ChipsChips
NVIDIA Releases Switchyard Rust Proxy for LLM Traffic
NVIDIA introduces an early Rust project for routing and translating requests between OpenAI and Anthropic APIs.
Topic
Chips
Compute hardware for AI: accelerators, processors, memory and networking. We follow vendor announcements, data-centre deployments and how hardware constraints shape the economics of training and inference.
Topic
Primary topic in 55 of 55 reports
- Chips
ChipsChips
NVIDIA Describes Practical CUDA Toolbox for GPU Computing
NVIDIA Developer Blog presented material on the practical application of CUDA, but optimization details are not revealed in the available summary.
ChipsChips
NVIDIA Introduces TensorRT Model Connect for Open Models
NVIDIA describes TensorRT Model Connect as a set of reference implementations designed to simplify the path from an open model checkpoint to inference.
ChipsChips
NVIDIA describes creation of an adaptive cybersecurity system
NVIDIA described an approach to adaptive cybersecurity, but available data do not reveal its results and details of implementation.
ChipsChips
NVIDIA Launched BioNeMo NIM in Claude Science for Proteins
NVIDIA described the launch of tools for protein structure prediction in Claude Science, but available data does not reveal results or usage conditions.
ChipsChips
CXMT produces first HBM3E chips in small batches
CXMT releases HBM3E in small batches for the first time, but scale and shipment readiness remain undisclosed.
ChipsChips
Nvidia invests $3,5 billion in Taiwanese MediaTek
Nvidia will invest 3,5 billion dollars in MediaTek amid the development of proprietary AI chips by major technology companies.
ChipsChips
Nvidia expands its lead with data-center systems
New efficiency of data-center systems is linked to data-flow management, not only to computing power.
ChipsChips
Neocloud Lambda raises $1 billion for Nvidia chips for Microsoft
Neocloud Lambda will direct loan funds to buy Nvidia chips for leasing to Microsoft.
ChipsChips
Databricks descried fault-tolerant PyTorch training in AI Runtime
Databricks described an approach to PyTorch training designed for regular failures in GPU environments.
ChipsChips
OpenAI and MHESI launched an accelerator for 10 startups
OpenAI and MHESI launched an eight-week program in Thailand for 10 startups in health, well-being, and education sectors.
ChipsChips
Nvidia buys Hugging Face for 12,9 billion dollars
Nvidia buys Hugging Face for 12,9 billion dollars on the backdrop of competition around open AI models
ChipsChips
Nvidia is in talks to acquire Hugging Face for 12,9 billion dollars
TechCrunch AI reports on a negotiated purchase of an open AI platform for 12,9 billion dollars, but there is no primary confirmation.
ChipsChips
NVIDIA Expands NVLink Fusion with Custom NVHBM Memory
NVIDIA has linked the expansion of NVLink Fusion to custom NVHBM memory amidst growing demands for AI infrastructure.
ChipsChips
NVIDIA Describes Shadow Engine Recovery for LLM Inference
NVIDIA described Shadow Engine Recovery to restore inference capacity after an LLM process failure, but detailed technical data is not yet available.
ChipsChips
OpenAI Unveils Jalapeño Chip, Surpassing Blackwell and Rubin
OpenAI's custom chip demonstrated superior performance against Nvidia's Blackwell and Rubin in tests recounted by The Decoder.
ChipsChips
NVIDIA Claims New Benchmark for Vera Rubin and Blackwell in AI
NVIDIA links Vera Rubin and Blackwell to a new benchmark for energy efficiency in agentic AI systems, but measurements are not disclosed.
ChipsChips
OpenAI unveiled the Jalapeño chip with InferenceX test results
TechCrunch AI linked OpenAI's new chip to InferenceX test results, but methodological details are currently unavailable.
ChipsChips
NVIDIA describes infrastructure for AI factories
NVIDIA linked the economics of AI factories to the efficiency of the entire infrastructure
ChipsChips
NVIDIA Vera Rubin NVL72 Sets an AI Efficiency Standard
NVIDIA claims a significant efficiency advantage for the Vera Rubin NVL72, but the comparison methodology is not disclosed in the available data.
ChipsChips
NVIDIA Describes the Role of Vera CPU in Agentic Compute Fleets
NVIDIA links compute fleet efficiency for agentic tasks to the role of central processors and workload variability.
ChipsChips
NVIDIA Introduces Groq 3 LPX for Interactive AI Inference
NVIDIA describes Groq 3 LPX as an interactive AI inference accelerator for the Vera Rubin platform, but measurements and independent confirmation are not presented.
ChipsChips
Intel Outlines Architectures for Agentic AI at Hot Chips 2026
Intel linked three architectures to the scaling of agentic AI from server racks to edge devices.
ChipsChips
Nvidia presented research on configuring AI systems
Nvidia showed the role of fine-tuning in the operation of AI systems