Anthropic Signs First Australian Data Centre Lease While Urging Slower Frontier AI

Anthropic has signed its first data centre lease in Australia, securing a campus with planned total capacity of 2.16 gigawatts. The data centre will be used primarily for AI inference, the operational stage that handles user requests and generates predictions, rather than for model training. The project is being developed by data centre developer Zerra DC, which plans to obtain power through renewable energy purchase agreements and to bear grid connection costs itself, while also planning to use a closed-loop air cooling system. The Australian government plans to impose restrictions on data centre resource use starting in 2027, while continuing to promote investment in AI infrastructure.

Anthropic chief executive Dario Amodei previously published a long article calling on the industry to slow the pace of advancing frontier model capabilities in order to buy more time for safety research and risk governance. OpenAI chief executive Sam Altman and SpaceX chief executive Elon Musk have echoed that call. Nvidia chief executive Jensen Huang rejected "AI doomerism," saying the AI industry "does not need any new laws." Meta chief executive Mark Zuckerberg argued that bringing in third-party evaluators is the industry's best practice, rather than advocating for AI to slow down.

Google DeepMind co-founder Shane Legg said on 16 September 2026 that current AI capabilities are advancing "very, very fast," but that "capability must not get ahead of safety." Regarding Amodei's proposed "slowdown," Legg said the direction is worth considering, while how it would be implemented still requires further discussion.

On 16 September 2026, Microsoft AI head Mustafa Suleyman published an article titled "A Warning About 'Model Welfare'," questioning Anthropic's approach of involving concepts such as "consciousness," "emotion," and "model welfare" in model training. Microsoft is a major investor in and commercial partner of Anthropic. In the article, Suleyman said he recognizes Anthropic's work on AI safety, and that his concerns focus mainly on "model welfare" and the specific path of anthropomorphizing AI.