Megaport creates GPU Pool, announces new contracts worth $458m

Megaport, a Queensland-headquartered automated infrastructure platform, has announced the creation of an on-demand GPU Pool.

Megaport aims to establish a globally-distributed AI inference cloud, of which the GPU Pool will form a foundational component, providing enterprise customers with access to high-performance GPUs, alongside the CPU, memory, storage, and networking infrastructure required to support production AI inference deployments.

Supported by A$350 million of investment, which will be commercialised through consumption-based pricing models, the platform will provide automated provisioning, billing, and operations.

The GPU Pool serves three objectives: attracting new customers seeking immediate access to AI infrastructure; providing existing customers with flexible burst capacity; and creating a pathway to convert short-term usage into longer-term contracted deployments.

By positioning infrastructure closer to end users, Megaport intends to address the need for ultra-low latency in AI inference, real-time workloads, and sovereign AI deployments, while supporting both contracted and on-demand consumption models.

It is expected to take six to nine months to procure and deploy all servers, then a further three to six months for the servers to ramp to optimal utilisation from deployment. Once they have achieved optimal utilisation, they have a 16-22 month payback target to recover the capex investment.

The company also announced the execution of four new AI infrastructure contracts with unnamed US-based technology providers running AI applications and inferencing workloads, representing a mix of new and existing customers.

The four committed contracts have a combined Total Contract Value of approximately A$458.9 million, irrespective of usage, expected to commence in H1 FY27. The contracts have a mix of terms, with c.55% of ARR and c.70% of TCV generated by contracts with terms of 36 months

The four signed contracts require approximately A$369.5 million in incremental capital expenditure primarily for high-performance NVIDIA GPUs, supported by Network and Storage components. Delivery of all hardware is expected in FY27, with deployment commencing on a phased basis once delivered.

To fund the capital expenditure required to fulfil these contracts and seed the GPU Pool, Megaport is undertaking a fully underwritten entitlement offer of new ordinary Megaport Shares to raise A$827.3 million.

“AI inference represents one of the biggest infrastructure opportunities of the next decade,” said Michael Reid, Megaport’s chief executive.

“The contracts announced today reflect the accelerating demand for globally distributed AI inference infrastructure. Megaport’s software-provisioned compute, network, and storage platform positions us strongly to meet that demand.

“The proceeds from the Entitlement Offer will enable us to fulfil contracted customer demand while building an on-demand GPU Pool that creates new opportunities across enterprise and sovereign AI markets globally."

Last month, Megaport announced that its wholly owned subsidiary Latitude.sh secured three major GPU, CPU, network and storage contracts with a combined TCV of A$254 million.