News
Updates, product notes and practical coverage for AI platforms and operators.
Meta has published a discussion with infrastructure chief Santosh Janardhan about why the company builds its own data centres for AI. The account highlights power planning, chip selection and custom infrastructure as strategic product inputs, underscoring how frontier AI competition increasingly depends on long-term physical capacity rather than software and models alone.
NVIDIA says its Vera Rubin NVL72 preview achieved up to 3.7 times GB300 NVL72 throughput on Qwen3-VL in MLPerf Inference v6.1. The same release reports strong Blackwell rack scaling and software gains. These are workload-specific benchmark results, not a universal production speed or cost guarantee.
Google has launched agentic video understanding for Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite. The API feature lets models actively inspect video frames, audio and transcripts, aiming to cut token use and analysis cost while improving retrieval, counting and anomaly-detection tasks for developers working with long recordings.
NVIDIA has added NVHBM to its NVLink Fusion platform, putting the memory controller inside the HBM stack. The company says the design can improve bandwidth and power efficiency while giving cloud and AI-chip builders a more standard route to custom, rack-scale systems for enterprise AI teams.
NVIDIA says its Groq 3 LPX inference system is now in full production alongside Vera Rubin NVL72. The company positions the LPU-based platform for faster token generation and long-context responsiveness in agentic applications, citing early deployments at Nebius and CoreWeave and a tighter integration with its wider AI-factory stack.
Firebird has opened an NVIDIA-powered AI factory in Armenia that the companies describe as the CIS region’s largest. The project combines Dell systems, NVIDIA DSX infrastructure and planned Rubin and Blackwell GPU deployments, signalling a push to build AI capacity closer to regional research, enterprise and public-sector users.
Baseten has launched a distribution and monetisation platform for developers of closed-weight AI models. Baseten for Model Labs combines managed inference, Model Library distribution and commercial support, giving specialist labs another route to production customers without building a complete serving business themselves.
OpenAI has reduced API prices for GPT-5.6 Luna and Terra and introduced Fast mode for GPT-5.6 Sol. The changes lower the cost of high-volume workloads while giving developers a new paid option for faster processing on latency-sensitive tasks.
Notion has acquired AI infrastructure company ZeroEntropy and is bringing its full team into the business. The deal creates a new Model Research group and builds on reranking technology that Notion says has already made its unified search substantially faster.
NVIDIA has introduced Spectrum-6, a 102.4-terabit-per-second Ethernet switch system designed for Vera Rubin AI factories. Early deployments are planned by major cloud and infrastructure operators, but NVIDIA’s performance, efficiency and reliability figures remain vendor claims that buyers should validate against their workloads.
NVIDIA says its Vera Rubin rack-scale AI platform is entering full production across a global partner network. The company reports large gains in throughput per megawatt and token cost, alongside new deployments by cloud and model providers, though independent workload testing is still needed.
Bristol Myers Squibb is deploying a second NVIDIA DGX SuperPOD built from eight Vera Rubin NVL72 systems. The pharmaceutical group plans to make the unified environment available across its research organisation for model training, predictions and agentic drug-discovery workflows using NVIDIA BioNeMo tooling.
Meta is expanding its Louisiana data centre to 5GW of compute capacity as part of a US$50 billion AI infrastructure investment. The project is also reshaping the local economy through jobs, supplier contracts, education funding and major infrastructure upgrades.