CME Group, the world's largest derivatives exchange, announced a partnership with Silicon Data to launch two Compute futures contracts tracking hourly rental costs of Nvidia's H100 and Blackwell B200 …
Taiwanese prosecutors in Keelung indicted nine people, including employees of Nvidia and Super Micro, on Monday for allegedly smuggling AI servers to China in violation of export controls. The case ma…
A developer argues that comparing GPU cloud providers by hourly rate alone is misleading, and that the true cost metric is cost per unit of work done, factoring in utilization, data transfer, provisio…
VLLM v0.27.1, the open-source inference engine, has unified its scheduler around a fixed token budget per step, making chunked prefill, prefix caching, and speculative decoding stackable by default, a…
Compute grew 106x since NVIDIA's P100 in 2016, while memory bandwidth grew only 10.9x, a 10x decline in bytes-per-FLOP that has left AI accelerators bandwidth-starved, according to a chip inventory of…
The AI infrastructure supply chain is experiencing a multi-year bullwhip effect, with shortages cascading from GPUs to memory, CPUs, and storage, and data center construction costs soaring to over $20…
Nvidia is partnering with data center developer Cloverleaf to build AI-native facilities optimized for its H100 and upcoming Blackwell GPUs, addressing power density and liquid cooling needs. The coll…
A new open-source tool, Mock Nvidia GPUs, lets Linux developers simulate NVIDIA GPU telemetry for monitoring software by intercepting NVML and nvidia-smi calls, without emulating CUDA or actual GPU ha…
Nvidia's latest demo shows that optimizing the inference stack, not the model weights, is the key to performance, achieving 4.2× higher throughput at half the latency on identical H100 hardware with L…
Zhipu AI's leaked Mythos benchmark suggests its upcoming GLM-4 model scores 87.2% on MMLU-Pro, 78.5% on GPQA-Diamond, and 99.1% on a 128k-context retrieval test, potentially outperforming GPT-4o on MM…
FleetSieve, a new profiling method for LLM serving fleets, reaches the oracle aggregate decision using 22,200 GPU-seconds, 6.9% less than uniform random profiling on a fixed H100 grid for a 31B-parame…
Liquid AI released DSpark draft model checkpoints for three LFM2.5 models, delivering up to 3.18x faster decoding on an H100 and up to 2.87x on an M4 Max MacBook Pro without changing model outputs. Th…
OpenAI has admitted that safety monitoring consumes 20% of inference compute, a significant operational cost that translates to $6,000 annually per H100 GPU. The admission signals that supervision is …
Jie Tang's team at Z.ai argues that LLM scaling should optimize for fixed inference budgets, favoring smaller, deeper architectures trained longer over wider models. Their ablation shows a 7B model tr…
Nvidia CEO Jensen Huang told customers on August 13, 2026, that A100 GPUs will remain effective through 2029, while the company simultaneously pushes its Blackwell, Hopper, and upcoming Rubin architec…
Nvidia's H200 AI chips are quietly entering China through gray-market channels despite U.S. export controls, with volumes in the hundreds, enough for cluster bring-ups and inference optimization but n…
The Commodity Futures Trading Commission (CFTC) is seeking public input on futures contracts for AI computing power, with a draft request for comment sent to the White House's Office of Management and…
Cerebras Systems claims its Wafer-Scale Engine 3 (WSE-3) delivers 1,800 tokens per second at 23 kW for Llama 3 70B inference, versus about 850 tokens per second at 56 kW for an 8×H100 DGX system, yiel…
The Commodity Futures Trading Commission has filed a Request for Comment on derivatives contracts tied to computing capacity, seeking feedback on cash market liquidity, manipulation risks, customer pr…
Researchers introduced PTXBench, a benchmark for evaluating and adapting large language models (LLMs) to use architecture-specific PTX for GPU kernel optimization, measuring functional correctness, ta…