Battery energy storage systems (BESS) are becoming essential infrastructure for AI factories, which require power-dense, fast-changing electrical loads that differ from traditional data centers. Prope…
Meta's PyTorch team released TorchCodec 0.14, adding a fast WavDecoder that bypasses FFmpeg for direct WAV file reading and HDR video decoding support for CPU and CUDA with full float32 precision. The…
TensorStack AI released Amuse v3.5.2, a 288MB local AI generation studio for Windows that allows users to run image, video, audio, and text models without configuring Python environments or complex wo…
Moonshot AI, developer of the Kimi AI assistant, saw its valuation surge sixfold to $30 billion in six months, driven by over $200 million in annual recurring revenue after its K2.5 model release. The…
A buyer who spent roughly $10,000 on both a Mac Studio M3 Ultra with 512 GB of unified memory and a dual NVIDIA DGX Spark setup ran the same 397-billion-parameter model on each machine and measured th…
A high-performance Expert Parallelism (EP) kernel is essential for running large Mixture-of-Experts (MoE) language models across multiple GPUs, as it handles the dynamic routing of tokens to experts l…
NVIDIA GPUs with Confidential Computing are now used for confidential inference in Apple's Private Cloud Compute, expanding beyond Apple's data centers to Google Cloud. The collaboration supports serv…
NVIDIA Isaac Lab on Amazon SageMaker AI now enables robotics teams to train reinforcement learning policies for humanoid robots at scale, using either SageMaker HyperPod for persistent cluster trainin…
NVIDIA launched Enterprise Manageability for its DGX Spark and GB10 systems, providing IT teams with a complete operational framework covering the full lifecycle from provisioning to end-of-life retir…
NVIDIA released a workflow for converting FP8-quantized CLIP model checkpoints into TensorRT engines, enabling faster inference and higher GPU throughput for production deployment. The process involve…
NVIDIA CEO Jensen Huang told Stanford students that his company achieved a 1,000,000x improvement in computing performance over the past decade through co-design — simultaneously optimizing hardware, …
NVIDIA introduced an automated, AI-driven research loop called Auto-FL within its NVIDIA FLARE platform to help researchers test and optimize federated learning strategies more efficiently. The system…
Chris Lattner, a lead engineer on Apple's original OpenCL implementation, explains why OpenCL and other C++-based GPU programming models failed to become dominant in AI, citing slow committee-driven d…
NVIDIA has introduced a new workflow using agent skills and Nemotron Speech to accelerate the evaluation of clinical automatic speech recognition (ASR) models, addressing the challenge of accurately r…
NVIDIA released a tutorial demonstrating how to build tiled GPU kernels for vector addition, matrix addition, and matrix multiplication using cuTile Python in Google Colab. The workflow includes envir…
NVIDIA's RTX 5090, priced at $1,999, offers 32 GB of GDDR7 memory with 1,792 GB/s bandwidth, making it the fastest consumer GPU for both gaming and local AI inference. The card excels at running 32B-c…
Apple overhauled its Apple Intelligence platform at WWDC 2026, announcing a strategic partnership with Google to co-develop Foundation Models and running its most powerful cloud AI on NVIDIA GPUs host…
NVIDIA has introduced the NVFP4 training recipe in TransformerEngine, enabling 4-bit mixed-precision pre-training on Blackwell GPUs with no measurable accuracy loss compared to FP8 baselines. The reci…
A developer reports that running gpt-oss-120B on an NVIDIA DGX Spark (GB10) failed due to MXFP4 quantization lacking native hardware support on SM121, causing kernel hangs and memory spikes. Switching…
NVIDIA Jetson Orin Nano Super 8GB benchmarks show 25W as the energy-efficiency sweet spot for sub-4B Bonsai LLMs, delivering 47-48% more tokens per second than 15W while maintaining or improving outpu…