↗
Qualcomm and Amazon Team Up on AI Data Center Silicon
Qualcomm and Amazon will develop custom silicon for AWS AI inference while testing optical links that can reach 1.6 terabits per second.
CH CHANNEL ARCHIVE
Tag: AI
Artificial intelligence models, systems, and products.
01 INCOMING
ARCHIVE
↗
↗
Arm Opens AI Portal for Optimized Models and Tools
Arm's new AI Portal lets developers and coding agents discover optimized models, performance data, and deployment workflows across cloud, edge, and physical AI.
↗
Google Puts AI Contrail Forecasts Into Cathay Pacific Flights
Google and Cathay Pacific are expanding an AI trial that helps ultra-long-haul flights avoid warming contrails using satellite and weather data.
↗
AMD Shows Threadripper Halo Station for Local AI
AMD's Threadripper Halo Station is a 2027 prototype with up to 2.6TB of memory and 576GB of HBM3e for local AI workloads.
↗
Databricks Proteus Specializes GPU Kernels for AI Inference
Databricks says its Proteus system generates and validates GPU kernels for specific inference shapes, with speedups reported on Qwen 3.5 122B.
↗
Extropic Adapts Transformers for Its Probabilistic Z1 Chip
Extropic describes Z1T, a sparse transformer research model built for probabilistic Z1 hardware, with early energy and latency estimates.
↗
GlobalFoundries Tapes Out GCRAM Test Chip for Edge AI
GlobalFoundries and RAAAM taped out a GCRAM test vehicle on FD-SOI, targeting denser, lower-power on-chip memory for edge AI.
↗
Google Introduces Gemini 3.8 Flash and Flash Cyber
Google launches Gemini 3.8 Flash for long-running coding work and Flash Cyber for controlled vulnerability discovery and patching.
↗
OpenAI Says Astra Crosses Its Critical Cyber Threshold
OpenAI says its upcoming Astra model can find and exploit unknown flaws, so advanced cyber capabilities will launch behind tighter controls.
↗
Gemini Adds Agentic Video Analysis to Cut Token Costs
Google’s new Gemini video mode searches only the moments that matter, cutting token use by up to 88% in the company’s benchmarks.