AMD looks to tighten grip on CPU crown with Venice customer-centric core power
www.sdxcentral.com — Leadership leans in on customer outcomes over benchmarks while still flexing big performance numbers
Infrastructure & Hardware
Infrastructure and hardware comprise the physical compute systems, specialized chips, and data center components that power artificial intelligence workloads. Truncated covers chip architectures, supply chains, energy requirements, and the hardware deployments that enable large scale computing.
How We Built TLX: A Hardware-Native GPU Compiler for Production ML at Scale
engineering.fb.com — Triton Low-Level Extensions (TLX), is an extension to Triton that helps engineers optimize GPU performance while preserving Triton’s productive programming model. TLX introduces explicit war…
A closer look at Verano, AMD's first step toward a Zen-based superchip to rival Nvidia's Vera Rubin
The Register — Cores, clocks, memory, and interconnects all perfectly balanced for a singular mission: keeping instinct fed.
NVIDIA CEO Huang's Troubles Send Shockwaves Across AI Chip Supply Chains As Chatter Starts About Firm Losing Dominance To Custom ASIC AI Chips
wccftech.com — Long-term TSMC contracts rarely free up capacity, but cheaper ASICs such as Google's TPUs could gain volume as US scrutiny of NVIDIA grows.
Private AI Needs Infrastructure That Can Evolve With the Workload
www.unite.ai — Two years ago, almost every enterprise I talked to assumed AI meant calling a frontier model through an API. That assumption is breaking down. Open-weight models have gotten good enough for a lot of day-to-day work, and...
SoftBank, Petros, and Grab plan Malaysian data center buildout with "long-term target" of 5GW
www.datacenterdynamics.com — Kicking off with an initial $1 billion investment
Fluidstack begins construction of 99-acre data center campus in Wythe County, Virginia
www.datacenterdynamics.com — Set for land on Progress Park
Data sovereignty is reshaping AI infrastructure
www.cio.com — Organizations scaling sovereign AI no longer need to trade compute speed for strict data residency. Modern on-premises architecture bridges the gap between compliance and small-file storage performance.
The CIO dilemma: Scaling production AI without breaking the power grid
www.cio.com — Shifting to production AI often increases energy demands. See how technology leaders leverage consolidation and liquid cooling to help optimize efficiency without compromising performance.
Boosteroid, Gaw Capital, and Firebird AI plan 480MW data center in Kazakhstan
www.datacenterdynamics.com — Would be part of phase two of “Data Center Valley” project in nation, targets 1GW by 2029
Record 350 megawatts of cooling gear ships in a month for data centers
interestingengineering.com — XNRGY Climate Systems has scaled up to meet the soaring cooling demands of AI data centers across North America.
Start and Restore GPU Workloads in Seconds with NVIDIA Snapshot
developer.nvidia.com — NVIDIA Snapshot, a new open-source component of the NVIDIA DSX AI Factory Platform, makes initialized GPU workloads reusable across inference replicas—speeding startup and recovery.
TensorBay launches 1MW containerized AIPods
www.datacenterdynamics.com — Powered by Nvidia Blackwell compute
Hybrid system could make cooling AI data centers cheaper and more efficient
techxplore.com — Data centers at the heart of the artificial intelligence (AI) revolution could be cooled more effectively and economically with a hybrid system under development at the University of Waterloo that targets hotspots on computer chips with a single drop of boiling coolant.
Germany's Deutsche Telekom to expand Munich AI data center by 20MW
www.datacenterdynamics.com — Company adding B300 capacity after launching original cluster in February
KORA builds AI models and agents into a new Linux OS bound for Qualcomm Snapdragon PCs
siliconangle.com — KORA builds AI models and agents into a new Linux OS bound for Qualcomm Snapdragon PCs - SiliconANGLE
Elon Musk & SpaceX's Next-Gen Space AI Satellites Could Boost Peak Compute To 400kW, Says Analyst, With 10,000 NVIDIA GPU Racks Expected In Compute Capacity By 2028
wccftech.com — Elon Musk's SpaceX should have 5 gigawatts of space AI computing capacity by 2028, believes analyst Jeff Pu of GF Securities. Pu's estimate relies on NVIDIA's space computing platform, which he dubs Kessel racks that rely on the latest Vera Rubin AI systems. The analyst believes that the thermal design point (TDP) of each of the GPUs present in the NVL72 platform for the Kessel racks will be 4,000 watts, a significant bump over what rumors have pointed towards as the power dissipation of standard terrestrial NVIDIA GPUs. SpaceX's Next-Generation AI Satellites Could Bump Solar Power Generation To 400kW Peak, […]
Going deep on local AI performance benchmarking with Apple’s Mac Studio with M5 Ultra taking on Nvidia DGX Spark — inference engine recipes mandatory for bigger models as MLX stack matures
Vertiv partners with Infineon to scale power semiconductors for AI data centers
www.datacenterdynamics.com — Inks three-year strategic agreement
Too big to trip
www.datacenterdynamics.com — Once viewed through the lens of individual facility protection, grid resilience now depends on data centers remaining connected and supporting system stability during power disturbances
Dismiss email signup
Subscribe to 
The top AI news in your email delivered each morning. Choose categories