• Telstra and Teneo executives said past traffic patterns may provide less guidance for planning networks built to support AI workloads
  • Operators may need closer coordination with customers before deciding where and when to add capacity

The fact

Executives from Telstra and network technology provider Teneo have warned that AI workloads are changing the information operators use to plan infrastructure. Speaking during a Capacity webinar, they said network demand can increasingly depend on where computing infrastructure is located and how applications move data between facilities and users.

Traditional network forecasts draw heavily on existing traffic and expected growth. The speakers argued that this becomes harder when new AI services have not yet established stable demand patterns. Operators may therefore need more information about customers’ planned workloads, locations and performance requirements before deciding where to expand connectivity. The discussion did not include a new Telstra network project, investment commitment or customer contract.

The assessment

AI traffic does not necessarily grow in the same places as existing network demand. It can concentrate around particular data centres, cloud regions or enterprise sites, leaving operators with a location problem as much as a capacity problem. Before approving a new route or upgrade, an operator needs to know which sites must be connected, how much bandwidth is required and when the service must be ready. General expectations of AI growth are not enough to support a specific build.

For BTW readers, the commercial issue is whether customer demand becomes firm early enough to shape network investment. Telstra’s argument only becomes operational when named workloads, locations and delivery dates begin influencing real capacity decisions.

What to watch

Watch for Telstra to apply this approach to a named fibre route, data-centre interconnection project or customer agreement. Details on locations, committed bandwidth, delivery dates and contract terms would show whether AI demand is beginning to shape actual network investment decisions.