Key Takeaways

  • Megaport (ASX:MP1) was caught in a global technology sell-off on 20 July 2026 amid fears the AI-infrastructure cycle has over-heated, with the stock at about A$18.23 on 17 July 2026, down 8.48 per cent.
  • It has pivoted from asset-light Network-as-a-Service toward a capital-intensive AI inference cloud after acquiring compute provider Latitude.sh in November 2025.
  • In May and June 2026 it announced AI contracts with combined total contract value above A$700m, and raised A$827.3m at A$14.30 a share.
  • H1 FY26 revenue rose 26 per cent to A$134.9m and guidance was upgraded, though the group reported a statutory loss of A$19.1m including acquisition costs.
  • The stock trades on demanding multiples, with a forward P/E ratio cited near 235 times, leaving it exposed to any de-rating of the AI trade ahead of FY26 results on 20 August 2026.

Megaport Limited (ASX:MP1) began the week under pressure as a broad retreat from technology shares raised fresh doubts about the durability of the artificial-intelligence investment cycle. Once a lean cloud-connectivity specialist, Megaport has spent the past year rebuilding itself around AI compute, striking large contracts and raising more than A$800m to fund the shift. That transformation has delivered a spectacular share-price run, but has also tied the company's fortunes tightly to AI-build-out sentiment.

From connectivity specialist to compute provider

Megaport built its reputation as the global leader in Network-as-a-Service, or NaaS, running a software-controlled global fabric that lets businesses create private, on-demand links to major cloud platforms including Amazon Web Services, Microsoft Azure, Google Cloud and Oracle. Rather than ordering fixed circuits that take weeks to install, customers can switch elastic connections up and down in minutes.

The footprint has expanded steadily: at the close of FY25 the network reached 983 enabled data centres, and mid-2026 capital-raising materials put the figure above 1,100 across 31 countries, though those counts come from company disclosures and may differ from audited numbers. The appeal has long been high gross margins, recurring subscription-style revenue and a largely fixed cost base.

The decisive change is that Megaport is no longer a pure connectivity play. In November 2025 it acquired Latitude.sh, a bare-metal and GPU compute provider, and through the first half of 2026 repositioned around what management calls a 'Globally-Distributed AI Inference Cloud', pairing its interconnection fabric with distributed GPU and CPU compute plus storage for latency-sensitive AI inference workloads. It is a far more capital-hungry and competitive proposition than the asset-light business investors came to know.

The contracts and capital raise that re-rated the stock

The catalyst was a rapid sequence of announcements. On 14 May 2026 Megaport disclosed three take-or-pay contracts with US technology providers running AI applications, valued at A$254.0m (US$182.9m) in total contract value and roughly A$90.6m in incremental annual recurring revenue, requiring about A$140.3m of capital expenditure.

On 3 June 2026 came four further AI-infrastructure contracts, valued at A$458.9m in total contract value and about A$199.0m in incremental annual recurring revenue, lifting the backlog materially and resetting the market's growth expectations.

To fund the pivot, Megaport launched an A$827.3m accelerated non-renounceable entitlement offer on a one-for-3.08 basis at A$14.30 a share, a 10.9 per cent discount to the theoretical ex-rights price, with institutional take-up around 99 per cent and an oversubscribed retail portion of roughly A$309m. Proceeds would seed an on-demand GPU pool of about A$350m alongside contracted capital expenditure of roughly A$369.5m, largely NVIDIA processors, with a payback cited at around 27 months. Pro-forma group annual recurring revenue was cited at about A$338m, up 49 per cent, a figure that appears to include Latitude and the new contracts rather than statutory reported ARR.

What the numbers show

The most recent full year, to 30 June 2025, showed revenue of A$227.1m, up 16 per cent, annual recurring revenue of A$243.8m, up 20 per cent, and EBITDA of A$62.3m, up 9 per cent, with net cash of A$87.8m and net revenue retention of 107 per cent.

The half-year to 31 December 2025 showed the pace of change: revenue rose 26 per cent to A$134.9m, EBITDA rose 28 per cent to A$35.3m at a margin near 26 per cent, and gross margin was 72 per cent. The statutory result was a net loss of A$19.1m, including A$15.8m of Latitude acquisition costs; the underlying loss was around A$3.3m.

The 65 on results day despite the top-line beat, on acquisition-cost and currency concerns. Management nonetheless upgraded FY26 guidance to revenue of A$302m to A$317m, an EBITDA margin of 21 to 24 per cent and capital expenditure of A$90m to A$100m.

Why the strategy raises the stakes

The legacy NaaS business is a classic operating-leverage story: because most costs are fixed, incremental recurring revenue at 72 per cent gross margins flows through strongly, letting EBITDA grow faster than revenue, as H1 FY26's 26 per cent margin showed.

The AI pivot changes the arithmetic: GPU hardware is expensive, depreciates quickly and carries risks around utilisation and obsolescence, tilting the business toward capital intensity. Brokers see the take-or-pay structure as a buffer, but returns still hinge on execution and the return on invested capital.

The macro backdrop and the AI debate

The wider mood on 20 July 2026 was cautious, with the S&P/ASX 200 reported around 8,796 to 8,825 points and the Australian dollar near US$0.6985. Brent crude reportedly jumped about 3.1 per cent to roughly US$90.82 a barrel amid escalating US-Iran hostilities and fears of Strait of Hormuz disruption. These backdrop figures were not independently re-verified.

Higher oil prices revive inflation worries, an unhelpful combination for high-multiple growth companies whose valuations rest on distant cash flows. Layered on top is unease about the AI theme itself: as Morgan Stanley has framed it, the debate has shifted from validating demand toward execution, utilisation and returns on invested capital.

The opportunity, if execution holds

Structural adoption of multicloud and hybrid-cloud architectures keeps driving demand for flexible interconnection, and the AI data-centre construction boom adds a powerful tailwind. By combining connectivity with distributed compute, Megaport argues it can occupy a differentiated position in AI inference, while the contracted backlog improves revenue visibility.

Analysts have offered constructive views: UBS has pointed to a potential return of about 45 per cent over the initial contract terms, Macquarie has suggested internal rates of return could exceed 20 per cent while noting Latitude's dual NVIDIA/AMD chip use mitigates supply risk, and Ord Minnett has estimated FY27 compute revenue could exceed US$130m. These are analyst estimates, not company guidance.

The risks investors are weighing

The bear case starts with valuation. On conventional metrics Megaport is expensive: statutorily unprofitable, with a trailing net loss of around A$20m and a forward P/E ratio cited near 235 times. Estimates diverge sharply: a June 2026 discounted cash-flow analysis put fair value at A$10.25, well below the market price, a community estimate sat around A$16.92, and JPMorgan's Overweight rating carried a A$16.70 target the shares had already passed by July.

Beyond valuation, the risks are tangible: AI demand could prove cyclical; GPU utilisation might disappoint and hardware could be superseded; and a capital-heavy pivot is harder to execute than an asset-light network. The A$827.3m raise dilutes existing holders, and the new AI revenue is concentrated among a handful of large US customers.

Competition is intensifying on two fronts: in connectivity, Megaport contends with Equinix Fabric, Console Connect and PacketFabric, plus hyperscalers' native interconnect; in GPU cloud it now faces specialists such as CoreWeave and Nebius alongside the hyperscalers. Trading near record prices, the stock is exposed to any de-rating of the AI trade, exactly the narrative that gripped the market on 20 July 2026.

What to watch from here

The next checkpoint is Megaport's FY26 full-year results, due on 20 August 2026. Investors will look for confirmation the company met its upgraded revenue guidance of A$302m to A$317m and delivered an EBITDA margin within the 21 to 24 per cent range, along with detail on how quickly the new AI contracts are converting into billed revenue.

Equally important will be early evidence on GPU fleet utilisation and the returns earned on rapidly growing invested capital, since these will determine whether the pivot creates or destroys value.

A balanced view on Megaport (ASX:MP1)

Megaport Limited (ASX:MP1) has engineered one of the more dramatic strategic reinventions on the local market, moving from a prized asset-light connectivity model toward a capital-intensive AI compute business in roughly a year. The rewards have shown in a soaring share price and a much larger contracted revenue base, but the cost is that the company is now hostage to the very sentiment cycle that drove its shares higher.

Whether Megaport vindicates its pivot will depend less on the AI narrative and more on execution: filling its GPU capacity, earning acceptable returns and managing obsolescence. The FY26 results due on 20 August 2026 will provide the first real test.