Ai-Media Technologies Limited (AIM) Earnings Call Transcript & Summary

[Audio Gap]presenting to the results. There is opportunity to ask questions at the end of the presentation. Please enter your questions via the function at the bottom of the screen. I'll now pass to Tony.

Thanks very much, Mel, and thanks, everyone, for joining Jason and me today to present our first year FY '26 results. I'd like to pay my respect to the traditional custodians of the land here in Chatswood, the Cammeraygal people of the Eora Nation. So this really is a very exciting time for Ai-Media. Five years ago, we acquired EEG, a hardware business in New York. And we've spent the last 5 years, essentially our first 5 years of listing, transforming the business from what was a human interpreted AI service to a fully AI native technology platform. We -- what's really important to understand, though, is that there are really 3 different things going on in this set of results. And if you just look at this at a point in time, it's a pretty confusing story. But if you can unpack what has been the 5-year transitions, you'll see very clearly that we've now completed that 5-year transition, and we are now setting the business up for the next decade at scale with fully AI-native solutions. And I'll go through a little bit about the metrics of that transformation a little later. But the key things that I really want to highlight here is that we see lots of opportunities with the LEXI product suite right across the 9 Squares that we've been speaking about in terms of expansion opportunities. So they are in the 3 industry categories being broadcast, being government and being enterprise and then the 3 regions being the Americas, EMEA and APAC. What we've now got, now that we have completely wound down that legacy business, and we shut the gate on that on the 31st of December 2025, and you'll see that big reduction in services revenue is now actually at 0. And so we finished that December run rate with 0 services revenue, and we're now just about building annual recurring revenue through a subscription model, and then selling in professional services on top of that in terms of implementation fees. So what we've, I think, want to really highlight is where is the kind of noise in the numbers from things that are either not going to be repeating, where have we made investments in product that we have expensed, not because we -- for any other reason than we want to just keep our balance sheet very, very clean. We have a number of U.S. investors, and those U.S. investors expect all R&D and product development to be expensed on the accounts. So what we have done with the advice of our auditors is to really add back or make it really clear how much of that investment was in products that were pre-revenue, but we expect to get revenue from in the near future, and that was $2.6 million for the half. So that's really the big swing in terms of the negative reported EBITDA number. And then finally, I think the other key thing that's going on in the numbers is when we acquired that business in New York, it was a hardware business, and they made about USD 8 million a year, selling pieces of hardware for anywhere between $10,000 and $12,000 each. There are 8,000 of those encoders in the field now, up from 5,000 when we acquired the business. And 5,000 of those 8,000 encoders are now up for a replacement cycle. Ever since we launched the FY '29 aspirational CEO target of revenue of $150 million and EBITDA of $60 million, this was really always the plan to see where we could, as quickly as possible, sunset the legacy business, convert what was a lumpy CapEx hardware business into annual recurring revenue. And just at our annual summit last month in Palm Springs, we had the opportunity to showcase those new encoders, showcase the new hardware as a software model to some of our largest customers, and they have confirmed that this is absolutely something that they prefer, bundling in, obviously, support on top of that. And they're also just very keen to get their hands on the new hardware that over the last 2 years, we've really been specking out based on all of the work we've been doing on LEXI Text and LEXI Voice, implementations in particular. But what we've also done is we've also developed some new LEXI family applications that will also be part of that annual recurring revenue growth going forward. So going forward, we will be reporting on annual recurring revenue. We'll be reporting on annual recurring revenue growth, we will be reporting on annual recurring revenue margin and EBITDA expansion. And this is really a far more traditional way for us to be reporting. And hopefully, this simplifies the business going forward now that we can pull out these 2 elements that are really clouding the underlying result, which is a very strong growth in annual recurring revenue with declining hardware sales, because it was end of life, and we're moving from a CapEx model to a brand-new generation of encoder and the sunsetting of that legacy services business that was very high in revenue, but very low in earnings. And so having that cost out of the system now allows us to have a model that can fully scale, and we're looking forward to continuing the success that we've had in that product development field. We just have got the go ahead to further increase some product and development investment in some of the key areas where we're getting some really exciting product-market fit. So in terms of the annual recurring revenue growth, we guided at the August results that we would be growing for the full year at 35%. The first half kicked strongly at 80%. A lot of that was actually the structural transition of the services revenue customers right across to LEXI. And so this includes our 2 largest customers, Channel Nine and Channel Seven, that are now fully LEXI and have been since the 1st of January. And you'll see in some of the later slides, the underlying scaling success that we've had with LEXI Text, that has really underlined a lot of this number. And so layering in those new opportunities and doing it across those 9 Squares are really how we see this sort of -- how we see this building on the success of LEXI Text. And there's some -- the slide right at the end of the deck that gives you the 20-year history of both iCap Scaling and LEXI Scaling shows that these products really that are an infrastructure, an ecosystem layer that sits across a live broadcast hardware system, it takes about 4 years to scale. And so what we're trying to do is compress that time line for some of these new LEXI upsell and cross-sells. And the key way of doing that is with the hardware refresh and these brand-new GPU-enabled devices that we'll launch next month at Las Vegas. So in terms of financial performance, you can see there the annual recurring revenue, $30 million at the exit run rate in December. Total revenue decline, as I said, contributed to by lower hardware sales and the decline in services revenue. And I've discussed the adjusted EBITDA, which also includes that $2.6 million in pre-revenue product and development. And so you can see that almost half of our total product and R&D spend is on pre-revenue products, which is, I think, the right level of investment given the opportunities that we're seeing. This is a great slide really, which shows what the company was before the transformation started, and where we are today. So effectively, [ 1H '21 ] was the business we took to IPO. It was the ReSpeaker led model. We had humans in the loop powering the AI. Now we're fully native AI. Our total revenue is up only 30% over 5 years. But you can see that gross profit is up 136%. But most importantly, look at that turnaround in the free cash flows. And in terms of the impact that we're having on our customers and the defensibility of our moats, just look at the growth in the LEXI minutes, that were over 1,200% increase over that period. We've got more customers delivering more services who want to do more and more with us as they replace their third-party vendors in other applications with our new LEXI suite. This is a slide you've seen before. So this really talks to how we add value to the broadcast chain. It's about making sure that we have a hardware solution, which is right next to the capture of the live video so that we can do things really quickly. And then the other thing that we do is we automate the orchestration of topic models and data to ensure that the AI is fed the right information at the right time second by second. So when we're going to the weather on for Brisbane, we need to know what the Brisbane suburbs are, or similar things for sport, the next 45 seconds. And then you need a different dictionary set. And so the AI automation layer in #3 acts a bit like a Juke box, and it moves different topic models in and out depending on the triggers that we receive from the rundown from our broadcast customers. And the rundown will include every single line item that is being played second by second from midnight to midnight. And this is all driven without any human intervention. And that's really been the critical architecture piece that's allowed our customers to get benefit from this, and it's also the moats that we have because there is no other company out there in the world that has an encoder set like this and in 8,000 points of presence right now. We're upgrading those encoders. We're putting really much more powerful GPUs in them, and we're layering on a series of other AI applications inside those encoders like AudioShake, which help us to denoise where there is very mixed and choppy audio, and to do it much faster than anybody else could, and with much better access to clean data, because of the 2 things being, we are closest to the signal, and we are orchestrating with the back-end systems that our customers have. And this slide is new, and this was actually a shareholder request, to really explain where does AI media fit in the overall architecture of camera to viewing. And so you really see at the top left, we have content creators. They will go and make content. Where we fit in is after the signal transport layer and feeds in directly to the broadcast workflow before going out to distribution. And so what we do is we do that secure ingest, real-time AI orchestration and language intelligence at scale that relies on having the cleanest possible data set. The most important thing is that we -- and what we work on with the professional services implementation layer with our customers is actually ensuring that we have the right systems to ensure that the data doesn't get corrupted. So are we spelling Fremantle the right way or are we spelling it the wrong way, for example, and making sure that any errors are picked up. And so the system gets better and better and learns. And so that delivers real-time captions, live translation, voice synthesis, accessibility, compliance, plus, plus, plus. And what we're doing is we're really future-proofing those -- that LEXI product suite now with the new encoder set. And so this is now the ecosystem. So right in the middle there, we have those LEXI encoders. We deliver support as a service around that with 2 tiers, business or enterprise. We've got the LEXI Text, which is the bulk of everything that has scaled so far. LEXI Voice, we've got some very, very good momentum on that. And obviously, the momentum on that has led to our confidence to build out the LEXI Voice Encoder line. LEXI AI, we're doing our first implementations with -- and POCs with existing customers to get more out of their data set, which feeds into the LEXI Library product, which gives customers access to all of their video archive data. LEXI Translate is the text-based translation engine. And then 2 new products that we're releasing also next month at -- in Las Vegas are, LEXI Audio Description, which is LEXI AD, which fits in really well with the Title II ADA regulation changes and government entities are really key on that one. And then the sleeper, but that's growing really nicely now is our LEXI Recorded product, which is offering effectively everything that we do in a live environment, also in a recorded environment. Big question everyone says is what about AirPods and things like that. They do the translation. How is that different to LEXI Voice? It's the same underlying translation technology. Let me be clear about that. But it's a very different workflow, and it's a very different set of circumstances. So that same API that's listening to something through an AirPod is listening perhaps to mixed audio. It doesn't have that orchestration layer. And you're not guaranteeing that everybody who's watching that broadcast is receiving the same translation. Everyone will receive a different translation where it's translated at the destination. So in environments where you need deterministic or consistent output, like at a conference or a broadcast, where you've got regulatory, contractual or reputational liability, where you've got audit consequences, that's really what we're talking about. And remember, our customers are the biggest broadcasters on the planet and the largest technology companies in the world and making sure that we've got this reliability plus all of the security layer that sits behind it, which comes with our recent SOC 2 and ISO certification. This is a key distinction and why we see there continuing to be a business-to-business rather than a direct-to-consumer application for what we can deliver. This is a 10-year graph here that shows the scaling of LEXI. The top line there is iCap, which is the network traffic, which was the very first product that EEG developed back in 2008, and it really paved the way for the implementation of LEXI. But you can see even when Ai-Media acquired EEG, we did so because we saw the promise of LEXI, but we did so back in December of 2020. So you can see just how much that's grown, and that's that 1,200% increase that you can see. Also the difference between the blue line and the pink is what our competitors deliver. So when we acquired the business, our competitors were delivering 84% of the total. And now our competitors are only delivering 1/3 of a much higher total. And so what we want to do is we want to replicate the success that we've had with LEXI Text with LEXI Voice and some of those other applications. And what we're now saying is that this will all roll up into annual recurring revenue, because we're moving everything to that monthly subscription model. This is, again, the unpacking under the old system. So this includes -- this has got apples and oranges in it. This has got services revenue. This has got technology revenue that includes CapEx and it's got annual recurring revenue. So that's why a slide like this can be a little confusing. We added an extra 7 countries in the last 6 months. Probably the most interesting for me was Mongolia. We didn't really expect there to be a lot of latent -- Well, I didn't expect there to be a lot of latent demand in Mongolia, but it turns out there is. And so we're actually doing our first partner conference. We've signed up 12 partners in the Asian region, and we're doing a first partner summit in Vietnam on the 3rd of June. To the extent that shareholders are in country around where we are doing a lot of these summits and events and trade shows, it's a really good opportunity actually to come along and get under the hood and talk to our customers, and we would encourage you to do that. Before I hand over to Jason, this is just the real -- this is the history of the EEG business that we acquired in 2021. It started with captioning encoders in 1981. And then, as I mentioned, the real key was around the connectivity, which was developed with the iCap network, which Bill McLaughlin, who is the son of the vendor of EEG, who's now our Chief Product Officer and I would say a tech genius wizard and leader for us all, and he really architected what has become the core underlying live broadcast infrastructure software layer on which everything else has been built. And then as you saw from that graph, it was really -- the beginning of scale was in 2018, and the acquisition that we made of this business has really supercharged that. So we're now delivering 12 times the volume that they were before and that we were combined. Jason, I'll hand over to you.

Thanks, Tony. The primary objective of the business has been to become a technology business, a technology-driven company. And if you look at our successes so far, it is heading in that direction. The tech revenue represented 71% of total revenue in H1 FY '26 on exit. That jumped to 80% in technology revenue and 20% in services revenue. We're shifting completely out of the services business and have closed down the legacy infrastructure. Going forward, you will see a services component in the revenue, but that will be professional services. We have had an increase in SaaS revenue of 61% on pcp. Typically, this has been around 35%. A lot of help has been through the transition period, and you'll see some of the numbers come out later on in my segment. Still very high gross margin at 84% for technology revenue. We are currently doing $2.5 million per month in recurring revenue, up from $2 million in the beginning of the half. We have a target in market for $3 million recurring revenue by the end of June per month, annualizing that to a $30 million ARR. That is the -- that's where we're heading to with our annual recurring revenue. There are other opportunities to support this recurring revenue over and above our target, which is transforming support as a service to ARR. Currently, support is offered on encoder repairs and replacement, not an enterprise-wide support agreement, which we're implementing to our current customer base of about 2,000 customers. Our encoder base, as Tony has said before, is just under 8,000 encoders out in the market, and a lot of these are over 5 years old, 3 years old and over is 5,152 encoders. Our plan is to release the new encoders at NAB and transform this model into a subscription model as well. Cash balance is strong at $16.7 million, up $2 million on pcp. Operating cash flow is up $2.6 million with free cash flow at $2.4 million. Next slide, please. There has been a lot of questions around what is our commercial model? And I think this snapshot really depicts our commercial model. So our objective is to become primarily all recurring ARR subscription-based revenue and 20% of that will be the professional services. So hardware currently sold upfront, typically is sold for around about $15,000 per unit AUD and the responsibility then transfers over to the customer to manage that -- any repairs, they have to pay for that and it really has a small warranty period. That hasn't been really received that well. There has been a lot of encouragement in transferring that responsibility over to Ai-Media and having a recurring model, which typically would be around about $500 per month. And every 3 years, they would receive an upgraded model or -- earlier or later, but that would be part of it, and that would just continue. In it would obviously be the support that would be tied in as a bundle over and above the $500, depending on what tier of support they would use. Then we have our LEXI products, which is LEXI AI, LEXI Voice, LEXI Traditional, LEXI Text, LEXI Insights, and that would be based on a selection basis. The professional services is largely implementation or engineering time that we would time -- that we would charge on time and materials at a 50% gross margin. Currently, professional services done free and support isn't charged. So those are opportunities we have in the market. I've been showing this graph for the last 3 halves, primarily to show that although tech revenue in the P&L is up 12%, overall tech sales is up 21% and deferred revenue is up 49%. As more and more customers transition to the long-term subscription revenue, we are going to see more deferred revenue and higher tech sales. In terms of our regions, we have 3 different regions, the Americas regions, Asia Pacific and EMEA. There is a lot of investment that has gone into EMEA and Asia Pacific and largely low revenue or sometimes even pre-revenue. In Asia Pacific, most of the revenue still comes from Australia, 96%. We are investing in the Asian region. We see a lot of opportunities through partners. As Tony has mentioned, we're going to be hosting events there. Again, there is development work required for different languages and similar to EMEA. As we expand the EMEA proportion of our footprint, we also need to invest further in different, different regions and languages and development of software and LEXI transcription and LEXI translation. Moving on to operating expenses. Positively, we have declined G&A by $1.1 million. This is part of a restructure that we did in FY '25. We've reinvested those funds into product, again, up 38%. A large part of our investment is in R&D. I think if we looked at it, it's around $2.6 million in pre-revenue investment whole of company for the half year. There is some investment that's gone into sales and marketing, but it's largely flat. In the P&L, the main thing that really stands out here, obviously, is the decline in hardware. If you're looking at the decline in hardware revenue, has declined $4 million. A lot of it is -- this is really to do with the new encoders that are due for release. We haven't released any encoders for the last 8 years. So we are releasing new encoders with higher specs, higher AI capabilities and our customers are waiting to transition or buy those encoders. So that's really what the reason is behind the drop in the hardware revenue on pcp. We expect that to increase substantially when we release the encoders in April 2026. SaaS revenue, pleasing result at 61% and services revenue declining at 33%. Services revenue now represents a gross profit of 34%. It is -- had become subscale and really not that profitable for us. In terms of our share-based payments, there was $1.3 million taken up in the P&L. It is based on the next 5 years. So the most expense that we would see in the P&L is in this half. It's based on a graded valuation model where you proportionately account for the next 5 years upfront. So we will see that number decline substantially over the period, leaving an adjusted EBITDA of $0.4 million, share-based payments being noncash. Operating cash flow is still strong at $2.6 million. And as I mentioned before, that cash in the bank has increased by $2 million. Our cash flow bridge will show the key things in terms of the net off between share-based payments and EBITDA. That's the delta of that negative $400,000. We have begun paying commissions on cash collections. That has helped a lot in terms of how our working capital, up $3 million on 6 months ago. And there are some small other expenses that worked against us, which was foreign exchange. We spent a little bit on CapEx, on -- mainly inventory and some CapEx on the production facility. We -- as a lot of you know, we have a facility in Farmingdale, Long Island, New York. We don't speak about it much, but we have invested significantly in that whole production capability facility. We now are able to do 3,000 encoders -- produce 3,000 encoders per annum. With a little bit more CapEx, we can increase that to 4,000. However, it is still subscale. It's current utilization is only 30%. As you know, typically, we would sell on a CapEx basis around about 1,000 encoders per year. Major improvements in the facility, less defects, less downtime. And a lot of our work happens on our encoders in this facility, R&D of the new encoders and the release of the new encoders, all borne from this facility. I'll hand over back to Tony to wrap up and talk about our prospective opportunities.

Thanks, Jason. And I want to leave a bit of a lot of time for Q&A. So the 3 strategic pillars remain unchanged. It's product expansion, it's geographic penetration and its segment differentiation. This is my favorite graph of the whole slide, and this is actually based on an Excel spreadsheet that Bill, my Chief Product Officer, has kept ever since 2008, which was where he recorded the number of iCap minutes each month, hours actually was -- his spreadsheet was in hours. The very first month of iCap hours, there were 4 iCap hours. And he has tracked that every single month since 2008 through to today. And what we really see is that the iCap Scaling, which remember was in Phase 2, that connectivity phase of the evolution of the business, really took 4 years to kind of prove. And then you see that beginning to take off. Similarly, LEXI Text was first implemented in around 2016, and it wasn't really until 2020 that you see this really lifting. So why -- what's the context of this graph? Well, what does LEXI Voice look like going forward, right? And so we launched that product a year ago. So we're 1 year into the product-market fit phase. And the exciting thing is that we found a couple of really valuable product market niches for that product where it is ready right now. Many other opportunities where the product still needs to wait for improvements in the underlying engines, but there's certainly an opportunity for us to invest more in go-to-market and in product improvements that are particular to those different use cases. So this is the -- then a little more on those -- the LEXI product suite. As I said, it's really just LEXI Text that is in those -- in that graph today. And so adding on voice AI encoders and support is going to increase that pink line and the annual recurring revenue even further. So where have we got in terms of the last 6 months on each of those 9 Squares. So remember, we pretty much started live, just in the top left square, in U.S. broadcast. The encoder refresh is going to be -- this is obviously still our largest market, the one we need to focus on the most in terms of near-term revenue opportunities and the encoder refresh is front and center of that. And so it's the -- we have the big show at Las Vegas, NAB, next month, we'll be getting feedback and preorders on that. LEXI Text now, as I said, it's now delivering 2/3 of the total, up from 16%. And so customers that had bought EEG hardware for a long time and then bought EEG hardware and iCap, but we're still using third-party captioning agencies are now looking to move everything across to us. And so we've got our eyes on that last remaining 1/3 of business as well. We are working on LEXI Voice POCs, proof of concepts, with our large American broadcasters, but there's still some work to do on getting that product ready for prime time in broadcast markets. So we are actually seeing the LEXI Voice opportunities in that bottom row of enterprise kicking probably the soonest, but also in government. And we've had some growth in the government sector. It's up from 5% to 7%. We do want to see that continuing to grow further. And the LEXI Voice and the new encoders will be a key part of that. In EMEA, it is that heavy investment to make sure that the right character sets for each country can turn up the right way as it is in Asia as well. So that's certainly where we will continue to increase our investments in product. In terms of where we sit in the competitive landscape, a lot of questions around what is -- what are the hyperscalers on the top left going to do to us? And the answer is, well, they're all our customers, and they use us for situations where they need things live and situations where they need that orchestration layer. So in summary, we've now transitioned the business to high-margin annual recurring revenue. We're now forecasting that growth up from 35% to 50% for the full FY '26 year. So that's going for an exit run rate of $36 million compared to $24 million in June of 2025. We are successfully penetrating these new markets with our technology. We are successfully deploying and leveraging AI advances like AudioShake and other software features onto our brand new encoder set, which reinforce our defensible moat. By moving out of just captioning into translation and voice, we've expanded the total addressable market from $2 billion to $70 billion, and we have 0 debt and a strong position to continue to invest. So with that, Mel, I'll hand over to you for questions.

Great. Thanks, Tony. Let's start with some of the questions we received just ahead of the webinar. Firstly, can we talk to why the FY '29 target was pulled? And also how you feel about revenue and EBITDA going forward given some of the comments in previous reporting periods?

Yes. So we are withdrawing those FY '29 CEO aspirational targets, because that's what they were. And instead, we're replacing them with Board approved near-term guidance on the 4 metrics of annual recurring revenue, annual recurring revenue growth, annual recurring revenue margin and EBITDA expansion. How I'm feeling about revenue growth versus EBITDA growth in the short-term is focus on scale, make sure that we are growing the revenue base first, make sure that we're investing enough in product. We've been managing significant expansions in our cash flow over the last few years, which have -- which we've matched with very prudent and successful scaling improvements in our product and engineering team, with our 2 new technology directors, Brad Bender, who has 25 years' experience at Google and Otto Berkes, who's got 25 years of experience at Microsoft, with both of them leading our product and technology committee and having now been directors for over 12 months. We're really starting to see a lot more rigor and confidence in the success of those discretionary expansions that we've made in terms of product development. $2.6 million spent on new products, which customers are telling us they really now want, right? And that's the LEXI Audio Description product, it's the LEXI Recorded product, it's LEXI Voice and it's the LEXI Encoders, all of which are pre-revenue. And so to the extent we expense all of that, as I said, because we want to keep our balance sheet clean for U.S. investors, we will continue to highlight what component of that R&D spending is on pre-revenue, so that shareholders can get a cleaner view on effectively what the underlying EBITDA growth is. And we'll continue to obviously report on the scale of LEXI as it grows out. And really, it's about making sure that we continue that -- the growth that we've seen of that 1,200% increase in the last 5 years. That's showing customer adoption, right? And what we want to see is that as a key underlying metric continuing improvements in cash generation with the subscription model. And if we can convert say, 3,000 of those 5,000 legacy encoder customers across in the next 2 years as well as selling new encoders, that is a target certainly that we've got. So we will be reporting on a pathway, hopefully, to much more than $150 million, but we're going to do it in annual increments that are approved by the Board. Remember, the old target was not approved by the Board. So this should hopefully give investors a lot more confidence as well as being able to really see into what is a much, much simpler business model.

What we might do is group some of these questions together. So I think if we start with our hardware. Can you touch upon some of the weakness you saw in hardware sales over the half? Were there any geographic isolations in the weakness? I think I remember the Europe expansion was expected to yield some uptick here. Could you also provide color on the expectations for this to bounce back in the second half of FY '26?

Yes. So the hardware sales is a number that we want to completely transition out of CapEx and towards OpEx. So on the old model, which is the model that we're reporting on today, right, because we didn't start the Hardware-as-a-Service model until January, right? So everything that you're seeing in terms of hardware is actually a hardware CapEx sale. Well, if a customer buys a piece of hardware in 1 half and then they replace it 5 years later, you're going to get that revenue this year once and you'll get it in 5 years' time once. So what we're actually doing is we're converting that over into a monthly recurring revenue subscription. We have held off selling the old units under the old CapEx model, because there's no rush to kind of replace units, like they do keep working. But what customers really wanted was an upgraded unit that was AI native,1 that had all of the latest security implementation on it. So for example, we've got real-time threat monitoring now built in to the underlying architecture of these new encoders. And what we are also doing is we're giving customers trade-ins on their old equipment if they bought it within the last 3 years. So we made a conscious decision not to push encoder sales in the first half of last year for customers that we would then just end up replacing that equipment with. So it was really just net new sales to new customers in that half. And then what we are hoping to see is an increase in the number of units that we ship under this new Hardware-as-a-Service model, which we'll be able to report on at the full year results. Jason, is there anything you wanted to add?

No, I think I covered it in my section. Look, we're expecting new encoders to be sold and obviously a transition of a portion of those 8,000 encoders out in the field. We have gone to our largest customers who hold 200 to 500 encoders each and have floated the idea of the subscription model, which they've embraced. They do like the idea of not being responsible for them, having the security with us, having the encoder refresh as scheduled. Any upgrades get pushed. If they ever have a problem with them, they send it back, we give them a new unit. So I think going forward, that model would be good for us. I think we found that the lumpy hardware revenue doesn't really fit our model very well. We want recurring subscription revenue that increases every single half. So moving to this model over the next, let's say, 12 months, I think we'll see some substantial uptick in hardware revenue.

Thanks, Jason. Sticking there, can you -- Laffin MST says, he can understand the hardware switch to HaaS. Can you clarify how long you expect the turnover of hardware to last? And are you largely just spreading the cost of hardware over that period?

Jason, do you want to take that one?

Yes, correct. We will be matching the costs, basically matching the cost with the revenue. So we won't see a massive margin erosion. I think we still -- for our margins for our hardware typically around about 80% as well. So yes, we'll be spreading the costs. We do see effectively just the compounding permutation year-over-year. So after you transition the first, say, let's say, 3,000, the next year, you will have that 3,000 recur, plus anything incremental that you'll sell. And the cost will follow that model as well over the period. The key difference here is that actually the hardware sale isn't one-off and it doesn't stop on a subscription. Effectively, what ends up happening is at the end of your first cycle of, we'll call it, 3 years, you will get a new upgraded encoder and that whole subscription will continue. So I think what customers do like is the fact that we've transformed the whole business into that monthly billing cycle for them. They get one bill for SaaS. Now on top of that, depending on what support tier they choose, that will be added on. And then they'll have their hardware as well as a recurring subscription. So for them to get through rather than spending $15,000 as a one-off as opposed to $500 per month, seems more palatable. And the thing that we've done to make it more enticing is kept at low risk. If they don't like that model, they can return at any time. There's no break fee or break costs. Having a lot of subscription contracts myself and approving them, the biggest thing that I find is they lock you in for very long periods of time, and it makes it unattractive. So we don't want that. We really want to encourage our customers to shift over to this HaaS model. I think that will have a good impact for us.

Great. Thanks, Jason. Also, can we just switch to voice here? Could -- in terms of the strong acceleration in SaaS revenue, was any of this driven by the LEXI Voice rollout? Could you provide some indications on whether you've had many client wins or strong uptake with the new product?

None of the SaaS growth is LEXI Voice. LEXI Voice is pre-revenue. So anything that LEXI Voice will deliver will be on top of that. As I said, this is a process of adding an orchestration layer to live video. And there are certain elements within the 9 Squares for which we have had great success with LEXI Voice at the moment. So they are American government. So when we're doing a lot of live translation, English into Spanish. So we reported, I think, at the AGM, it was last year. So what would that be? 4 months ago, we had our first city of customer who was replacing Spanish interpreters in well, effectively was like city hall council meetings. So we've got several more of those, which have been really good. But probably the one that I'm most excited about in terms of LEXI Voice, and it's a great question by the way, is the enterprise sector. And it's the enterprise sector across all regions. And so if you think about it in terms of Fortune 500 companies, I won't name the company, but it's a large pharmaceutical company that's got headquarters in the United States and in Germany. And we've just completed a 1-month unpaid POC where local, all of the internal meeting -- meetings that they were having that were secure, it -- just internal meetings, had 4 language options delivered with LEXI Voice. So you had English, French, Spanish and German. And the feedback -- and this is pretty fresh, right? Like this is -- we just got this feedback last week. I think we just finished the pilot -- the 4-week pilot last week. And the feedback was, yes, we want it for the next year, but the German sounds really robotic still. And could you improve the way that the German sounds. We really like the way that the English sounds. The French sounds great. The Spanish is perfect. The translations are better than anyone. We love the fact that it's always on, right? And they're talking about potentially 3,000 hours, for example, to book -- pre-book effectively for next year. So what Jason will do is he'll package up a deal that will have LEXI Voice encoders. We'll have one encoder or 2 encoders in New York, a couple of encoders that will sit in Germany. We will get Kate and the sales engineering team to ensure that it's reading from all the right document libraries and archives and where they sit in the pharmaceutical company's knowledge management system. And then we meet with them through effectively a sales and support pod structure, which feeds into that support as a service model that Jason is talking about. Where we meet with them and we review the progress on both the implementation and where they're looking for product improvements. And a large part of the marketing investment that we make is in these conferences and summits. So at the Ai-Media Annual Summit in Palm Springs last month, we got all of our top 20 customers together, and they gave us feedback on what they really want to see in terms of priority. And that all came down to better denoising, which is why we're putting the AudioShake application on the brand-new encoders. They wanted lower latency, so a beefed-up encoder with more GPU grunt, and this is really important, the data needs to be sovereign. Everyone was perfectly happy, even the Europeans to still buy 100% made in U.S.A. hardware, but the data must never leave their environment. So what we're effectively doing is we're kind of building a series of different contained networks that mean if, okay, you need to ensure that this data never leaves your organization, that's one particular layer. What if you also have data sovereignty requirements so the data can never leave the EU? That's another layer. What if you want to make sure that personal identifiable information never, ever, ever gets out or into the system and you want to make sure that if there is any PII, it immediately gets scrubbed? That's done in real time. How do you know that you haven't got a threat actor that's come into your live stream and is lurking there? Well, we've got real-time threat detection. All of these things are the moat and scale that we've built, and it's why we're now delivering 1,200% more volume than EEG was, before the acquisition. So that's what kind of where we're seeing voice immediately the opportunity, I would say, where we can already plug into an existing organization's requirements. Where we're not seeing it yet because the technology is not quite good enough, the sentiment isn't quite good enough, none of our, say, broadcast partners think that they could yet spin up a football match in Japanese and get an equivalent kind of commentative view on that, but that's where we're heading, and that's what a lot of the R&D going forward is going to be about.

Can you just touch on, again, how is the push into Europe going following the regulatory change and recent marketing efforts? Do you have any specific contracts to talk to or activity?

Again, so the regulatory change happened in June of 2025, and that was a requirement for the very first time that -- under the European Accessibility Act that European broadcasters introduce captioning. That supercharged encoder sales into that region as a one-off. Remember, that was as a CapEx model. Those encoders are going to last for 5 years. Many of those customers will do trade-ins on this new model going forward. What we have been doing in Europe, though, is we've been steadily expanding the LEXI services that we've been turning on for these customers, and you can see that in the reported numbers. Continuing to -- so -- say, for example, we haven't yet solved the problem in Greece, right? Greek doesn't work. Albanian does. Don't ask me why. But these are still areas where we're continuing to focus. And when we do finally get Greek sorted, then we've got another market there that we can work with. Similarly, in Mongolia, we're working with our first index customer there, CNBC Mongolia, but there are some Cyrillic characters that just aren't showing. And so this is the way that we're really focusing our product and R&D efforts as well as taking the feedback directly from participants at the Ai-Media Summit as to what they really want to see in their workflows right away. And so if you're looking for near-term revenue growth, it's encoders.

Okay. Great. I guess let's stick to HaaS then, given that encoder comment. There's a few questions here on that. What are the balance sheet implications for the HaaS transition? Jason, I'll direct that towards you.

We're still working through that with our auditor, but it effectively is not a leasing option. It is -- we have to carry some of the costs on our balance sheet, obviously, with the inventory, because the right-of-use asset sits with us and no longer with the customer. So on the balance sheet, you will see a large amount of right-of-use assets as that increases. But again, that will flow into the P&L.

Great. Okay. And then, if we switch back to ARR, can you just explain how your contracts have historically been structured? And how will they change going forward in terms of how you're calculating ARR?

ARR, we look at recurring subscription contracts on a monthly basis, that is locked in long-term. So we remove out any subscription services or Software-as-a-Service that isn't a subscription, that is a one-off LEXI product, for example, a one-off event like an Olympics in a stadium. That -- obviously, that comes across every 4 years. So we remove all of that. We remove anything that is one-off, transactional and just stick to the traditional subscription model. But a lot of our services and non-subscription are transitioning into subscription. It is becoming more and more attractive for the customer. So eventually, I think that 70% of our revenue will be all based on a subscription annually recurring model. I think this has been a conscious decision for us to transition completely into this model. And then all you'll have left over is the nonrecurring SaaS component, which happens -- a customer can still buy hours for 1 month, let's say. That's still possible if that suits their need. Or in some instances, customers are -- I think, for the time being, are attached to the CapEx model for encoders. So we're not going to turn that away. We're going to allow for that, but encourage a transition over to the HaaS model. And then lastly, you will have your engineering hours that we charge on time and materials for implementation or any support work that we haven't been charging for. So I think the business in the near-term -- near to medium-term will be largely a recurring business. We will -- that's one of the reasons we're now directing everybody to the ARR metric rather than a target that's out 3 years from now, that we do have confidence in the ARR model. We're working towards it internally, and we want the market to be focused on it.

Great. And then Stella has asked, can you confirm the target exit ARR for FY '26 is $3 million a month from its current $2.5 million a month?

Yes, confirmed. So we...

Yes, which would be annualized at $36 million, yes.

And our guidance is on pcp, which was 24%. We're aiming to grow that to $36 million.

Michael has asked, if you are moving to ARR, will you also measure net revenue retention and percent of churn so that investors can understand the quality of the business?

The churn has been minimal, as you guys know. I think once you get the encoders into the environment, our churn -- in one of my slides, we have not seen churn in our top 20 customers where bulk of our revenue comes in the last 5 years. We do not see churn as a big factor in the business. But yes, we will be reporting on any churn.

Okay. So Ken has asked what happened to LEXI DR? He hasn't seen it in a diagram in this presentation.

LEXI -- it's a good question. So LEXI DR is one of our encoder products. That's our disaster recovery encoder product. We also have LEXI Local Encoder products, and we have Alta Encoder products. They're all part of the encoder family in the middle.

Okay. You spent $2.6 million in the development of the new encoder. Do you expect that number to increase or decrease in the second half? And once the product is launched in April, do you expect this R&D cost to cease?

I didn't say that. I didn't say that we spent $2.6 million in the last half on the new encoders. I said we spent $2.6 million on pre-revenue products, which included encoders, LEXI Voice, LEXI AI and LEXI Recorded. We did not spend $2.6 million in R&D on the new encoders in the last half. So let's just be clear about that. We will continue to be spending money on pre-revenue products, and we will increase, in fact, the amount of R&D that we do as a percentage of revenue because of the success in the initial proof of concepts and the need to now scale the go-to-market function and to continue to add features, implementations and improved security architecture for these target customers.

To add to that, to Mel, I think we spent $2.6 million on company-wide pre-revenue product development. The other thing to note is that there is also a low revenue investment that we've been making into Asia and EMEA. We're still very new territories. A lot of investment in go-to-market has been made into those regions that we have not encapsulated in that $2.6 million number. So currently, the investment base, and that's why you see a flat EBITDA is high. And we do see that switching to a point where the revenue will come in as well.

Can we switch to hardware? Stella has asked, how much hardware was sold in the first half? And when do you expect the upgraded version to start selling? Can you repeat how much year or by month customers pay under the HaaS model?

I don't have the exact number for hardware sales, the hard number. What I do -- what I can offer is the HaaS model has 3 different types of encoders that we're going to be offering, the HD492, which is our current encoder that we're using, and that will effectively go from roughly around $14,000 upfront to $395 per month. Then we've got an upgrade to that version that has some voice features, which is the HD592, which will attract an additional $100 per month on that to $495. And then finally, we have our Rolls-Royce type encoder, AIX1 which we've been working on sometimes, which can do a lot of different things that all the other encoders do, and more, and has multiple languages on it, and that will be selling for $595 per month.

And the USD -- I should just say the USD prices as well.

Okay. And then just to continue with Stella, how much of the service fee is now an implementation fee? In first half, service gross margin was 35%. Is this the level we should expect for future implementation?

No, no, no, no, no. That's -- Stella -- Stella has got it -- It is confusing, right? This is -- this half was legacy services. So we were still doing ReSpeakers. We were still doing stenographers. We still had legacy human infrastructure services that we switched off in December. That is what that line is. Completely gone, completely separate from what Jason was talking about in terms of professional services. And Jason, I'll let you talk about that model, the professional services model.

Yes. So like Tony said, our professional services has now started to roll out. We will see another line in the P&L that's going to be professional services. We will separate out any overhang of the legacy services going forward. So we typically have priced our professional services at 50% gross margin.

Great. Jason, going back to ARR, can you reconcile how the old ARR measure reconciles to the new ARR measure in terms of the FY '26 target? Do you think in the future we could maybe put a reconciliation bridge in future results?

Yes. When we first came out with the ARR metric, we apportioned the total revenue and apportioned what portion of that was ARR, which came out at $17 million. But when you look at the traditional mathematics on ARR, it's on exit on the final month. So on exit on the final month, actually for FY '25, our ARR, our monthly recurring revenue was $2 million and our annual recurring revenue, therefore, was a multiplication of 12 months, came out at $24 million. So we will be following the normal method, which is on exit. So on exit on a monthly basis by this end of this financial year, we're expecting $3 million in recurring revenue -- monthly recurring revenue multiplied by 12.

Great. That brings us to our last question for the session, Tony. So I mean, Nicholas has asked, what's your message to shareholders given the removal of the FY '29 target and the share price fall?

Look, the reported numbers here show a clear transition that's now been completed. I think the underlying economics and the underlying trajectory of the scale of the business show a platform that has a lot of value and increasing value to our customers. We are moving now with certainty to an annual recurring revenue forecast model, one that is Board approved, one that is backed by the data and one that gives a lot more granularity than a point-in-time CEO forecast 3 years down the track. So I think this is a significant upgrade in precision. And if I were able to have done this 2 years ago, I would have done this 2 years ago. It was really because we were still in the midst of that transition. Remember, we still had 2 years to go to shut down the legacy services business. We were still selling hardware in terms of CapEx. We've now completely simplified this business to one that is going to benefit from that scale, that you can see that we've delivered over the last 20 years as we begin to turn on these new products and in these new geographies. So I think the business is in a great place. I think a lot of the hard work that we've done is behind us. And now it's a question of working week-by-week really with a lot of our customers to kind of panel beat the final elements of what is an orchestration layer that fits in with their live broadcast system and in a way that can deploy a fully integrated solution that looks at the best hardware and the best software with the best AI tools to deliver a secure, low latency, highly reliable service that is auditable and delivers a consistent result to consumers. That's what we remain focused on, and that's what we're going to continue to report on. So hopefully, shareholders see this as improved commitment, improved precision and more and more consistency, so that they can really see how we will get to that trajectory of growth that remains our ambition.

Great. We might end there with any closing comments you have. If anyone has any other questions on the results, feel free to e-mail me later today or this week. I'll pass back to you, Tony.

Thanks, Mel. Just say thanks, everyone, for joining. Very happy to take follow-up questions. I'm on ausbiz this afternoon, and I know we've got a lot of meetings with all of you scheduled in the next few weeks. I really think what these numbers show is the completion of the structural transformation of what was effectively 2 declining business units and one business unit growing strongly. We have demonstrated a high-margin, high-growth scaling recurring infrastructure platform tailored to live video. We've had 0 churn of our customers in the last 5 years, and our volumes have grown 12-fold in that period. The feedback that we had from those key shareholders at our annual event was very exciting. We've just got the feedback from -- back from the session, and we got an average of 4.8 stars from those customers participating. And we've made significant changes to the road map to really get faster product-market fit, now as a result of it. So hopefully, over time, we'll just continue to see everything moving up and to the right and people can just have confidence that the business is on the right path because we're now focusing on those very traditional simple metrics of annual recurring revenue, annual recurring revenue growth, ARR margin and expanding EBITDA. So we'll continue to report on that. We'll continue to separate out any of those R&D and product investment decisions. Half of that product and R&D investment was in pre-revenue products. We are going to increase our investment in product and R&D because we're seeing immediate opportunities for accelerating product-market fit. And I'll continue to update on LinkedIn frequently with lots of the operational updates and product updates that we've got. And stay tuned for the big release of the new encoder products in Las Vegas. So thanks, all, very much for joining and look forward to the follow-up conversations. Thanks, Mel.