Moonshot AI released the Kimi K3 open-weight model with 2.8 trillion parameters on Hugging Face on July 27, but only 1.8% of parameters (about 50 billion) are active per forward pass due to Mixture-of…
AWS raised H200 GPU instance prices by 15% on January 4, 2026, the first GPU price increase in roughly two decades, while average GPU utilization across more than 23,000 Kubernetes clusters sits at ju…
Moonshot AI's Kimi K3, a 2.8-trillion-parameter model that topped the Frontend Code Arena leaderboard on day one, releases its full open weights on July 27. The MXFP4 weights require approximately 1.4…
Alibaba's Qwen team previewed Qwen3.8-Max-Preview, a 2.4 trillion-parameter multimodal model, on July 19, 2026, at the World AI Conference in Shanghai, claiming it is 'second only to Fable 5' among be…
Moonshot AI's Kimi K3, a 2.8-trillion-parameter open-weight multimodal model released on July 16, 2026, generated an H100 CUDA kernel that runs 14.82 times faster than optimized PyTorch, nearly matchi…
AMD's Instinct MI350P PCIe AI accelerator, featuring CDNA 4 architecture and HBM3E memory, has been showcased at multiple trade shows including Dell Tech World, HPE Discover, and Computex 2026. The ca…
Average GPU utilization in production Kubernetes clusters is just 5%, according to the Cast AI 2026 State of Kubernetes Optimization Report, meaning 95 cents of every GPU dollar goes to idle silicon. …
A hardware analysis reveals that deploying 70-billion parameter models in FP16 requires moving 140 gigabytes of weights across the memory bus per token, creating a memory-bound bottleneck that limits …
GPU utilization averages just 5% across production Kubernetes clusters, according to the Cast AI 2026 State of Kubernetes Optimization Report, meaning 95% of provisioned GPU capacity is idle at any mo…
Thinking Machines Lab, Inc. released Inkling, a general-purpose multimodal model with 975 billion total parameters and 41 billion active parameters, on July 15, 2026 under an Apache 2.0 license. The m…
Kalshi, a CFTC-regulated prediction market, launched compute forward curves on July 14, allowing traders to bet on future GPU rental prices for Nvidia's B200, H200, and A100 chips. The contracts enabl…
WhiteFiber disclosed initial R&D results for Project Redwood, a distributed GPU supercluster architecture that linked two NVIDIA H200 clusters 83 km apart into one logical supercluster delivering 111.…
US Commerce Under Secretary Jeffrey Kessler told the House Foreign Affairs Committee on Tuesday that Nvidia has shipped only a "very small quantity" of H200 AI chips to mainland China and Hong Kong, c…
U.S. Commerce official Jeffrey Kessler confirmed that very few of NVIDIA's H200 AI chips have reached China or Hong Kong, months after the Trump administration announced it would allow limited sales. …
Kalshi, the first federally regulated prediction market exchange in the US, has launched contracts allowing traders to speculate on the per-hour cost of running NVIDIA's H100, H200, B200, and RTX 5090…
China's integrated circuit exports surged more than 96% year-on-year to US$177.28 billion in the first half of 2024, according to data from the General Administration of Customs, driven by global AI d…
Nvidia is halving its list of approved customers across Asia and tightening distributor vetting as US export controls squeeze its China market share, which is projected to collapse from 66% in 2024 to…
An engineer's analysis reveals that LLM inference throughput on NVIDIA GPUs is bottlenecked by memory bandwidth, not compute. A 7B model achieves ~15 tok/s on a T4 but over 3,500 tok/s on an H100, a 1…
Researchers have developed a memory-efficient training stack for Mixture-of-Experts models that achieves 4.7x to 8.2x higher per-GPU throughput compared to a finely-tuned FSDP2 baseline, using fewer t…
Solidigm and MinIO have developed a new memory tier that uses ultra-dense solid-state drives to offload key-value cache blocks from GPU high-bandwidth memory, addressing the AI memory wall in large la…