Capex now comes with a B

FY26 result (year ended 30 June 2026). Reporting currency: A$

It’s hard to get your head around how much this business is set to step change, particularly in terms of capital deployment.

Capex went from A$35m in FY25 to A$153m in FY26 and is guided for A$1.3-1.4bn in FY27. Compute of course being the driving force.

In the Michael Reid era I have received remarkably consistent positive industry feedback on the networking business and Megaport overall. The networking business doesn’t move on a dime (sticky customers cuts both ways) but my industry feedback suggests good inertia and the numbers seem to suggest that (eg NRR expansion +5pts). Although I would have thought organic constant currency Network ARR +24% in the year would suggest more momentum than the FY27 revenue growth guide of +17-21%. Maybe conservative, something for me to run down.

More work is needed on my part on the compute side but so far conceptually I’ve had supportive industry feedback. Extending into GPU from a bare metal CPU base is a natural extension as previous CPU customers increasingly have AI use cases. Customer demand for instant provisioning and user-friendly front-ends is relevant (remembering a lot of developers came to age in an AWS environment). And having an integrated private network can be a relevant competitive advantage for those customers where the networking side is important to their use case.

On another note, more M&A is interesting. M&A likely remains a recurring theme. Sounds like something in storage is to come. And I would envision software acquisitions to layer on the GPU pool business.

But for the one currently in due diligence (in the networking and security space), the target’s revenue and expected valuation range are both ~US$100m, not an onerous multiple.

At a guess, maybe Aryaka Networks? Smarter minds than mine pushed back to me on that one, but it would do a few things. It would physically extend MP1 to the last mile of the customer (vs ending at the data centre). It would add a couple hundred enterprise customers. And it would add for the first time owned security features. Last ARR datapoint on them I’ve seen is US$100m in June 2024. They’ve had challenges hence maybe acquirable now.

Such a big capex build-out naturally carries risk. Although within the emerging mini-neocloud sector on the ASX, as Firmus and Sharon AI both head to the local sharemarket, I can’t help but point out MP1 is a comparatively more established operator, with a track record of delivering on uptime, and an excellent industry reputation based on my conversations. From a de-risking perspective for investors, that surely should help them stand out in the space.

FY27 guidance is the first issued including the recently announced large compute ramp of the on-demand GPU pool.

Compute revenue is a reasonably wide range, with revenue driven by timing of contract delivery.

On revenue recognition timing within FY27: “the strategic deal is take-or-pay…as soon as it’s delivered, it instantaneously starts revenuing.

Of the GPU pool specifically: “we’re not expecting a whole range of that pool to be revenue in this financial year.”

Delivery risk: “each one of these components needs to be delivered in order to turn on the paying component. And if any of that delays, which, as you know, is a massively supply chain challenged world at the moment across all these elements, it can move it.”

FY30 and beyond aspiration: Sustainable 20%+ growth — Highly Profitable — Sustained profitability and sustained FCF.

FY27–FY29 phase framing: Accelerate revenue — Capitalise on prior transformation investment; Expand TAM; Aggressively grow market share; Continue investment with revenue growing faster than costs; Accelerate revenue through reinvestment.

Due diligence is underway on the potential acquisition of a US-based technology company with an estimated annual revenue of ~US$90-100m and breakeven EBITDA.

The potential funding requirement (including capex) is estimated at US$75-110m.

“Discussions remain at an early stage…”

Re M&A targets:

Syndicate facility - Binding commitments secured for an A$825m syndicate debt facility, inclusive of refinancing the existing A$150m facility.

Contracts funded from available liquidity — cash at bank A$435m plus the A$825m debt facility. "Fully funded for all committed capital expenditure…in addition to the potential acquisition".Compute investment update

Supply chain shortages > contract variation: supply constraints meant equipment for two of the 3 June 2026 contract announcements could not be secured. Alternative arrangements were agreed and two new contracts were entered for higher-grade GPUs, described as "an improved commercial outcome for Megaport". Aggregate TCV increases by ~US$87m versus previously announced, no material change in ARR and capital expenditure requirements.

Total strategic contract TCV since April 2026 to approximately US$0.9B (A$1.3B). Full ARR now expected to be A$435m, supported by A$826m in capex. Equipment, space and power has been procured for all contracts.

GPU pool deployment metrics unchanged: Time to procure and deploy the GPU Pool 6-9 Months; Ramp from deployment 3-6 Months; Payback target 16-22 Months.

A sample size (<5%) of GPUs will be deployed for testing in Q2 FY27, with the remainder planned to be deployed in line with the above.

FY26 delivered an in-line to above result vs latest guidance entering August.

FY26 is the first year to consolidate Latitude.sh, acquired 26 November 2025, and Extreme IX, acquired 12 December 2025.

Revenue. Accelerating core Network revenue, driven by strong NRR growth and new logo acquisition. Increase in revenue also driven by new Compute revenue stream following the acquisition and expansion of Latitude.sh. Compute Services revenue is reported entirely in the Americas.

Gross margin. Partner commissions 12% of standalone Network revenue, consistent with FY25. Direct network costs, +34% to A$51.5m against revenue up 37% reflect ongoing investment, 155 net new DCs and continued 100G/400G core backbone upgrades.

Costs. Employee costs reflect rolling addition in headcount, concentrated across GTM roles, together with the inclusion of headcount from integrated acquisitions; "H2 FY26 employee costs increased 34% on H1 FY26. Equity-settled employee costs were A$34.0m, an increase of A$14.5m.

Underlying EBITDA. Excludes equity-settled employee costs A$34.0m, acquisition-related costs A$16.2m (Latitude.sh A$15.0m, Extreme IX A$1.2m), fair value loss on contingent consideration A$3.0m, finance costs A$6.3m and FX losses A$3.3m.

Investing cash outflows. Includes $115m cash outflow towards acquisitions of Latitude.sh and Extreme IX, $158m capex investment and $172m of advance payments for capex. The advance payments sit in non-current deposits.

155 net new data centres were added, up 35% YoY, and new DCs in FY26 landed with higher first year ARR per DC.

Incremental ARR was +103% on a constant currency basis, the largest ever increase y/y.

The FY26 customer cohort generated ARR 55% higher than the previous record.

Network LTV:CAC of 5.8x remains above investment zone target range even with continued Go-To-Market investment.

Q: Does the FY27 capex range cover all kit for all contracts and the GPU pool?
A: “Yes, you are correct on the CapEx...the only time it triggers CapEx is actually when the equipment is delivered...you might pay 20% up front deposit. And then once you deliver the kit, then your CapEx actually triggers at that point.”

Q: Why could equipment not be procured for two contracts?
A: “...it's largely around technical specs.” “One particular contract was associated for a particular hardware that we weren't comfortable with the spread of the different hardware for the automation system. So we changed that with multiple customers, in effect, to a larger TCV and a much better contract”.

Q: Entering FY28 run-rating the full compute ARR plus the pool?
A: “...you would have a ramp period as all the GPUs start to get taken up, which is different to a strategic deal...The strategic deal is take-or-pay the entire piece. As soon as it's delivered, it instantaneously starts revenuing...ARR is different to revenue, i.e., ARR, let's say, ARR hits at December mark, we'll only get 6 months of that booked in revenue.”

Q: Does the bottom end of guidance assume minimal GPU pool?
A: “...It's really just around timing when it's deployed and when it's taken up and the...price that customers are paying at that point.” “...1-month delay makes a material difference...it takes 6 to 9 months to procure and then deploy. And once we've deployed it, there is a 3- to 6-month ramp...we're not expecting a whole range of that pool to be revenue in this financial year.”

Q: Run-rate EBITDA above $600m on twelve months of contracts and pool?
A: “...the timing on the range for revenue is driven by timing of strategic contract delivery, and then the EBITDA is a fall-through of that...we have to keep reinvesting in the new group business. So you don't just drop every single dollar to the bottom line.” “The network revenue is tight and there's a broader range for compute revenue because of the sheer difference between how it ramps.”

Q: Reasonable expectation for reinvestment of the $435m of contracted ARR?
A: “...you would start to see a higher margin, but I wouldn't necessarily be baking that in at this stage...with a higher CapEx in business, you would expect to see higher margins over time, but give us time to grow.” “...we will continue investment with revenue growing faster than costs...what is the long-term future for FY '30 and beyond, very sustainable 20% growth...We're going to do 100% plus in revenue next year.”

Q: What other options to fund future compute contracts?
A: “...we have the following retail entitlement offer that landed on the 2nd of July.” “...we're covered for what we've announced...this is funded exact thing by debt for this contract. That's not how it is. We've got a general corporate facility for those purposes.”

Q: Prepared to increase gearing to something like 2 times?
A: “We're well funded, so we'll continue to manage within thresholds. We've only just got the commitment letters at this stage.” “...we're going to grow into the debt rather than just from day one.” “...this is just step 1 in...our capital management framework plan.”

Q: Who are you competing against in compute tenders?
A: “I haven't come across any of the CDN players that I'm aware of...every 2 weeks, the pricing changes. Vendors have gone away from long pricing. They'll give you about a 2-week window...So there's no tenders, so to speak...it's a timing game...the pie is expanding so rapidly and so large that there's plenty of pie for anyone and everyone in this space.”

Q: Do the storage partnerships cover the need, or is storage still an acquisition area?
A: “...from a storage element...I would say we are comfortable with what we have today for everything, but uncomfortable in that we need to constantly add more and more innovation to it...we're constantly exploring options in that space, and we will forever be exploring options in that space.”

Q: Which space is the potential acquisition in?
A: “We're currently looking - I think it would be the network and security space...this isn't just a compute business here. This is a giant network business that we'll continue to invest in, a giant CPU business and a GPU business, and then there will be the storage elements that come with it as well.”

Q: Plans for liquid cooling, given Vera Rubin requires it?
A: “It will be a mix...99.1% of data centers are not built for liquid cooling...if you want to deliver Vera Rubin, you'll need liquid cooled elements just because of the sheer heat that they create...data centers are catching up...You can also deploy B300s liquid cooled...GB300s don't sit in an air-cooled facilities, but B300s sit in there. One has the Grace Blackwell on 72 in 1 rack.”

Q: Rollout of the new compute platform?
A: “...we rolled all of that out...we have 100-something plus sites or locations that deploy virtual edge. If anything, that equipment is probably getting old and in the future, we'll look to replace that...that gives us that 400-gig connectivity up and then 100-gig opportunity to run.”

Q: What gives confidence the compute customers are quality counterparties?
A: “...the important part about adding the pool is to constantly diversify our customer base”. “We do a mix of credit checks, obviously, speak to the customer, understand the financials...it will be a mix of terms, but a mix of upfront payment and then payment in advance.”

Q: Take-up of pooled capacity and rollout of the initial compute contracts?
A: “So customers are very happy because they keep asking us for more. We are being disciplined around how we accept more, because...you could just sell everything to one customer and you end up with this concentration risk...We're hitting the time frames that we're telling our customers...Latitude and Megaport combined, we run 26 data centers plus...as well as...1,100 physical data centers all around the globe with automated platforms.”