NVL72
PulseAugur coverage of NVL72 — every cluster mentioning NVL72 across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
NVIDIA revises Rubin GPU specs, reducing memory and NVLink bandwidth
NVIDIA has revised the specifications for its upcoming Rubin GPU, reducing the HBM4 memory bandwidth and NVLink interconnect bandwidth compared to earlier projections. The Rubin GPU will now offer 19.2 TB/second of memo…
-
Grace Blackwell scheduler boosts usable GPUs by 27% through rack-locality optimization
A new scheduler, dubbed Grace Blackwell, has been developed to optimize GPU utilization for large-scale AI training and inference. This scheduler focuses on maintaining rack-locality for multi-node jobs, which is crucia…
-
SpaceX secures $40B debt financing for 2027 GPU acquisition
SpaceX has secured $40 billion in debt financing specifically for the acquisition of graphics processing units (GPUs), with these chips anticipated for delivery in the first half of 2027. Elon Musk confirmed this signif…
-
NVIDIA unveils containerized DSX PODs for rapid AI infrastructure deployment
NVIDIA announced its DSX POD reference design at GTC Berlin 2026, a system that integrates multiple Vera Rubin NVL72 racks within a shipping container. This modular design aims to streamline the deployment of AI inferen…
-
SemiAnalysis proposes new GPU cluster rental SLAs for reliability
SemiAnalysis is proposing new Service Level Agreement (SLA) terms for GPU cluster rentals, focusing on clear definitions of downtime and credits for failures. The proposed SLAs aim to ensure that restored capacity is ac…
-
CoreWeave deploys NVIDIA Vera Rubin systems, Cognition reports 4.8x speedup
CoreWeave has begun deploying NVIDIA's Vera Rubin NVL72 systems, making them available to customers. Cognition, an AI company, has reported a significant performance increase, achieving 4.8 times higher token throughput…
-
AI interconnect demand surges, reshaping copper vs. optical debate
The demand for AI infrastructure is driving up the need for high-speed interconnects, pushing companies like Astera Labs to develop solutions that can handle increased data exchange between GPUs. This has led to a re-ev…
-
Together optimizes inference on Nvidia Rubin GPU with new features · 4 sources tracked
Together has announced advancements in their inference and OSS capabilities, leveraging new features from Nvidia's Rubin GPU. The company has updated its b200 gemms to incorporate Rubin's wider MMA steps, increased tens…
-
Nvidia and Palantir partner on AI-driven supply chain optimization · 4 sources tracked
Nvidia and Palantir have partnered to leverage AI for optimizing supply chain operations, with Nvidia's own complex, million-part supply chain serving as the initial test case. The collaboration utilizes Palantir Foundr…
-
OpenAI chip reportedly outperforms NVIDIA Rubin NVL72 on perf per watt
OpenAI has reportedly developed a new chip that offers superior performance per watt compared to NVIDIA's latest offerings. Despite lacking specific decoding capabilities, OpenAI's chip reportedly outperforms NVIDIA's R…
-
SemiAnalysis releases AgentX dataset for agentic inferencing
SemiAnalysis has released AgentX, an open-source dataset designed for agentic inferencing, featuring a 1 million token context length and multi-turn capabilities. The dataset aims to test the resilience of CUDA's domina…
-
TPU partners with Mooncake for inference optimization
SemiAnalysis reports that Tensor Processing Unit (TPU) is collaborating with the open-source inference optimization library Mooncake. This partnership aims to integrate TPU capabilities with Mooncake Store, enhancing pe…
-
AI Hardware Focus Shifts to Energy, Supply Chains, and Networking
The AI hardware landscape is undergoing a significant shift, moving beyond just GPUs to focus on energy efficiency, supply chain stability for critical minerals like copper, and advanced networking capabilities. Former …
-
Cursor releases open-source MoE training megakernel, Mixture-of-Kittens
Cursor Research has open-sourced Mixture-of-Kittens (MoK), a specialized training kernel designed for Mixture-of-Experts (MoE) models. This megakernel fuses MoE communication and computation into a single deterministic …
-
AI hardware advances with new clusters; open-source debated
Ineffable Labs has acquired its first Vera Rubin-NVL72 clusters, powered by Google Cloud and NVIDIA hardware. This move signifies a generational leap in AI hardware, accelerating the pace of advancement in the field. Se…
-
Kimi K3 model's scale and efficiency boost AI hardware demand, says SemiAnalysis
SemiAnalysis argues that the Kimi K3 model, despite its linear attention and lower KV-cache requirements, is beneficial for NVIDIA and the broader AI hardware ecosystem. The model's massive 2.8 trillion parameters neces…
-
SpaceX boosts AI Sat V1 power specs for enhanced compute capacity
SpaceX has increased the power specifications for its AI Sat V1 satellite. The peak power output has been raised to approximately 250kW with battery assistance, and the average power will be around 160kW. This upgraded …
-
Nvidia CEO denies Vera Rubin AI platform delays, cites 'giant amounts' in production
Nvidia CEO Jensen Huang has refuted claims of production delays for the company's upcoming Vera Rubin AI platform, stating that it is already in production and will be delivered in 'giant amounts.' While Huang confirmed…
-
AI Value Capture Shifts: From Infrastructure to Model Labs
SemiAnalysis reports that the AI industry's value capture is shifting from infrastructure providers to model labs. While 2023-2025 saw significant gains for companies like TSMC and NVIDIA due to infrastructure build-out…
-
NVIDIA's VR NVL72 System Features Cableless Design, Rubin GPUs
SemiAnalysis has provided a detailed breakdown of NVIDIA's upcoming VR NVL72 system, highlighting significant design changes from previous architectures like GB200. Key innovations include a cableless compute tray desig…