Topic
Explore the latest artificial intelligence news and research.
Versity has added RDMA support to its open-source S3 gateway, enabling direct transfers from tape archives to GPU memory. Amid soaring NVMe prices and server shortages, this low-cost setup aims for 20GBps using 50 parallel drives—how will it reshape storage design for AI training?
Silicon Data's GPU rental index shows B200 climbing to $5.66 per hour, while the same H100 splits into different tiers depending on provider. Here's the framework for comparing hourly rates without conflating them with equipment lifespans or depreciation costs.
The first OCP specification standardizes HBF's host connections, reliability, and software operations. After product samples ship, verification of interoperability and power/durability characteristics will continue on actual hardware running alongside HBM.
StatCounter shows North American Linux web-browsing share exceeding 10%. A closer look at the sudden daily jumps and country-by-country differences shows this can't yet be confirmed as PC market share or a migration away from Windows, and the data remains subject to revision for 45 days.
AI has evolved beyond a mere tool for attackers into a weapon that autonomously hunts targets. As attack automation eliminates the grace period before vulnerabilities are exploited, and as massive alert volumes and new threats targeting AI infrastructure surge, defenders are being forced to build machine-speed countermeasures that go beyond manual response.
Six CVEs alleged against SQLite were rejected within four days—after CISA's supplementary values had already been attached—once verification against the actual code and PoCs failed to hold up. Automated handling requires vendor confirmation.
MiniMax H3 has taken the top spot in a video editing ranking with audio, and the model weights have also been released. However, the local version only goes up to 768p, and the Context-IR and Regenerate modules used in the evaluated 2K system remain undisclosed.
TrendForce revised its 2026 AI server shipment forecast upward to approximately 31% growth. Combined capital expenditure from nine major CSPs exceeds $886.7 billion, with demand expanding into power, cooling, and proprietary ASICs.
Alibaba has announced its latest AI model, "Qwen3.8-Max," boasting 2.4 trillion parameters. By adopting a MoE architecture, it achieves high computational efficiency, and through a self-evolving loop that leverages visual feedback, it possesses a high degree of autonomy capable of completing complex development work and practical tasks over the course of several days without human intervention.
The Pixel 11 lineup that Google is expected to announce will reportedly see a uniform $100 price increase across the board compared to the previous generation. While increased storage capacity and the latest chip are expected, leaks also suggest the standard model may lose its telephoto camera, with reduced memory on certain models and a smaller battery capacity.
