Stargate has become the clearest warning flare in the AI boom, as Norway, Australia and a handful of hyperscalers turn the race for compute into a high‑stakes battle over who will own, power and ultimately control the global inference economy.
The UK’s 2025/2026 Cyber Security Breaches Survey shows 43% of businesses and 28% of charities reported a cyber incident in the past year. The headline is not just persistence; it is operational exposure. Phishing remains the dominant route in, education is absorbing heavier pressure, and supplier-r
SAP npm packages poisoned with credential-stealing malware in "Mini Shai-Hulud" attack. Malicious preinstall hooks harvest GitHub tokens, cloud keys and CI/CD secrets. Attackers weaponise AI agent configs for persistence, turning Claude and VS Code settings into execution paths.
Stargate has become the clearest warning flare in the AI boom, as Norway, Australia and a handful of hyperscalers turn the race for compute into a high‑stakes battle over who will own, power and ultimately control the global inference economy.
They are writing the contracts, steering national grid plans and, increasingly, deciding how much diversity developers and enterprises can realistically have in the supposedly “open” AI era. Stargate, Norway and Australia are no longer separate stories; together they show how a small cadre of hyperscalers is becoming the de facto ownership class of the global inference economy.
Stargate began as a moonshot that matched the mood of peak AI exuberance: a 500‑billion‑dollar joint venture to build about 10 gigawatts of data‑centre capacity in the United States, marrying OpenAI’s models with SoftBank’s capital, Oracle’s cloud footprint and Abu Dhabi’s MGX money. It was pitched as the AI equivalent of a national highway system, only with teraflops instead of asphalt. On paper, the structure looked neat, with OpenAI and SoftBank holding the big equity, Oracle and MGX taking smaller slices, Masayoshi Son as chair and Sam Altman as operational commander.
The first phase in Texas set the tone. Five new US sites were announced in 2025, lifting planned capacity to nearly seven gigawatts and more than 400 billion dollars in committed spend on the way to that 10‑gigawatt, half‑trillion target. SoftBank’s SB Energy was contracted to deliver a 1.2‑gigawatt solar‑plus‑AI campus in Milam County, explicitly sold as proof that scale and sustainability could coexist in one package. It was Silicon Valley ambition written in steel and concrete.
Then the frictions started to show. Governance and financing questions about who carries the debt, who has first call on capacity and who gets paid first if demand disappoints turned the joint venture from a clean equity story into a tangle of competing interests. The grand structure that once promised to bind a handful of big names into an enduring AI infrastructure cartel has quietly loosened into a more prosaic world of leases, bilateral deals and off‑balance‑sheet vehicles. Stargate still exists as a banner, but much of the real action now happens in contracts between OpenAI and its landlords.
Norway was meant to be Stargate’s European showcase. A petrostate that exports hydrocarbons yet runs almost entirely on domestic hydropower at home, it offers that irresistible mix of cheap, clean electricity and cold climate that data‑centre developers dream about. TikTok’s 90‑megawatt build and a Google campus heading toward 240 megawatts helped turn it into a Nordic hub for cloud and AI workloads. Against that backdrop, a 230‑plus‑megawatt Stargate‑linked project in Narvik, via Nscale, inevitably hit the political radar.
The Storting’s response was telling. Parliament asked the government to examine the “wider effects” of data‑centre expansion on employment, value creation, national security and the grid, and pushed for a ban on new crypto‑mining facilities. MPs complained that data‑centre customers are “often unknown” to local authorities, which they described as untenable when allocating scarce power, land and water. What had been treated as an unalloyed opportunity suddenly became a source of strategic anxiety.
OpenAI’s move was equally telling. It stepped back from renting Narvik directly and signalled that it would instead work with Microsoft to secure compute in Norway “similar to our operations in other regions”. The facility, the megawatts and the wires are still there, but ownership and grid‑facing risk now sit with Microsoft, while OpenAI becomes a tenant. For all the talk of transformational joint ventures, the practical end‑state is familiar: a hyperscaler with the balance sheet holds the keys and the AI lab rents.
Australia is playing out a parallel drama in fast‑forward. It has gone from peripheral node to what several brokers now rank as the second most attractive data‑centre investment destination after the United States. Built capacity sits around the mid‑gigawatt range and is set to double within a few years. More than 250 facilities are already operating, with scores more in the planning system. Sydney’s “cloud alley” and Melbourne’s western fringe are being remade as AI and cloud corridors at astonishing speed.
The cast of characters is broad and ambitious. Domestic champions like NEXTDC, Macquarie Data Centres and CDC are scaling hard. Global incumbents such as Equinix, Digital Realty, AirTrunk and Global Switch are deepening their footprints. Property groups like Goodman have reoriented their development books toward “digital estates” with hundreds of megawatts of planned IT power. At the edge, a new wave of neocloud and infrastructure‑AI players is appearing. Firmus, for example, is using its Project Southgate partnership in Australia to scale towards gigawatt‑class AI‑optimised capacity, while newer outfits badge themselves as “AI clouds”, selling integrated stacks of compute, storage and model serving to enterprises that might once have gone straight to the US giants.
Downstream, that shapes the choices available to developers and enterprises who say they want diversity. A world in which a small cadre of hyperscalers underwrite most of the bricks‑and‑mortar AI campuses, hold the longest‑dated contracts and control the pipeline of advanced GPUs is a world where multi‑cloud can easily collapse into “different badges on the same handful of landlords”. Neocloud players may innovate at the edges with better latency or integrated tooling, but if their economics depend on long‑term hardware leases and wholesale cloud contracts from Microsoft, AWS or Google, the real leverage lives elsewhere.
Microsoft’s latest announcement in Australia crystallises that shift. The company has unveiled a A$25 billion programme through 2029, its largest investment in the country, to expand Azure AI and cloud capacity, strengthen cyber‑defence and train millions of Australians in AI skills. It plans to more than double local AI infrastructure, effectively writing itself into the centre of Australia’s digital story for the next decade. For Canberra, the package ticks boxes because it promises jobs, skills and security. For the market, it raises the bar for everyone else. Competing with, or even just co‑existing alongside, a landlord of that size requires either partnering on its terms or finding niches it does not yet care about.
The Commonwealth has started to grapple with that tension. March’s “Expectations of data centres and AI infrastructure developers” sets five national principles, including contributing to grid reliability and water security, supporting local innovation and “enabling access to compute for Australian start‑ups, researchers and not‑for‑profits on favourable terms”. The Australian Energy Market Commission has followed with draft technical standards that would force large data centres to be more grid‑friendly, flagging the risk that inflexible loads could undermine stability and affordability. In parallel, analysis commissioned by the market operator has exposed a yawning gap between connection requests, around 44 gigawatts, and plausible demand, perhaps six gigawatts this decade, warning of “phantom demand” fuelled by AI hype more than hard usage curves.
Norway’s debate is, at heart, the same one in a different accent. How many hundreds of megawatts of AI load can a small, renewables‑heavy grid absorb before other priorities, such as electrified transport, industrial decarbonisation and basic household resilience, start to lose out? Who gets to decide which projects count as “strategic” and which are just speculative bets dressed up as national infrastructure? And how comfortable are voters with the idea that the decisive voice in that conversation may belong not to their own utilities or regulators, but to a handful of global platforms and their shareholders?
From the macro balcony, institutions are still politely upbeat. Global forecasters talk about AI‑driven investment propping up growth and promise that productivity gains will, over time, justify the spend. Energy agencies sketch scenarios in which efficiency and demand management keep the lights on despite rising AI and cloud load. Equity analysts, meanwhile, find themselves modelling a strange combination of 1990s‑style infrastructure overbuild risk and 2020s‑style platform dominance.
From an editorial vantage point, 2026 looks more like an inflection point. Norway’s parliamentary interventions, OpenAI’s quiet slide into Microsoft’s shadow at Narvik, Australia’s new expectations and grid rules, the rise of Firmus‑style neoclouds with giga‑scale ambitions and Microsoft’s A$25 billion bet on Australia all suggest the same thing: the AI infrastructure boom is no longer being waved through on the assumption that all compute is inherently good. Governments are starting to ask who owns it, who pays for it and who might be left holding the bill. Investors are starting to ask which projects really earn their cost of capital.
Which leaves a simple, slightly uncomfortable question hanging over the rest of the decade. Are we building a plural, resilient inference economy, where many different providers own, operate and govern the digital factories that matter, or are we sleepwalking into a world where a tiny club of AI landlords writes the cheques, sets the terms and nudges entire national power systems and sustainability plans around its own build schedules? It is the sort of question a weekend anchor might ask with a hint of irony. The harder part, for Norway, Australia and everyone else, will be answering it before the concrete sets and the grid ties are sunk.
References (selected): Microsoft Australia AI investment announcements; OpenAI and partner statements on the Stargate project; Norwegian parliamentary resolutions and legal commentary on data‑centre development; Australian government “Expectations of data centres and AI infrastructure developers”; Australian Energy Market Commission consultation on data‑centre grid standards; AEMO‑commissioned analysis of data‑centre demand; industry research on Australian data‑centre capacity and investment.
Get the stories that matter to you.
Subscribe to Cyber News Centre and update your preferences to follow our Daily 4min Cyber Update, Innovative AI Startups, The AI Diplomat series, or the main Cyber News Centre newsletter — featuring in-depth analysis on major cyber incidents, tech breakthroughs, global policy, and AI developments.
SAP npm packages poisoned with credential-stealing malware in "Mini Shai-Hulud" attack. Malicious preinstall hooks harvest GitHub tokens, cloud keys and CI/CD secrets. Attackers weaponise AI agent configs for persistence, turning Claude and VS Code settings into execution paths.
Australia’s A$25bn AI wager, Bezos’s leap into “physical AI” and Musk’s push to shift data centres into orbit turned this week into a defining moment in the AI global industrial contest, with the Global South emerging as both proving ground and prize in the new AI steel age.
Vercel confirms a security incident after a compromised third-party AI tool's OAuth token allowed attackers to pivot into internal systems, exposing environment variables and API keys across its platform.
Anthropic is scrambling to contain fresh questions over its Mythos AI after online users reportedly accessed the ultra‑powerful model through previously mapped pathways, sharpening Pentagon supply chain concerns and spooking markets already on edge about AI‑driven cyber risk
Where cybersecurity meets innovation, the CNC team delivers AI and tech breakthroughs for our digital future. We analyze incidents, data, and insights to keep you informed, secure, and ahead. Sign up for free!
