Megaport raises $594M to build AI inference cloud, secures four US contracts
What's the deal? Australian network infrastructure company MegaportDealroom has a profile for this one. Try Dealroom → is raising US$535.3M to build a globally distributed AI inference cloud and has secured four new AI infrastructure contracts worth US$296.9M with US-based technology providers.
The contracts are expected to start in the first half of 2027 and require nearly US$239.1M in capital expenditure — primarily for high-performance Nvidia GPUs, network, and storage infrastructure.
Megaport plans to set up an on-demand GPU pool backed by US$226.5M in investment, offered to enterprise customers through contracted and consumption-based pricing. Under the entitlement offer, it will issue shares at US$9.25 each — a 13.9% discount to the stock's last close.
Why now? Enterprise AI adoption is shifting from model training to latency-sensitive inference workloads, creating a surge in demand for distributed GPU compute. Megaport is betting its network of more than 1,100 data centres in 31 countries gives it a natural advantage in delivering AI compute closer to end users.
"Megaport is positioning itself as a picks-and-shovels player in the AI gold rush — benefiting from the next wave of AI infrastructure buildout," said Hebe Chen, a market analyst at Vantage Markets.
What could go wrong? The company is making a substantial bet on inference demand materialising at scale by 2027. GPU prices could shift, competition from hyperscalers is intense, and enterprise AI spending patterns remain hard to predict. The 13.9% share discount also signals dilution risk for existing shareholders.
The signal: Megaport's pivot from pure connectivity into GPU compute mirrors a wider trend of infrastructure players moving up the value chain to capture AI-driven demand. With four US contracts already locked in before the inference cloud is even built, the company is betting that its footprint across 1,100+ data centres gives it a distribution edge that hyperscalers — concentrated in fewer, larger facilities — can't easily replicate.
Read more: The Star
