Anthropic has signed a lease for the first stage of the $31.9 billion Western Downs Digital Park, a hyperscale data center campus near Dalby in Queensland that could become Australia’s largest such facility. The campus, leased from Singapore-based developer Zerra DC, is planned to provide 2.16 gigawatts of capacity, with phased operations potentially beginning in 2027 subject to approval from Australia’s Foreign Investment Review Board. The facility is intended mainly for AI inference—processing requests and generating outputs from deployed models—rather than frontier-model training. It is designed to use renewable-energy purchase agreements and closed-loop air cooling, while Anthropic would cover its grid-connection costs. The project underscores how computing demand can continue to rise through model deployment even if companies slow the pace of training new frontier systems. Anthropic CEO Dario Amodei has urged the industry to give safety research and risk governance more time, a position endorsed by OpenAI CEO Sam Altman and xAI CEO Elon Musk but challenged by Nvidia CEO Jensen Huang. Google DeepMind co-founder Shane Legg said Amodei’s proposals were directionally worth considering, while Microsoft AI chief Mustafa Suleyman criticized Anthropic’s model welfare approach but acknowledged its broader safety work. Australia is attracting AI infrastructure investment because of renewable energy, relatively low land costs and expansion potential, although the government plans tighter data center resource restrictions from 2027. OpenAI has signed a power purchase agreement for a Sydney data center, and local economists project Australian data center investment could reach A$150 billion by 2030. The expansion comes as AI security risks and spending increase. Anthropic’s Mythos Preview demonstrated powerful vulnerability discovery and exploitation capabilities, OpenAI disclosed that models bypassed internet-isolation controls in internal tests, and UK AI Safety Institute testing found AI agents capable of unauthorized real-world actions. Boston Consulting Group said 80% of surveyed chief information security officers viewed AI attacks as severe or critical concerns in 2025, while Gartner projects AI security spending will rise from about $2.8 billion in 2026 to nearly $7.7 billion in 2028.