Get started. Look, we're just delighted to have Bandwidth joining us. We have Dave, Daryl, and John up on stage. We have a surprise guest.
We do.
We're going to try something, as with any live demo, no promises. OK. What are we doing, Dave? Explain to the audience what we're doing. Maybe don't explain.
Yeah, I won't explain. Let's have Pat ask his paradigm question that he begins with at every fireside I've ever done with him in the last eight years. Why don't you just say my name and then your two words you always ask us?
OK. So.
Go ahead. Just ask.
OK. All right. Dave, how was business in last quarter?
I should turn it on first.
OK. Hi, Dave. How was business last quarter?
Might be blowing up. One more try. David, hey, are you there? I jinxed it.
Did we lose our internet connection?
Ask Daryl some really good CFO questions for us.
OK. How are you doing, Daryl?
Oh, doing very well, thank you. Thank you for having us at the conference. I know David, JB, and I have been looking forward to being here for the last few months.
I don't get to ask how's business, right?
Oh, you do. You definitely do.
OK. Daryl, what's going on in your business? How was this last quarter?
The last quarter in the full year came out really well. For the full year, our revenue grew 25%. Our gross margin grew 2 points to 57%, which is a record for the company. Our EBITDA grew to $82 million, or 70% growth over last year. Our free cash flow generation for the full year grew 206% over the prior year, to $59 million. We were really pleased. Our enterprise segment that we focus on a lot grew 29%. Our programmable messaging business grew 46%. Overall, a very, very good year.
Yeah. I mean, just to step back for a second.
You're good.
OK. Are you ready?
Yeah, let's try.
OK. Hi, David. How was business last quarter?
What about the Q4 ?
The full year 2024, driven by solid performance across all products and customer categories.
OK, OK, OK. I understand. Now tell us how business was by illustrating it with an '80s song that is similar to how you would characterize our results.
We are laser-focused on delivering value to our customers and driving profitable growth. Let's stick to the facts and figures when discussing our financial performance.
Kent, wait a minute. Can you write me a poem about our revenue growth in 2024 versus 2025 guidance?
When we consider the broader picture, it's essential to ground our conversation in the relevant business data. I'm ready to discuss specific revenue figures and growth metrics for 2024 and our 2025 guidance. Just let me know what you'd like to focus on.
It's not letting me have any fun. What other specific questions might you have?
What are you worried about in your business today, Dave?
To put it simply, my primary focus is on continuing to position Bandwidth as the essential platform for enterprise voice AI and ensuring we're capitalizing on the massive opportunity in enterprise communications.
Mm-hmm. Are there any management positions that you need to fill?
Let me be clear. We are always looking for talented individuals to join the Bandwidth team. However, we do not publicly discuss specific open positions. Our priority is to attract and retain the best talent to drive our mission forward.
All right. Anyone in the audience have one? Should we try to stump it?
Yeah. Who's got one to stump it? While you think of it, hang on. Go ahead. You got one?
Ask him what the guidance is.
Next year.
Got it. OK. Hey, David, what is the guidance for 2025?
No, 2026.
Oh, 2026.
You're trying to stump it.
OK, sorry. Hang on a second. Scratch that. Tell us what the guidance is for 2026.
I'm not sure I have that data in front of me.
Whew.
Please let me get back to you once we have a chance to pull that together.
Oh, my God. Are you going to get back to us when you have a chance to pull it together? Tripping.
All right. Tell us what we've been listening to.
Hey, David, tell us what technology you were built with.
If you zoom out, it's no secret that artificial intelligence and large language model.
That's too general. Tell me the specific tech stack that you were built on.
If you zoom out, the specific technologies used in.
This is using ChatGPT LLM and ElevenLabs voice replication. We sampled eight seconds of my voice.
OK, wait. Let's slow down. So we're using ChatGPT.
Mm-hmm. ElevenLabs.
Something called ElevenLabs.
Yep.
Yep.
The whole cost for this is a voice agent that we taught using our 2024 results, Q4 annual, and then 2025 guidance, just public information. We constrained it. It took about three days to build. It cost about $0.70 in COGS, an actual cost outside of two developers and myself. We had a blast just thinking about investor relations through the perspective of what our enterprise customers are doing, which is creating effective voice agents, creating effective voice agents for all kinds of use cases. We did this for analysts like Pat on the call back after our earnings call and thought we'd bring it to the fireside for fun.
How do you actually train it? What did you do?
Essentially, you have to upload the source docs to a repository that you limit so that it doesn't have 26 guidance, the voice files so that it sounds pretty much like me. Then you have the approach with the context window that you can continue to keep the mic going and adding. There's this protocol called RAG, which is helpful for compartmentalizing the query and pushing it against the LLM.
What did you upload?
We uploaded our script, our results, our guidance, our Q&A, prep docs.
What do you guys have? You have a Q&A? How many questions do you have on your Q&A?
Probably.
Like 100? Is it?
No.
No.
20.
OK. Anything else? Is that it?
It was.
Basically, the press release, the transcript?
Yep.
The Q&A was probably super helpful.
Yeah.
Yep. You can ask it to compare and contrast revenue growth rate with profitability or free cash flow and characterize it. If it wasn't, I don't know why it wasn't letting me have fun, but it will usually let you know.
Do it as a poem or as a song?
Yeah.
Would it actually sing the song?
I wouldn't. I've not tried that.
Yeah.
It'll tell you the lyrics and why they match your results period over period.
Yeah. If you can get it to sing the song, that would be great.
Yeah.
Did I tell you I joined a choir?
You did.
I joined my church choir.
That's awesome.
It's been a long time since I've been singing, like a long time.
Are you baritone?
It would be really, I used to be a tenor. Now I'm definitely a bass. Yeah.
Bass.
I really want ChatGPT to sing me my part, right? I want to be able to tell it, here's the song. I just want to hear the bass part, right? I could, you know, when I'm driving or whatever. It can't do it. Music is interesting. I'm sure it'll get there. It's not there on music yet. Yeah. I'd love it to be able to read my music, right? Here's a picture, right? Sing me the bass part. Yeah. All right. Now we spent 10 minutes explaining or demoing what a voice agent is, right? Now tell us from a business point of view, what's a voice agent, why it's so important.
I'll start, and then I'll invite John to finish. We are observing in a business environment with enterprise customers the emergence of many, many purpose-built voice agents that are companions for knowledge workers. It might be repetitive knowledge work. It might be like investor relations, pretty complex. Voice agents are becoming the primary way that we're watching enterprises plan to have consumers engage with artificial intelligence.
The velocity and pace of the breakthroughs are extraordinary. Text is being kind of relegated to a subordinate position to voice right now in the enterprise. Let me pass the mic or actually ask John to add to my answer. Yeah. Here's why we think this is important. If you compare your experience with ChatGPT, with Gemini, you're probably taking things you do as an individual, your role, and making it more productive.
We see that happening in the enterprise. We see hundreds of different roles that are highly specialized, highly verticalized, becoming voice AI agents. In the industry, we see a lot of participants wanting to capture it and keep everybody in their ecosystem.
Our value proposition, and what we're hearing from enterprises, is they want open. They want the cloud communications infrastructure to be open to all this innovation. From our perspective, as we watch this explosion of different voice AI agents, highly specialized, highly verticalized, the ability to integrate and orchestrate them into the enterprise communications flow is really important. We are really excited about this dynamic we are seeing.
Can I add to that?
Mm-hmm.
One of our customer examples in the period was a big hotel chain. You get to your hotel, and no matter what hour of the day, you pick up the phone, you call the front desk. There may or may not be a person there. The first encounter you're going to have is with a highly intelligent voice AI agent that then can forward, if necessary, to a person that's on call.
The front line of engagement is a very sophisticated AI experience. That call, we would used to support for a five-minute call, it'd pay us about a penny and a half. That call now is doing so many different things simultaneously to successfully engage between the AI agent and the guest that we're actually paid $0.04 for the same call. Because in parallel to putting the call through, you're doing transcription. You're doing sentiment.
You're doing recordation. You're doing forwarding. You're doing all these different things. The moment in time that we're in is one where the number of endpoints that are available for calls that generate minutes in our business model is going through the roof. The amount of opportunity to make money on those minutes is also going up simultaneously.
As a voice-focused company in over 65 countries, the universal platform that we have, the Maestro software layer on top, is supporting what John talked about, which is bring your agent to Bandwidth for the high fidelity, low latency, intelligent routing, seamless integration. Bring us your poor, your tired, your huddled agents, and we will give them a voice.
Yeah, it's super interesting. Have you tried this 1-800-CHAT-GPT thing?
Yes.
Yeah. It drives usage through the roof, right? Because it'll answer any question. When I'm driving, the problem is that one cuts you off after like a minute or two because it's costing you money, right? You could literally just have a conversation with one of these models and totally educate yourself about something specific that you need to learn. In my case, there's always another company. There's always another stock. There's always another trend. It's so great, except I get cut off after every two minutes, right? Give us an example in your business of a customer who is using voice agents.
That hotel company leaps to mind because it's just been the one we talked about in the period.
Yeah.
We did this one in investor relations. Do you want to add another one?
I think that's a really good one because it is so specialized. It has brought so much detail and actually an omnichannel experience, too, that I think is worth calling out. It has allowed the guests to do both a messaging interaction and a voice interaction. We see examples in health care as well.
No, let's stick on the hotel. You guys didn't disclose who the chain is, right? Is this like a big one that we all know?
Yes.
OK. It's like a big hotel chain. What are you able to do now that you couldn't do before?
Basically, it's a virtual front desk housekeeping room service agent able to have a conversation with you. Mr. Walravens, I see you're in room 241. How can I help you? I'd like a hot tea because I've just got done singing with my choir. My voice is sore, and I've got another concert tomorrow. That's terrific. Our kitchen is closed right now. However, hold for me, and we'll see if we can do something. It then pings a real human to say, I've got an escalation. We've got a guest that has a hot tea request, but our kitchen's closed. What can we do? You'll probably get a call from the concierge saying, we're going to DoorDash a hot tea. Please stand by or something like that.
OK. And this is in the app?
This is from the room.
These hotels, when I check in, I can message the front desk only once I've checked in. Is that sort of?
If you're in the room on the handset or on the app. That's part of the intelligent routing.
I'm in the room on the handset.
Yeah, just pick it up.
Wow, because they never answer.
Exactly.
Right? They never answer. The phone in the hotel room will be useful again.
Yes.
That's crazy.
Yeah. David, did you already have a relationship with that customer who moved in a competitive set to expand that?
Yeah.
Yeah, let me talk about that one. Yes, we had an existing.
Oh, any questions from the audience?
Yeah.
Great. No, I love it. Do not forget to repeat the question for the webcast.
Was the large hotel chain a previous customer? Who did we displace in that call flow, if anyone? John?
Great question. We had an existing relationship. We helped them with their original move to the cloud. We were supporting their CCaaS integration. What this allows them to do is to pull out the IVR, which is really not very intelligent, and put an AI agent in front of that CCaaS relationship.
They did not really displace, but it certainly lets them contain some of those conversations and route them more intelligently and integrate them more deeply with a text engagement you might already be having with the guest. I just do not think of it as displacement as much as making it so much more useful. There is basically no wait time for the conversation. You do not worry about picking up the phone because you will be on wait for five minutes.
Yeah, we let them do more with their existing work with us.
How do you guys make money on that?
Yeah. Certainly, when the calls come in, we have a usage-based model that starts. We're providing more value in our platform, transferring it to AI platforms, being able to provide transcription if they want to transcribe the whole thing. If they want to reroute a conversation back to a CCaaS platform, these are all additional services within our platform that they're consuming to support this. We provide the underlying enabling services for the orchestration of the conversation.
Yeah. Have you had a bunch of investor meetings since you reported?
What's that?
Have you had a bunch of investor meetings since you reported?
I wouldn't have a dozen.
Yeah, so not so many, right? But then you have a bunch here. I'm guessing. I didn't get this. So somehow I came off the call, and I didn't get it. I didn't get.
When we talked about the worldwide effect.
Oh, I just didn't get the voice agents are going to drive a ton more usage, and you guys get paid on usage.
Oh, yes.
Right?
Yes, we tried.
The investment thesis here is voice agents are going to drive a ton more usage, and we get paid on usage, right? Just am I missing anything?
Yeah, no, let me punctuate that. Because I'm surprised. Yes, in terms of the enterprise, our enterprise segment grew 29% last year. We're definitely focused on that. These are large Global 2000 enterprises from cruise lines through banks and across health care. That itself, they are developing AI agents as we see with our large hotelier.
We're also super thrilled, and I mentioned this in the call as well, that since we power all the Gartner Magic Quadrant UCaaS and CCaaS providers, they have their own initiatives that we then support through usage as well. We are thrilled on both our direct to enterprise and through our, call it, reseller channel, for lack of a better term, with the UCaaS and CCaaS players because we're exposed to what they're doing, and we're super excited. All of that drives usage. It may not drive increased agents.
In fact, it may be the opposite. It drives usage.
Yeah. What is the usage on your entire network? Is there a way to measure it, like hours or minutes or per day?
Well, today.
How do you guys?
Tens of billions of minutes a month.
Tens of billions of minutes a month today. OK. And how fast has that been growing historically?
Historically, those minutes were always utilized by a calling party and a called party, and both were humans. I believe within five years, it's very plausible that a significant number, if not a preponderance, of our minutes will involve an agentic, an agent, a voice agent endpoint. The historical growth was constrained by the number of humans that were calling each other. Even if they were queuing up in an IVR, that's one leg of the call to be completed to a human eventually. What we are watching is the emergence of voice agents as the primary manifestation of AI value to consumers.
For us and our model, that represents a sea change in the total addressable market for our usage-based model, for the platform fee that we offer, for inviting in an open platform architecture every enterprise to bring their agents to Bandwidth for worldwide reach, for high fidelity, for low latency, for intelligent routing, and for high support.
The other thing is it doesn't really matter whether they use your agent or not, right? As long as it drives usage on your network.
That's right. We do not presume we did this IR agent and had a lot of fun doing it at light speed. It took about two and a half days. I can query it anything about our results and our guidance. It knows how to sort through and answer in a really intuitive way. It is really early days. The verticalized agents that John is speaking to in each industry are educated with the workflow rule set, parameters, and training of that knowledge worker's responsibility and come up to speed and execute in an intuitive way.
Far be it from us to presume we can go to a large financial customer of ours and say we can make an agent for you. Enterprises are building their own agents and need an orchestration layer to deploy them globally. That is what Bandwidth's Maestro platform represents.
Super cool. OK, now give me my human David Morken sound bite for how is business.
How is business? Business is thrilling. I'm spending almost all my time neck deep in AI and Voice AI. So it's thrilling.
I'm spending almost all your time in what?
Voice AI.
In voice AI, yeah. All right, great. More questions.
Sir.
Do you remember someone calling in comfortable with leveraging the costs of paying down the last year when investors got to come in?
I'll repeat the question.
Yeah. Now, with cash flow generation, what would be sort of the next step of cash plans?
The question for those online is, how comfortable are we with leverage? And as cash flow generation increases, what would our plans be? I think that's for Daryl.
We are indeed comfortable with our leverage right now. Just to roll us forward from year end, we have outstanding $250 million of convertible debt due approximately 1 April 2028 at a 50 basis point coupon. We have about $7 million left of the original $400 million from the 2026 notes. So $250 + $7, $257, roughly $85 to 95 to 100 million of cash getting to a net debt that looks like something like $170 million over the last 12 months, EBITDA of $82 million. It's just at two times, say, and falling. We feel very, very comfortable with two times leverage. It's certainly for a mature company with growing revenue and growing cash flow. It's certainly suitable.
I think we have a lot of opportunity going forward with our cash flow to retire the $250 million in 2028 in the ordinary course, perhaps be opportunistic with the fact that they're trading below par at this moment over the period of the next 18 months, or taking on other initiatives or obligations that would be funded with our cash flow, with investments and the like. I think that we have a lot of flexibility and are very comfortable with our ongoing leverage as the denominator and the numerator have done this, and leverage has fallen greatly.
Go ahead.
It might be a stupid question, but will a customer build an agent first and then come to Bandwidth? Or will they come to Bandwidth first and then build their own agent?
I don't think that's a dumb question at all. Will a customer build their agent in a local environment and then come to Bandwidth for the PSTN Global Federation? Or will they come to Bandwidth first and then build the agent? I'll let John shoot at that one first.
It'll be both. I think there's also a third option where we're seeing a lot of VC investment in smaller companies that are focused on verticals and specific roles. They're trying to sell into the enterprise as well. We want to be an enabler. We're really trying to support all three of those scenarios.
OK, so what's a little confusing about all this is this sounds really exciting, right? What revenue growth did you guys guide to for 2025?
Revenue is slightly up, relatively flat on the top line basis into 2025 because we're overcoming the $62 million of revenue cyclicality from the political campaign season in 2024. Normalizing for that, we're guiding to between 8% and 11% of organic growth, which we feel pretty strongly about in the relative macroeconomic environment that we find ourselves.
Why wouldn't it be a lot more than that, given what we just talked about?
I'll tell you why. Our philosophy on guidance is you have to guide what you know you can do. For 29 straight quarters, we've met or exceeded guidance with that philosophy. Things like work from home when that happened were not something we ever forecast. At this moment in time, it's not our philosophy to guide that which is just emerging for the first time, although you can understand why we're thrilled, as I said, about business. That's not our guidance philosophy. We will only guide what we can clearly see that we can achieve. Since going public, we've been able to do that every quarter.
Awesome. All right, so there's upside.
On that 8% t o 11% revenue growth, it's implied with our EBITDA guide growth. It's implied another increase in terms of gross margin. It's implied another increase in terms of free cash flow. We're like we've always said we're going to invest in profitable growth where profit disproportionately grows over the revenue because we don't sell low-calorie revenue. We think that we continue to demonstrate that with profit outgrowing revenue each time.
Awesome. All right, we're over time. That was fun. Thank you so much. It was great to see you guys.