News & Insights
A content hub for AI/HPC engineers, research institutions, and industry partners: from the AI-HPC expert perspective, we distill real Q&A, track industry frontiers, and produce actionable insights.
📡 Subscribe: RSS Feed (English) 🗓️ Weekly: every Monday we publish the latest AI×HPC industry watch (bilingual); subscribe to the RSS feed to follow.
Columns
- 🎙️ Curated Q&A: real answers from the AI-HPC assistant, curated into searchable articles—from InfiniBand network design to Slurm heterogeneous scheduling.
- 📊 Industry Watch: weekly AI×HPC industry roundup plus quarterly AI×Science frontier tracking (converged AI×HPC compute, agentic research, validation bottlenecks).
- 📄 Paper Selections & Trend Watch: structured paper deep-dives and AI/HPC trend tracking, with an engineer's lens and actionable judgments.
- 📑 Deep Reports: white papers, benchmark observations, annual trend summaries.
- 🎤 Interviews: conversations with scientists, system architects, and industry leads.
- ⚖️ Policy Briefings: compute infrastructure, data governance, open-source compliance tracking.
Editorial Principles
- Neutral: evidence-based, no marketing-driven conclusions.
- Reusable: every article provides terminology, methods, and reference links.
- Actionable: actionable recommendations for research groups and engineering teams.
Curated Q&A
Real Q&A from the AI-HPC assistant (chat.ai-hpc.org), curated and expanded into searchable content.
- Curated Q&A: How to Design InfiniBand Networks for 10K-GPU Clusters? — topology selection, port/rate planning, RDMA tuning, co-design and acceptance (source: chapter 03-network).
- Curated Q&A: How to Configure Slurm Heterogeneous Scheduling? CPU/GPU Hybrid Cluster Practice — GRES definition, GPU resource management, MIG/vGPU multi-tenant sharing, and Docker submission templates (source: chapters 07/06).
Industry Watch
Every Monday we publish the latest AI×HPC industry watch (bilingual); past articles are tracked via the RSS feed and the sitemap.
- 2026-10-19 (this week) AI×HPC Weekly Watch (2026-10-19): 100K-Card Supercluster Debuts, RMB 4T Compute-Network Investment, NVHBM Standardization, and AI Scores 98% on Theoretical Physics — China's first all-domestic 100K-accelerator supercluster debuts with ten application achievements, the '15th Five-Year' compute network draws RMB 4 trillion in new direct investment, NVHBM supply chains standardize under NVIDIA's spec (Samsung adopts TSMC 3nm for base dies), GPT-6 Astra scores 98% on the TheoPhysEval benchmark, and Token exports and e-commerce categories arrive.
- 2026-10-16 AI×HPC Weekly Watch (2026-10-16): 10K-P Domestic AI Compute Goes Live, the Supernode Density Race, and AI4S Enters Its Verification Era — Pengcheng Cloudbrain III puts 10,000 P of domestic AI compute into trial operation, domestic GPU supernodes push single-rack density to 144 GPUs, a Chinese foundation model cracks the 3D Ghys conjecture with formal verification, Google opens its most powerful models to 200 European scientists, and space compute satellites go into orbit.
- 2026-10 (summer–autumn roundup) AI×HPC Fusion Enters Harvest Season: 2026 Summer–Autumn Industry Roundup — LingSheng tops TOP500, the Sugon 8000 (Dengfeng) 100K-GPU cluster goes live, China's AI datacenter capacity passes 24 GW (2nd globally), Vera Rubin NVL72 begins global delivery, two "no brute-force" Nature breakthroughs, and three industry consensuses.
- 2026 Q4 AI×Science Quarterly Review (2026 Q4 Kickoff): From Discovery to Verification — formal verification as the new frontier, MatBrain lightweight routes, Google RAISE open ecosystem, compute satellites in orbit, and the "trust problem" of scientific AI.
- 2026 Q3 AI×Science Quarterly Review (2026 Q3): From Prediction to Discovery — converged AI×HPC compute, agentic research systems, real breakthroughs in materials/drugs/weather, and validation bottlenecks.