Google is exploring AI data centres in orbit, and while that sounds distant, the pressure behind it is already visible in Australia. Data centre demand in Sydney and Melbourne is rising quickly, and electricity is the constraint everyone talks about.
What is the Concept
Google calls the research effort Project Suncatcher: solar-powered satellites carrying its AI chips, connected by lasers, running workloads in space. Early prototype launches are planned with a partner for around 2027, so it is an experiment, not a service Australian businesses can buy.
For an Australian SME the useful reading is simple. The biggest AI providers are searching for power, and that search shapes the pricing and capacity of the cloud regions you use, including those in Sydney and Melbourne.
Why It Matters Now (2025–2026 Context)
Australia has become a major data centre market, with large investment concentrated in Sydney and growing activity in Melbourne, Perth and Brisbane. Grid capacity and approval timelines are real constraints, and the Australian Energy Market Operator has flagged data centres as a growing source of demand.
The contrarian view: orbital compute will not lower your AUD cloud bill soon. The costly mistake is assuming scale will fix inefficient AI usage on its own.
How AI Is Changing This
AI agents and copilots multiply the number of model calls each employee triggers. A professional services firm in Brisbane that rolls out AI drafting for fifty staff can see usage compound across the month.
The non-obvious idea is to treat tokens like electricity: meter them by team, not just by company. Visibility changes behaviour faster than policy documents.
Real-World Examples
Australian operators such as AirTrunk and NextDC have announced large expansions, and global providers continue to add capacity here. These projects show demand is real and local.
A realistic scenario: a Melbourne logistics SME uses AI to classify freight documents. By moving routine classification to a smaller model and keeping a larger one for exceptions, it reduces its monthly AUD spend while keeping accuracy.
Practical Insights / Actions
Use the Compute Resilience Ladder: measure, route, cap, diversify. Measure AI cost per completed task, route work to the cheapest capable model, cap agent loops, and keep the option to switch providers.
Also check data residency. If your clients expect data to stay onshore, confirm that your AI vendor processes prompts in Australian regions, and review obligations under the Privacy Act before expanding use.
Future Outlook
Orbital data centres face launch cost, radiation, heat and maintenance hurdles. Expect terrestrial growth, including renewable-backed sites, to meet most Australian demand this decade.
The hidden opportunity is local efficiency. Australian firms that run lean AI workflows will be better placed if power-driven price pressure continues.
Conclusion
Google's orbital plan is a signal that AI is energy-bound. Respond by measuring and trimming your own consumption rather than waiting for a technological leap. RP SoftTech can help Australian teams audit their AI workloads and set sensible cost controls.