centre in Germany. The CPU module…
