Anthropic seeks $15 billion Australian data centres deal ahead of trillion-dollar IPO

Anthropic Pursues Massive Australian Data Centre Footprint

Anthropic is moving aggressively to establish a major presence in Australia, seeking at least 1.4 gigawatts of Australian data centres capacity in what could become a $15 billion infrastructure investment 2

2

. The AI startup has positioned Australia as its second home for AI model training, marking a strategic expansion beyond its US operations as it prepares for a highly anticipated public offering valued at over $1 trillion 1

1

.

According to confidential tender documents, Anthropic's Anthropic Australian expansion plans target at least 1 GW of capacity operational by the end of 2027, with the company rushing to secure any available AI data center capacity deliverable by mid-next year 1

1

. This aggressive timeline reflects the intense competition for AI infrastructure as companies race to secure the computational power needed to train increasingly sophisticated models.

Major Data Centre Operators Compete for Anthropic Contract

The tender process has attracted Australia's leading data centre developers, with CDC Data Centres, AirTrunk, Nextdc, Iren, and Stack all submitting initial proposals in March 2

2

. Shortlisted bidders met with Anthropic executives in Canberra in April, signaling the government's involvement in facilitating this major AI infrastructure deal. Sources familiar with the discussions indicate Anthropic aims to finalize its first Australian data centre contracts this month 1

1

.

Anthropic's preferred approach involves partnering with a developer capable of constructing a large-scale 1.4 GW-plus data centre campus, though the company remains open to jointly developing a suitable site if existing options prove insufficient 2

2

. The AI startup is expected to reach a final decision within six weeks and may distribute the project across four or five providers rather than selecting a single developer, potentially spreading both risk and opportunity across Australia's data centre ecosystem.

Copyright Deal Critical for Claude AI Model Training

Beyond securing physical AI infrastructure, Anthropic is simultaneously negotiating a copyright deal with the Albanese government that would enable legal training of its Claude AI model in Australia 1

1

. This parallel negotiation highlights the complex regulatory landscape AI companies must navigate as they expand internationally. The copyright agreement would provide Anthropic with the legal framework necessary to process Australian data for AI model training, addressing concerns about intellectual property rights and data usage that have complicated AI development globally.

The timing of these negotiations, concurrent with Anthropic IPO plans, suggests the company views regulatory clarity and infrastructure commitments as essential prerequisites for its public market debut. Establishing Australia as a legitimate training ground for AI workloads could differentiate Anthropic from competitors and demonstrate geographic diversification that investors increasingly value. For Australia, hosting a major AI startup's training operations could position the country as a significant player in the global AI race, potentially attracting additional investment and talent to the region while raising questions about energy consumption, environmental impact, and the sustainability of such massive computational infrastructure.