AI news, 19 July: China’s Moonshot unveils giant open model rivalling US leaders

Moonshot AI launches Kimi K3, claiming near-parity with top US systems. Chinese startup Moonshot AI unveiled Kimi K3, described as the world’s largest open AI model with about 2.8 trillion parameters. The company and independent evaluators say it matches or approaches leading models from OpenAI and Anthropic on key tasks such as coding and reasoning, according to The New York Times, BBC and AP. Moonshot plans to release the full open weights later this month (around 27 July), which would let anyone download, run and customise it. The announcement, coinciding with China’s World AI Conference and remarks by President Xi Jinping promoting open international AI cooperation, prompted market jitters and renewed debate over how quickly China is closing the gap despite US chip export controls. Full independent verification of all claims will come with the open release.

Nationwide US protests target AI data-centre boom. On 18 July, the grassroots group HumansFirst coordinated demonstrations at more than 125 sites across dozens of states—the first large national day of action against rapid data-centre construction. According to Reuters, protesters cited higher electricity bills, water use, local secrecy around deals, and lack of community control. A recent Reuters/Ipsos poll found only 14% of Americans would support an AI data centre in their own community. Organisers described the issue as non-partisan; industry groups have previously said they aim to be responsible neighbours. The actions highlight growing public tension over the physical footprint of AI infrastructure.

Databricks nears $188 billion valuation in new funding. Data-and-AI company Databricks announced it has signed a term sheet for a strategic funding round that values it at $188 billion, led by existing investor Coatue, with the deal expected to close later this summer. Reports from the Wall Street Journal, Reuters and TechCrunch put the raise at roughly $3 billion and note a roughly 40% jump from its prior valuation. The move underscores continued strong investor appetite for firms supplying the data and AI platforms that enterprises use to build and run AI applications, even as some market valuations face scrutiny.

Alibaba’s T-Head open-sources AI chip software stack. At the World AI Conference in Shanghai, Alibaba’s chip unit T-Head said it is open-sourcing SAIL, the full software stack for its Zhenwu AI processors. Reporting by the South China Morning Post says the move aims to make it easier for developers to adopt the hardware and reduce reliance on Nvidia’s dominant CUDA ecosystem. T-Head claimed programmers could adapt SAIL to mainstream AI frameworks in under a week. The step is part of broader Chinese efforts to build alternative AI computing stacks amid export restrictions, though real-world adoption and performance parity remain to be proven at scale.