Nvidia, OpenAI and Anthropic line up for the Australian AI data center push

US hyperscalers are running short of power, land and political patience at home, and Australia is selling itself as the escape hatch. Canberra has spent recent weeks courting Nvidia, OpenAI and Anthropic with incentives to anchor AI infrastructure on Australian soil, and the pitch now carries projections of A$150 billion in investment and up to 10 gigawatts of new data-center capacity by 2030. The Australian AI data center push is, in large part, a response to the permitting fights and grid constraints that have made the United States a harder place to build during the current 2026 capacity cycle.

That courtship makes Australia one of the more aggressive Asia-Pacific suitors for hyperscale AI capital in this cycle. The companies have answered with a concession on location, promising to build at remote sites away from the residential communities and local opposition that have stalled data-center approvals across the US. In exchange Australia offers the input now scarcest in AI: reliable power at scale, on a stable grid, in a political climate that has so far welcomed the build-out. Distance from major markets is a cost the companies appear willing to accept in return for approval certainty.

The scale under discussion dwarfs anything Australia has built before. Emails released by the New South Wales government show Anthropic pursued up to 5 gigawatts of capacity in that state alone, roughly three times the size of the country's entire existing data-center estate. OpenAI chief Sam Altman has said Australia could become a data-center capital of the world if it wanted the role, and the 10 gigawatt ceiling attached to the current proposals is the same order of magnitude as the single Ohio campus OpenAI has committed to on a 20-year lease.

Australian AI data center push: where the projections diverge

The headline figures remain pre-deal projections that outrun anything signed to date, and they diverge sharply depending on who is counting. Treasury, in a briefing released by Treasurer Jim Chalmers, estimates that A$150 billion could be spent on data centers in Australia by 2030. Oxford Economics, whose Australian impact work is led by Michael Brennan, models a more conservative A$60 billion by 2030, up from A$20 billion in 2026.

Independent tallies of announced project pipelines run higher still, past A$155 billion. The spread between the most cautious and most optimistic estimates reflects a market that is still largely unwritten: the gigawatt figures are ceilings that proposals could eventually reach, and reaching them depends on power contracts, water rights and state approvals that have not yet been granted. The Australian AI data center push will ultimately be measured in signed leases and energized campuses.

Part of the divergence is a question of scope. A figure in the A$150 billion range describes what the pipeline would be worth if nearly every proposal now in planning moved forward, while Oxford Economics models the spending it expects to actually occur given grid, water and approval constraints. Both readings can be true at once, which is why any single headline number is best treated as a bound on the outcome rather than a forecast.

Nor is the country starting from zero. Australia already operates 162 data centers, with about 90 more proposed, and it can point to a mature ecosystem of grid connections, fiber and operators rather than a greenfield promise. That installed base is part of the argument Canberra makes while competing for the same pool of hyperscale capital that is circling other Asia-Pacific markets in this cycle.

A single proposal shows the size of what is being contemplated. A 1-gigawatt campus planned for western Sydney would rank among the largest data centers on earth and become Australia's biggest single electricity user. The power and water demands of a facility that size explain why the boom is contentious: AI data centers are projected to grow from about 3 percent of national electricity use today to 13 percent by 2035-36, a swing that will compete with the electrification of transport and industry and feed directly into the power prices Australian manufacturers pay. Cooling demand raises the same competition for water that communities in the US cite when they resist these projects.

Canberra attaches strings to the AI gold rush

Australia is not offering its grid as a blank check. The Albanese government has said it will require AI giants, including Microsoft, Google and Anthropic, to make their Australian capacity available to local startups and innovators, a so-called sovereign AI bargain intended to keep part of the economic upside onshore. Suitors are discovering that Canberra is transactional: negotiations cover where facilities sit, how many jobs each project carries and who gets access to the compute they install.

Microsoft and Google are in the same conversations, so the courtship is effectively a contest among five hyperscale players for a finite amount of Australian power. The arrangement also shifts the economics for each participant, since capacity reserved for local firms cannot be monetized at global rates. That is a price Canberra appears willing to impose in exchange for domestic compute security, and the direct winners if the top of the range materializes would be the owners of the land, grid connections and campuses that hyperscalers lease for decades.

A capex cycle that punishes oversized promises

The wider 2026 cycle is a warning against treating any of these totals as settled. Nvidia first floated a $250 billion guarantee behind OpenAI's Ohio campus, scaled that figure back during negotiations, and now provides financing of up to $105 billion for the project. It has also agreed to invest $3 billion in the renewable developer SB Energy to power that campus, a reminder that electricity supply, not chip supply, now decides where AI gets built. Nvidia has become a financial backer of the firms that buy its chips, funding OpenAI, Anthropic and SpaceX, so its courtship by Canberra matters twice over: it decides where processors are sold and, increasingly, where capacity gets financed.

Anthropic's own position explains why its Australian interest is taken seriously rather than read as posturing. The company spent the past year scrambling to line up cloud capacity after demand for Claude Code and its agent tools outran its forecasts, signing third-party deals such as a reported $45 billion agreement with Nscale alongside cloud arrangements with SpaceX and Google. It is also weighing a possible record-setting initial public offering, and investors are asking whether Anthropic can secure enough compute to support its valuation. Its 5 gigawatt ambition in New South Wales reads as one answer to that question.

Why this matters

For Australia the bet is that geography converts into a lasting share of the global AI economy rather than a construction boom that rents out land and power, then leaves. For Nvidia, OpenAI and Anthropic the country is an insurance policy against US power and permitting risk, with the added benefit of a government that states its price up front. The figure to track is the distance between the A$150 billion headline and the A$60 billion baseline, because that gap closes only when proposals become signed contracts and powered campuses.

Related Articles

• Microsoft and Amazon Commit $45 Billion to Australian AI Infrastructure Investment
• Nvidia OpenAI Circular Financing: $250 Billion Guarantee Reshapes AI Infrastructure Risk
• Anthropic Nscale compute deal: $7.5B a year before its IPO

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.