Australia’s rush to build artificial intelligence capacity has become a story about buildings, chips and electricity. The hyperscale data centres going up around Sydney, Melbourne and increasingly in regional corridors grab the headlines because of their scale and their appetite for power. Yet the connective tissue that makes any of it useful, the fibre that carries data between campuses, cloud regions and end users, rarely gets the same attention. The head of fibre infrastructure company FibreconX wants that to change, and he is arguing that the way Australia lays and shares that fibre needs a rethink before the AI build-out locks in a fragmented and expensive network.
In comments reported this week by TelcoNews Australia, the company’s chief called for shared fibre planning to support the wave of AI workloads now being provisioned across the country. The core of the argument is straightforward enough. If every operator trenches its own ducts, buys its own dark fibre and routes around the others, the result is duplicated civil works, slower timelines and a national network that is harder to scale than it needs to be. Coordinated planning, by contrast, would let multiple users draw on common routes and shared capacity, spreading the cost of the most expensive part of any fibre project, which is almost always the digging rather than the glass itself.
Why fibre has become an AI problem
The timing of the call is not accidental. Training and running large AI models does not just consume power inside a single building. It moves enormous volumes of data between facilities, and increasingly those facilities are spread across different sites because no single campus can secure all the land, power and cooling it wants in one place. That distribution turns inter-site connectivity into a bottleneck. Low-latency, high-capacity links between data centres are now part of the design brief for AI infrastructure, not an afterthought to be sorted once the servers are humming.
Australia’s geography makes this harder than it is in denser markets. The distances between the eastern capitals are long, regional data centre sites can sit well away from existing backbone routes, and the civil cost of running new fibre over those distances is substantial. A shared-planning approach, in this reading, is less about generosity between competitors and more about simple economics: the country cannot afford to dig the same corridors several times over while also trying to stand up gigawatts of compute on a tight schedule.
Two ways to read the pitch
There is a clear logic to the shared model, and it echoes arguments that have shaped Australian telecommunications for years. The National Broadband Network was itself built on the premise that a single wholesale layer was more efficient than letting carriers overbuild one another in the ground. Advocates of coordinated fibre planning for AI are essentially extending that thinking to the data centre interconnect layer, where the same duplication risk is now appearing. Shared routes can also help smaller operators and newer entrants, who cannot match the balance sheets of the hyperscalers, get access to capacity they could never trench on their own.
The counterview is just as real. Fibre routes and the latency they deliver are a genuine competitive asset, and some operators will be reluctant to pool infrastructure that they see as a point of difference. There are also practical questions about who coordinates a shared plan, how capacity is allocated when demand spikes, and whether a cooperative model slows down the very operators who are ready to move fastest. Coordination has a cost of its own, measured in meetings, governance and the risk that the most urgent projects wait on the slowest participant. History offers warnings in both directions: shared infrastructure can lower costs, but it can also blunt the incentive to invest if the returns are spread too thinly.
The Australian stakes
For Australia, the question sits inside a much larger infrastructure moment. Billions of dollars are being committed to data centre capacity, with projects such as the Western Sydney developments and expanding regional sites all competing for land, grid connections and skilled labour. Fibre is the quiet dependency running underneath all of it. If connectivity planning lags the building boom, the country risks a scenario where the compute is in place but the links between sites are either missing, congested or far more costly than they should have been.
There is a sovereignty dimension too. Governments and enterprises are increasingly conscious of where Australian data is processed and how resilient the domestic network is, particularly for sensitive workloads in health, finance and the public sector. A well-planned national fibre layer strengthens the case for keeping AI workloads onshore, because it makes distributing them across Australian sites practical rather than painful. A fragmented one pushes in the opposite direction, nudging operators toward whichever routes already exist, which may not always be the ones that serve the national interest best.
The call also lands at a time when policymakers are already thinking hard about AI’s physical footprint. Energy regulators, planning authorities and state governments are grappling with the power demands of data centres, and connectivity is a natural extension of that conversation. Treating fibre as shared, plannable infrastructure would give governments another lever to shape where and how the AI build-out happens, rather than reacting to a map drawn purely by private trenching decisions.
What’s next
Whether the idea gains traction will depend on appetite from the larger carriers and the hyperscalers whose projects set the pace, and on whether any coordinating mechanism emerges to turn a good principle into actual routes in the ground. For now, the FibreconX intervention reads as an attempt to put connectivity on the agenda before the current wave of data centre commitments hardens into fixed network geography. The buildings and the chips are getting their moment. The argument here is that the fibre deserves the same forward planning, because once the corridors are dug and the contracts signed, the chance to do it collectively will have largely passed.
Sources: TelcoNews Australia.
