Australian networking firm Megaport announced a A$827.3M ($594M USD) equity raise on Wednesday to fund a globally distributed AI inference cloud, alongside four new AI infrastructure contracts worth a combined A$458.9M ($329M). The move marks the company’s formal pivot from network connectivity into GPU compute.
The Contracts and the Capital
All four contracts are with US-based technology providers running AI applications, according to TNW. Revenue starts in H1 2027. The contracts alone require an estimated A$369.5M in capital expenditure, primarily for high-performance Nvidia GPUs plus networking and storage infrastructure.
The raise is a fully underwritten 1-for-3.08 entitlement offer at A$14.30 per share, a 13.9% discount to Monday’s A$16.61 close, according to Stocks Down Under. It represents roughly a 32% increase in shares on issue.
Reuters reported that the move marks a significant step in Megaport’s push into AI infrastructure, with the company betting on surging demand for GPU-based compute as enterprise AI adoption shifts from model training to latency-sensitive inference workloads.
The GPU Pool Strategy
Beyond the contracted deployments, Megaport is committing approximately A$350M (~$252M USD) to build an on-demand GPU pool offered on consumption-based pricing. Until now, the company only deployed GPU hardware when a customer signed a 24-to-36 month commitment. The pool flips that model: Megaport takes on utilization risk directly, betting that enterprise burst demand and sovereign AI workloads will absorb capacity at premium rates.
Management is guiding to a 16-to-22 month payback at optimal utilization. The pool will be deployed across Megaport’s existing footprint of more than 1,100 connected data centers in 31 countries over the next six to nine months.
The Infrastructure Thesis
The geographic angle is the core differentiator. Most GPU capacity today sits in a handful of enormous data centers optimized for training. Megaport is targeting inference, which benefits from being close to the user. A distributed network of smaller GPU pools, spread across data centers the company already connects, fits inference workloads better than centralized mega-campuses and slots into the gap between hyperscaler clouds and single-location GPU specialists like CoreWeave and Lambda Labs.
Pro forma group annual recurring revenue now sits at A$662.9M, with compute ARR alone at A$385.2M, a 6.4x jump since the Latitude.sh acquisition. The core network business is also accelerating: network ARR for April 2026 hit A$277.7M, up 25% year-over-year on a constant currency basis.
The Execution Window
This is Megaport’s second major capital raise in 18 months, following the November 2025 Latitude.sh acquisition. Investors who bought into that raise are being asked to write a second, larger check before the first has fully proven itself. CFO Leticia Dorman indicated that debt will become a more permanent part of the capital structure going forward, suggesting this should be the last large equity ask for some time.
Hardware delivery runs through FY27, with ramp to optimal utilization expected by late calendar 2027. The decisive question is whether speculative GPU deployment can attract enough on-demand and sovereign workloads to justify the utilization risk before the inference compute market gets crowded by hyperscalers building their own distributed capacity.