Annie is seeking infrastructure partnerships with Australian sovereign data centre operators. Rather than building or leasing its own compute capacity, Annie deploys on infrastructure operated by Australian partners — bringing high-value AI inference workloads to sovereign hardware.
The partnership model works as follows:
- Annie's enterprise customers choose their deployment environment at onboarding
- Annie supports deployment on certified Australian sovereign cloud and on-premise environments
- Infrastructure partners host the compute capacity on which Annie runs; they receive AI inference workloads that are high-utilisation, high-value, and long-tenured
- Customers receive verified sovereignty — their AI runs on Australian soil, under Australian law, operated by Australian providers
Infrastructure partnerships are in active discussions with providers across the Australian sovereign cloud market. If you operate or represent a sovereign data centre and are interested in discussing a partnership, contact us via the enterprise briefing request form.
Annie's compute requirements depend on the number of pipeline roles active simultaneously, the scale of the enterprise deployment, and the frequency of Cognition Stream fine-tuning runs.
Indicative minimum requirements for a production deployment:
- GPU — NVIDIA A100 or equivalent; minimum 8× GPU for a full pipeline configuration (more for larger organisations or high-throughput use cases)
- RAM — 512 GB minimum for model weights and inference context
- Storage — 10 TB minimum for model weights, task data, and Cognition Stream artefacts
- Network — high-bandwidth internal networking between GPU nodes; external connectivity is optional
These figures are indicative. The enterprise briefing process produces a detailed infrastructure specification based on your organisation's use case, task volume, and pipeline configuration.
Infrastructure partners hosting Annie deployments for multiple enterprise customers should expect GPU utilisation profiles characteristic of sustained inference workloads — distinct from the bursty patterns typical of public cloud consumption.
Enterprise AI deployments have a very different compute utilisation profile from typical public cloud workloads. For infrastructure partners, Annie generates demand in three ways:
Sustained inference load. Enterprise organisations running production AI workflows generate continuous, predictable inference demand — unlike bursty consumer traffic. This produces high, steady GPU utilisation, which is economically more efficient for infrastructure operators than on-demand cloud patterns.
Off-peak fine-tuning. The Cognition Stream runs overnight fine-tuning cycles. This generates additional compute utilisation during off-peak periods — filling GPU hours that might otherwise be idle — and is scheduled to avoid daytime inference peaks.
Long tenancy. Enterprise AI deployments are sticky. Once an organisation has deployed Annie on a given infrastructure environment and begun accumulating fine-tuning data, migration carries a significant operational cost. This creates long-tenured, stable infrastructure relationships rather than short-term cloud commitments.
For infrastructure partners, Annie represents a route to sovereign AI workloads — a growing category of regulated-enterprise demand that is specifically seeking Australian infrastructure for compliance and sovereignty reasons.