Highlights

  • Megaport (ASX:MP1) is evolving from a pure cloud-connectivity business into a full-stack AI infrastructure platform spanning network, compute and storage.
  • Its AI Exchange lets neocloud and GPU-as-a-Service providers, such as SHARON AI, plug into Megaport's global footprint and reach enterprise buyers on demand.
  • In June 2026 the company raised A$827.3 million to fund four new AI contracts (A$458.9 million total contract value) and a speculative on-demand GPU pool.
  • Pro forma group ARR reached about A$662.9 million, with compute ARR of A$385.2 million and network ARR of A$277.7 million (up 25% year on year).
  • The strategy shifts risk onto Megaport's balance sheet for the first time, making execution, utilisation and capital discipline the key things to watch.

Megaport Limited (ASX:MP1) has spent more than a decade as one of Australia's most successful global technology exports: a Network-as-a-Service (NaaS) platform that lets enterprises spin up private, on-demand connections to the world's major cloud providers in minutes. Now it is attempting one of the boldest reinventions on the ASX technology bench, repositioning from a connectivity specialist into a full-stack provider of AI infrastructure.

The pivot has been rapid and capital-intensive. Over roughly eight months the company has acquired a bare-metal compute business, launched an "AI Exchange" that plugs GPU providers directly into its global fabric, added a storage layer, and tapped shareholders for A$827.3 million to bankroll a distributed AI inference cloud and a pool of GPUs. For a business long prized for its asset-light, high-margin model, this is a meaningful change of character — and one that shifts both the opportunity and the risk profile for investors.

What Has Happened

Megaport is layering an AI infrastructure strategy on top of its connectivity business. The centrepiece products are the Megaport AI Exchange, a marketplace that lists AI compute, network and storage services on a single procurement page, and a "globally distributed AI inference cloud" placing compute close to where data and users sit.

The building blocks came together quickly. In November 2025 the company acquired Latitude.sh, a bare-metal compute provider, giving it the GPU and server capability it had lacked. In mid-2026 it launched Megaport Storage — powered by a partnership with Wasabi — completing an infrastructure "trifecta" of compute, network and storage. Then, in June 2026, it announced a fully underwritten A$827.3 million entitlement offer to fund newly contracted AI work and a GPU pool.

The AI Exchange has also begun signing up partners. SHARON AI, one of Australia's leading neoclouds and a GPU-as-a-Service operator, became the first Australian company integrated into the exchange in October 2025, and smaller AI software players have since begun listing services on it.

Why It Matters

The significance is twofold. First, Megaport is trying to ride one of the largest waves of infrastructure spending in years. As enterprises move from training AI models to running them — the "inference" phase — demand is shifting toward compute that is distributed, low-latency and located near end users. Megaport already operates the connective tissue between more than 1,100 data centres, which management argues is the ideal foundation for such a cloud.

Second, the strategy changes what kind of company Megaport is. For most of its life it has sold software-defined connectivity on subscription terms, earning high margins without owning much physical kit. Funding a pool of GPUs on a consumption basis means owning utilisation risk directly for the first time: if the GPUs are busy, the returns can be attractive; if they sit idle, the capital is stranded — a genuine departure from the capital-light model that made MP1 a market favourite.

From Cloud Connectivity to a Neocloud Platform

To understand the shift, it helps to understand the base business. Megaport's NaaS platform connects more than 3,000 customer logos across data centres in roughly two dozen countries, letting them provision private links to AWS, Microsoft Azure and Google Cloud through software rather than lengthy telco contracts. That business is still growing solidly, with network annual recurring revenue (ARR) of A$277.7 million as at April 2026, up about 25% year on year, and net revenue retention of 113% by logo.

Chief executive Michael Reid has described the emerging model as a "quasi-bare-metal neocloud" — a lean operation run by around 80 staff across some two dozen global locations, leaning heavily on automation. The AI Exchange is the commercial front door to that vision. Rather than trying to become a hyperscale cloud itself, Megaport lets specialist GPU and neocloud operators list their capacity and reach its enterprise base, with customers treating AI compute as another checkbox alongside the connectivity they already buy.

The AI Exchange in Practice

For a neocloud such as SHARON AI, the appeal is distribution. Plugging into the exchange means an enterprise, government or research customer anywhere on Megaport's footprint can establish a private, high-speed connection to that provider's GPU fleet in minutes. For Megaport, each connection drives network usage and places the platform at the centre of AI supply chains it does not have to build entirely itself.

Key Financial and Operational Details

The June 2026 raise put hard numbers on the ambition. The A$827.3 million entitlement offer was priced at A$14.30 per share, a 10.9% discount to the theoretical ex-rights price — dilution of about a third. Management earmarked around A$369.5 million for capital expenditure tied to AI contracts and roughly A$350 million (about US$252 million) for the GPU pool.

Those contracts are the anchor: four new AI infrastructure agreements carrying total contract value of about A$458.9 million. On the recurring-revenue side, pro forma group ARR reached approximately A$662.9 million, split between compute ARR of A$385.2 million — a 6.4-fold increase since the Latitude.sh acquisition — and the A$277.7 million network ARR. For FY26 the company guided to consolidated revenue of roughly A$307–315 million and an EBITDA margin of 21–24%, below its historically higher connectivity margins as investment ramps.

For context, in FY25 (reported August 2025) Megaport delivered revenue of about A$227.1 million, up 16%, and EBITDA of roughly A$62.3 million. The first half of FY26 showed the transition's cost, with a net loss of around A$19.1 million reflecting integration and build-out, and management has signalled that debt will become a more permanent part of the capital structure.

Potential Catalysts

Several developments could support the story. The most direct is utilisation of the new GPU pool: management has flagged a payback of roughly 16–22 months at optimal utilisation, so evidence that the pool is filling would validate the balance-sheet bet. Conversion of the four contracted AI customers into recognised revenue, and the signing of further contracts, would show the pipeline is deepening.

Growth in AI Exchange partners is another lever: each new neocloud or GPU provider that lists, and each enterprise that buys AI compute through the platform, reinforces the marketplace flywheel. Continued strength in the core network business, which still funds the group, would reassure the market that the AI push is additive rather than a distraction.

Material Risks and Uncertainties

The risks are equally clear-eyed. By funding a GPU pool on consumption-based pricing, Megaport is owning utilisation risk directly for the first time; if demand disappoints, expensive hardware could sit idle. The A$827.3 million raise was heavily dilutive, and management has indicated more debt lies ahead.

Integrating Latitude.sh, standing up storage and scaling AI infrastructure simultaneously is operationally demanding for a lean team. The company earns much of its revenue offshore, creating currency exposure, and it competes for AI infrastructure spend against far larger, better-capitalised players. The share price has also re-rated sharply on the AI narrative, which can amplify volatility if execution stumbles or guidance is missed. None of the projected returns on the GPU pool or new contracts are guaranteed.

What Investors Should Watch Next

For Megaport Limited (ASX:MP1), the near-term signposts are specific. First, watch GPU pool utilisation and any disclosure of realised revenue-per-GPU-hour, the metric that determines whether the capital bet pays off. Second, track conversion of the four contracted AI customers into recurring revenue. Third, monitor AI Exchange momentum — the number of neocloud partners and enterprise buyers transacting through it.

Beyond that, the FY26 revenue and EBITDA-margin outturn against guidance, the group's cash and debt position, and how quickly the distributed inference cloud moves from strategy to live deployments all matter.

Conclusion

Megaport Limited (ASX:MP1) is making a deliberate, high-conviction move to convert its role as the connective tissue between data centres into a broader AI infrastructure franchise. The AI Exchange, the Latitude.sh compute engine, the new storage layer and the A$827.3 million war chest together sketch an ambitious full-stack platform aimed squarely at the inference era.

The prize is participation in one of the defining spending cycles of the decade. The catch is that the company is trading some of the capital-light elegance that made it a market darling for direct exposure to utilisation and execution risk. Whether that trade pays off will depend less on the AI narrative and more on the unglamorous details — GPU utilisation, contract conversion and disciplined capital allocation.