Mobilne mreže News: 2 септембар 2025 - 7 октобар 2025

November 18, 2025, 5:22 AM EST. Google Fi Wireless is rolling out AI-enhanced audio on Android and iPhone (iOS 11+). The feature, complementing recent HD/HD+ calling, uses Google AI to filter background noise and clarify speech on both ends for more natural conversations. It's enabled by default and can be turned off in the Google Fi app under account > Phone settings > Call quality. On Android, the Phone by Google screen shows AI-enhanced by Google Fi, plus a notification shade alert. Changes take effect on the next call. Privacy note: audio is processed on Google servers to improve quality (see Privacy Policy). Live with Google Fi app version 146 on Android.

November 18, 2025, 5:20 AM EST. Stone Fox Capital weighs whether AI hype has inflated NVDA stock beyond fundamentals. The analysis contrasts AI momentum with NVDA's fundamentals, discusses potential catalysts and risks, and notes that the author currently has no position but may initiate a long position or call options on NVDA within 72 hours. The piece emphasizes transparency through disclosures and the Seeking Alpha caveat about past performance. It offers a cautious lens on AI-driven momentum, advises readers to do their own research and consider diversification and risk. Overall, the article frames NVDA as a high-visibility tech name where sentiment can swing with AI news and quarterly results.

November 18, 2025, 5:18 AM EST. Global supercomputing centers are embracing NVIDIA NVQLink, a universal interconnect that tightly couples quantum processors with GPU computing to enable hybrid quantum-classical workflows. The rollout across Asia, Europe, and the U.S. leverages CUDA-Q to connect quantum control with GPU acceleration, addressing low-latency, high-throughput needs. NVIDIA reports a throughput of 400 Gb/s and latency under four microseconds, enabling scalable real-time decoding for quantum error correction as demonstrated with Quantinuum's Helios processor. This open architecture unifies diverse quantum hardware with existing HPC infrastructure, accelerating research in materials science, cryptography, and beyond.

November 18, 2025, 5:14 AM EST. Japanese AI startup Sakana AI has raised ¥20 billion ($135 million) in a Series B at a $2.635 billion valuation to scale frontier research and drive applied AI in Japan. Founded in 2023 by former Google Brain researchers, the company pursues biologically inspired, efficient AI that eschews brute-force scaling. Its self-evolving and model-merging approaches aim to enable smaller models to synthesize capabilities, transfer knowledge and recombine skills with fewer large training runs, aided by automated architectural search. Sakana AI plans to tailor foundation models to Japan's linguistic, cultural, and regulatory environment for enterprises, finance, manufacturing and government. The round included MUFG, Khosla Ventures, NEA, Lux Capital and others, with MUFG stressing AI implementation across Japanese firms. The firm seeks to blend research with real-world deployment and a sustainable ecosystem.

November 18, 2025, 5:12 AM EST. Nvidia-backed efforts are expanding supercomputing worldwide. In the U.S., three Nvidia-accelerated systems-Minerva, Janus, and Tara-will support AI inference and workforce development. At LANL, the new Mission and Vision systems from HPE will run on the Nvidia Vera Rubin platform and Quantum-X800 InfiniBand networking; Mission handles classified work for the NNSA, while Vision enables open science, including foundation models. In Europe, the Jülich Supercomputing Centre's JUPITER exaflop system joins the Top500 echelon, powered by 24,000 GH200 Grace Hopper chips and interconnects. In Japan, RIKEN plans two Nvidia-powered machines: one for science with 1,600 Blackwell GPUs on the GB200 NVL4 platform, and a second quantum system with 540 GPUs, tied to Quantum-X800 InfiniBand.