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…
A developer argues that the real money in AI is in the physical infrastructure layer—chips, servers, power, and cooling—not the chatbots. Using Groq as an example, the piece explains the 'memory wall'…
Unsloth achieves up to 7.3x training speedup over standard Transformers for MoE models like gpt-oss-20B on an NVIDIA B200, according to published benchmarks, while Axolotl delivers up to 1.45x speedup…
Cisco Foundation AI released Antares, a family of small language models (350M and 1B parameters) that localize known vulnerabilities inside real codebases. Antares-1B achieves 0.209 File F1 on the Vul…
Moonshot AI paused new sign-ups for its Kimi K3 model after demand pushed its GPU cluster to capacity. The Chinese lab said on X that demand over 48 hours "pushed close to the limits of our current ca…
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…
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…
The Compute Exchange Inc. launched a secondary marketplace for used and refurbished Nvidia H100 and A100 GPUs, connecting buyers with suppliers as enterprises and AI startups seek cost-effective AI ha…
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. …
Average GPU utilization across tens of thousands of production Kubernetes clusters is just 5%, according to the Cast AI 2026 State of Kubernetes Optimization Report. Default Kubernetes scheduling uses…
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…
RightNow AI open-sourced bonsai-turbo, a batch-1 decode engine that runs PrismML's Bonsai 27B ternary LLM 1.76x faster than the official llama.cpp fork on an H100, achieving 151 tok/s (ternary) and 15…
Fine-tuning a 70B parameter model like LLaMA 3.1 locally requires significant VRAM, with a full fine-tune needing roughly 1.1 TB before activations and realistically 8× H100/A100-80GB GPUs with ZeRO-3…
Alternative GPU cloud providers Together AI, Runpod, and Nebius are gaining market share as AWS faces GPU shortages, with Together AI offering H100 clusters at $6.79 per hour on-demand and reportedly …
The PyTorch-Triton 3.7 release introduces the Triton Plugin Extensions system, a framework for dynamically loading custom compiler passes, dialects, and DSL extensions into upstream Triton at runtime …
NVIDIA released Nemotron-Labs-TwoTower on July 1, achieving 2.42 times faster inference throughput at 98.7% of the baseline model's benchmark quality by adding a second neural network tower trained on…
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…
Tech companies are missing market expectations due to a slowdown in customer spending on AI infrastructure, reflecting deeper economic challenges in scaling AI models. The volatile GPU supply chain, p…
Nvidia has slashed its list of Asian buyers for advanced AI chips by more than half, implementing a stringent new 'white list' after tightening compliance checks to block shipments from reaching China…
Infrastructure strategy has shifted from optimizing known variables to making decisions amid relentless uncertainty, driven by rising AI demands and volatile cloud pricing. Companies face massive cost…