Megaport secures A$827 million to develop a distributed AI cloud and pursue opportunities in the inference market.

      Megaport has spent the last ten years as a company facilitating connections to various cloud services. On Wednesday, it revealed its intention to transform into a cloud provider itself. The Australian networking company secured four new contracts for AI infrastructure, totaling A$458.9 million (approximately $329 million), and initiated a fully underwritten entitlement offer aimed at raising A$827.3 million (around $594 million), as detailed in its filing. This funding will support its shift from a plumbing model to a computing focus.

      The contracts are the first priority. All four agreements involve US-based tech companies operating AI applications, are anticipated to commence in the first half of 2027, and will demand nearly A$369.5 million in capital expenditures, primarily for high-performance Nvidia GPUs, as well as networking and storage solutions. This represents a significant commitment for a company of Megaport's size, explaining the substantial nature of the fundraising relative to its operations.

      The capital is primarily directed towards the strategy underpinning these contracts. Megaport plans to establish a globally distributed AI inference cloud, supported by an on-demand pool of GPUs that is expected to be underpinned by approximately A$350 million in investment, which will be available to enterprise clients under both contracted and consumption-based pricing models.

      The GPU pool is set to be implemented across Megaport's existing network of over 1,100 connected data centers in 31 countries, with deployment scheduled to occur over the next six to nine months.

      The core of EU technology is highlighted by the latest developments from the EU tech ecosystem, a narrative from our insightful founder Boris, and some controversial AI-generated art. It’s available for free every week in your inbox. Sign up now!

      The strategy is geographically focused. The majority of GPU capacity is currently concentrated in a few massive data centers optimized for training large models. Megaport is directing its attention to the other half of the AI workload: inference, which involves executing a trained model to respond to queries, benefiting from proximity to the user.

      Megaport's proposition is that a distributed network of smaller GPU pools, strategically positioned across the data centers it already connects, is more suitable for inference compared to centralized mega-data centers, thereby filling a niche between hyperscale cloud providers and specialist single-location GPU companies.

      This approach reflects a realistic understanding of the future of AI infrastructure. As models transition from research prototypes to actual products integrated into various applications, the financial dynamics will shift from training to serving, with serving favoring close proximity and distribution.

      Megaport already possesses the network linking the locations where this computing capacity will reside, providing a genuine structural advantage if this concept proves valid.

      Early coverage surrounding the fundraising presented some confusion regarding the figures, which is important to clarify. The total value of the four contracts is A$458.9 million; the amount to be raised is A$827.3 million; and the commitment toward the GPU pool is approximately A$350 million.

      Several headlines mixed these figures into a singular amount. However, they represent distinct categories: contract values, the funds required to finance them, and the specific investment earmarked for computing within that funding.

      Megaport has also revised its revenue guidance for 2026 to A$307 million–A$315 million and forecasted a combined pro forma annual recurring revenue of A$662.9 million once the compute division is incorporated. Trading of shares was suspended while the fundraising was organized, which is a typical procedure for a transaction of this magnitude on the ASX.

      The inherent risk lies in the fact that any company investing heavily in Nvidia GPUs based on contracts set to commence in 2027 faces uncertainty: the demand and pricing for AI infrastructure could change by the time the hardware is deployed and begins generating revenue.

      Megaport is investing capital now against future revenue that will arrive later, in a marketplace evolving swiftly, where 18 months can be a considerable span of time. The contracts provide a baseline, but the ambition for the inference cloud must grow, and that is what the A$827 million is ultimately betting on.