AI News from September 3, 2026

Nvidia is acquiring Hugging Face for $12.93 billion, pairing its AI computing reach with the leading open-model platform used by millions of developers.

Meta’s Muse Spark 1.3 improves agentic and coding scores, but its low per-task cost may matter more than its still-unfinished frontier challenge.

NVIDIA says Vera Rubin NVL72 can deliver up to 30x more agentic AI throughput per megawatt than GB300, targeting long-context inference costs.

A 36Kr cluster maps China’s AI industry and reports a 1.4 billion yuan Keling investment, while limited evidence leaves terms and context unverified.

MBZUAI’s Institute of Foundation Models says K2 Horizon opens weights, code, data and methods, challenging closed AI on reproducibility.

Moonshot is reportedly pursuing a $3 billion Hong Kong IPO as Kimi K3 raises the stakes for Chinese AI firms and open-model rivals before US listings.

Marvis has launched a custom model feature with support for Kimi and Zhipu GLM, signaling a broader push toward multi-model AI integrations.

Meta has launched Muse Spark 1.3, citing better coding and agentic-task performance as it competes with OpenAI and Anthropic.

TechCrunch Disrupt 2026 will bring founders and investors together to examine fundraising, hiring, AI agents, and the hard realities of startup scale.

A Semianalysis report links South Korea’s proposed sovereign AI push to Nvidia demand, while raising questions about SK Hynix’s role and evidence.