Ladies and gentlemen, thank you for standing by and welcome to Agora Inc.'s 3Q 2020 financial results. At this time, all participants are in only-listen mode. There will be a presentation followed by a question-and-answer session at which time if you wish to ask a question, you will need to press star and one on your telephone. I must advice you that this conference is being recorded. I would now like to hand the conference over to your speaker, Ms. Fionna Chen. Thank you. Please go ahead.
Thank you, operator. Good evening and good morning, everyone. My name is Fionna. I am the Investor Relations Director at Agora. Thank you for joining Agora third quarter 2020 earnings conference call. Joining me today are Tony Zhao, Founder, Chairman, and CEO, and Jingbo Wang, CFO. Our earnings result press release and a slide deck can be found on our IR website at investor.agora.io. Reconciliation between our GAAP and the non-GAAP results can be found in our earnings press release. During this call, we will make forward-looking statements about our future financial performance and other future events and trends, including guidance. These statements are only predictions that are based on what we believe today, and actual results may differ materially. These forward-looking statements are subject to risks, uncertainties, assumptions, and other factors that could affect our financial results and performance of our business.
We will discuss them in detail in our filings with the SEC, including today's earnings press release and risk factors and other information contained in the final prospectus relating to our initial public offering. Agora assumes no obligation to update any forward-looking statements we may make on today's call. With that, let me turn it over to Tony.
Thank you, Fionna.
Tony, please.
Welcome, everyone, to our. Hello, you hear me?
Yes.
All right. Thank you, Fionna, and welcome everyone to our earnings call. This is the first quarter in 2020 that was impact-free from the COVID situation in China as COVID-19 gradually faded off between late May and early June. Even without the short-term demand caused by COVID-19 in China, we still deliver another quarter of outstanding results with revenue of $30.8 million for the third quarter, an increase of 81% year-over-year. Our number of registered applications reached more than 240,000 at the end of September, adding nearly 10,000 per month in Q3. Our number of active users reached more than 1,800, up 95% year-over-year. Outside of China, we continue to see strong demand for our real-time voice and video product in both mature use cases such as entertainment and gaming, and emerging use cases such as education and virtual events.
We also made solid progress in our product, brand, and ecosystem in this quarter. In particular, I'm proud of our marketing and design team, who created an amazing experience at our RTE2020 virtual conference. With more than 8,000 registered attendees and tens of thousands of unique registered viewers, 170 guest speakers, and over 100 companies all over the world participating. We also had our first RTE startup challenge with more than 130 startups joining the competition. More than 75% of these startups are already funded, and 50 of them were selected to join our startup program, where they will receive training, platform usage credit, and dedicated technical support. At Agora, we constantly support startups at an early stage to cultivate innovation and accelerate the emergence of new use cases for RTE technology.
At RTE2020, the discussions will focus not only with what exists today, but how to reimagine tomorrow's innovation or innovative RTE application and build the future together. The traditional concept of communication is all about exchange of information, creating a successful Real-Time Engagement use case requires a lot more than that. It is about creating a shared moment and space for people to experience and interact with. To do that, we focus our attention on the three pillars of Real-Time Engagement: shared context, interactivity at any scale, and ubiquity. To create shared context, we offer more than just video and voice, also use case-specific tools such as signaling virtual environment, augmented reality objects, content moderation, voice recognition, and so on. We can't build all of this ourselves, we have partnered with leading third-party service providers.
Now, many of those features can be easily used by developers as plugins in our platform. Our plan is to make Agora platform the engine for developing any real-time engagement application. Shared context also depends on having smooth, lag and jitter-free experiences that simulate real life. That is why we are rolling out our experience level agreement or XLA.
With XLA, Agora will not only let you know that the number of minutes you are running on are of premium quality, but we will also guarantee our performance on the four core metrics that are critical to end-user experiences. Channel join successful rate, audio fluency, video fluency, and transmission latency. Since its launch, we have signed up dozens of well-known internet companies for XLA. Of course, fully immersive experience should be available for any size of gathering, which brings us to the second pillar of real-time engagement, interactivity at any scale.
Since the very beginning, we have designed our platform to be highly scalable. For example, recently we supported an education customer in South Asia with more than 50,000 concurrent video users in one fully interactive classroom. At Agora, we feel real-time engagement should be available anywhere to everyone and in any app. That brings us to our third pillar, ubiquity. For this to become reality, we have been making our technology broadly available and easy to adopt. We are committed to supporting all popular operating systems, browsers, and development frameworks. Earlier this year, we added React Native and Flutter SDK to our list of supported SDKs. In this quarter, we also released our Unreal SDK for gaming and RTOS SDK for IoT devices, which will release more space for RTE cases to grow.
As we discussed in the past, we have seen aspiring entrepreneurs who have some great ideas but lack the time and resources to build an app from scratch. We want to make it even easier for anyone to create their own RTE apps. For this reason, we launched the Agora Education Application PaaS and Agora RTE App Builder. With Agora Education aPaaS, you can create an online education application with very limited coding, whether it's one-on-one classroom, small class, or interactive large class with massive number of attendees. Agora's RTE App Builder goes even a step further. Whether you are a seasoned developer or a product manager, you can now create a production-ready conferencing app with no coding at all in a matter of minutes, and you can then customize the design and its specification of the app in many ways based on your unique requirements.
Last but not the least, security compliance and privacy protection are critical to our success. In this quarter, we made steady progresses in enhancing our operational standards, such as best practice in software development, data access management, and end-to-end encryption. Recently, we were awarded three ISO certificates on our information security management system, which further attests to our rigorous security practice and strengthen the trust developers have in Agora. Now let me turn this over to Jingbo, who will reveal our financial results.
Thank you, Tony. Hello, everyone. Let me start by first revealing financial results for the Q3, and then I will discuss our outlook for the full year. Total revenues grew 81% year-over-year to $30.8 million in the third quarter of 2020, which is mainly due to the increasing volume for both our video and voice products. At the end of September, we had powered more than 400 billion minutes of Real-Time Engagement per month on average in this year. As you probably already know, the COVID-19 pandemic has been well contained in China by late May. After which, most offline activity and travel gradually returned to normal. As a result, I would say usage in Q3 had almost no direct impact from COVID in China.
That means that the usage in Q3 was still significantly higher than the same period last year, reflecting the increasing penetration of contextual Real-Time Engagement in our activities and the long-term behavior shift. The usage in the U.S. and [rest of the world] , on the other hand, was still at heightened levels due to both our market expansion efforts and the prolonged impact of COVID-19. Demand for use cases such as education, virtual event, and entertainment continued to grow monthly. Our trailing 12 months constant currency dollar-based net expansion rate was 188% at the end of September. This number has been positively impacted by the spike of the usage this year due to COVID-19. Fortunately, the situation has stabilized in China and hopefully the other regions as well in the near future.
Afterwards, this expansion rate will likely come back to levels similar to what we saw in 2018 and 2019. Turning to cost, expense, and margin. Please note that the usage of our products and timing of our hiring and marketing initiatives might fluctuate from quarter to quarter, which cause our profitability to fluctuate significantly as well. We'll remain focused on capturing long-term growth opportunities and keep making investments on innovation. For my following comments, I will focus on non-GAAP results, which excludes share-based compensation expense. Non-GAAP gross margin for third quarter was 62.5%, which was 6.8% lower than Q3 last year and 6.1% lower than Q2 this year. The decrease in gross margin was mainly due to our international expansion into regions with higher infrastructure costs, such as Southeast Asia, North America, and Oceania.
To that extent, the gross margin was also affected by higher bandwidth costs in China, as one of the main Chinese telecom companies raised their nationwide prices in this quarter. Non-GAAP R&D expenses were $11.6 million in Q3, up 18% year-over-year as we continue to build our R&D team. Non-GAAP R&D expenses was relatively flat at 37.5% of total revenue in the quarter compared to 37.6% in Q3 last year. Innovation is critical to our success. We'll continue to invest significant resources in our R&D capabilities in order to further strengthen our technology leadership and broaden our API portfolio. Non-GAAP sales and marketing expenses were $5.9 million in Q3, up 35% year-over-year, mainly attributable to team expansion and increased advertising expenses. Sales and marketing expenses represented 19.2% of total revenue in the quarter compared to 25.8% in Q3 last year.
We believe we have again demonstrated the efficiency and scalability of our developer-centric go-to-market model. Non-GAAP G&A expenses were $3.3 million in Q3, up 135% year-over-year, mainly due to team expansion and professional service fees related to information security management. G&A expenses represented 10.8% of total revenue in the quarter compared to 8.1% in Q3 last year. Non-GAAP operating loss was $1.3 million, translating to 3.2% non-GAAP operating loss margin this quarter compared to a loss margin of 1.9% in Q3 last year. Adjusted EBITDA was - $105,000 in Q3 with a - 0.3% margin compared to a 1.1% margin in Q3 last year. Turning to cash flow, our operating cash flow was - $1.9 million in Q3 compared to - $54,000 last year. Free cash flow was - $5.1 million compared to a - $1.3 million last year.
Net cash outflow in Q3 was mainly due to the purchase of server and network equipment as we continue to expand our software-defined real-time network. Moving on to balance sheet. We ended Q3 with $675 million in cash equivalents, and short-term investments compared to $641 million at end of Q2. The decrease was due to the negative free cash flow, $5.1 million, and payment of professional service fees related to our IPO. Now turning to guidance. COVID-19 is still an unprecedented variable for our business model, where historical experience may not apply. Our guidance on full year revenue reflects a number of assumptions that are subject to change based on uncertainties related to the impact of the COVID-19 pandemic.
With that, for the full year of 2020, we maintain our previous guidance for total revenue, where we see it expected to be in the range of $125 million-$130 million, which would represent approximately 94%-102% year-over-year growth. In closing, the execution performance in Q3 met our expectations. With this, we are making long-term investments on innovation and will continue to support our developers and customers around the world. Thank you to the entire Agora team and everyone, and hope you're healthy and safe. Let's open up for questions.
Thank you so much. Ladies and gentlemen, we will now begin the question- and- answer session. As a reminder, if you wish to ask a question, you will need to press star and one on your telephone and wait for your name to be announced. If you wish to cancel your request, please press the hash key. Again, it's star and one if you wish to ask a question. Our first question comes from the line of Yang Liu from Morgan Stanley. Your line is now open.
Thanks for the opportunity to ask questions. Three questions from my side. The first one is, could management update us in term of the progress of penetrating online education customer in term of their big class use cases? With the big potential there and previously management mentioned that some big customers doing testing, could you please update us the progress? The second question is related with competition. We saw some fundraising movement from your peers in the capital market, private capital market. Could you please update us if you have any more intensified competition while pitching customers or penetrating customers? The third question is related with the gross margin outlook. As Jingbo mentioned that one Chinese telco raised the price. Do you see any risk of the other two will also raise the price in future? What will be the gross margin trend in the next few quarters?
Thank you.
Okay. I'll talk about the progress on use cases, especially for the larger classroom experience. It's true that we are actively working with customers on what we call super small class use cases because this is one step further in enhancing the user, student, and teacher experience in large classroom use cases. The use case is like this. Students can take a class with hundreds or thousands or even tens of thousands of other students in one same classroom so that there could be a lead teacher teach them together. During the classroom experience maybe every 10- 15 minutes, the students will be formed small groups of four to six in each small group, and they can work on assignment and discuss with each other or even like a formal competition within the small group to compete on who's going to win on answering certain questions.
After that, come back to the large classroom session. This new format combines the efficiency of large classes and the interactivity of small classes. The competition could launch within the small classroom will further enhance the students' attention or focus in the content being taught in that classroom session, which can significantly improve the learning experience and the outcome. There are several other interesting use cases like in virtual workplace and also in some gaming and social use case as well. Outside of education, we are also working on some other interesting use cases as well. Now, about the competition, it's true that I think this year maybe due to both COVID-19 caused demand and of course our build in the public market.
There are more companies looking at the space and there are fundraisings for different small companies or new product rollouts in this space. We think those indicates the potential and opportunities in RTE PaaS industry are recognized by more and more people today. As we point out early, those developments, I think, also shows the industry are gradually coming to a consensus that this is not old stuff similar to video web conference or other product offered by traditional company. Rather, it's a whole new industry that focus on real-time engagement. The concept of real-time engagement is more and more being adopted by all industry peers instead of the old concept, RTC. We're happy to see our early predictions and judgment have proved to be right. Meanwhile, with the newcomers and the competition, we also think our competitive position quite clear and strong.
Like in technical edge, RTE involve a lot of technical challenges from performance to scale to features. Each use case have different requirements in order to support. As I mentioned, our platform actually support hundreds or thousands of different use cases. It also requires different priorities among various aspects of the solution. Working with our customers, we have developed a solution for more than 100 use cases with certain volume already, which has been used in millions of end users. Our data analytics and XLA assurance also unique in the industry today. PR for Mindshare side, we also by far have the largest developer community for RTE. It's nearly 10,000 new apps in just every month. We support by far the widest range of platforms and development frameworks.
I'll take the third question on margin. As to the telcos in the price, we do not expect that to happen with the other two telcos, given the one that did the price this quarter unfortunately had the lowest price amongst them, and the effort was really to close the pricing gap. After the recent adjustments, the prices are more similar in these main telcos, we do not expect any short-term changes. As to the general trend on gross margin, as I mentioned earlier, it was affected by both the international expansion and also the incoming costs. If we look slightly longer term, I think we can enhance our gross margin in three ways. Number one is always innovation. To create more innovative products and features. That's always the pathway to create more value for customers and generate higher margin.
That's why we have been investing R&D very heavily. Secondly, XLA would be a great way to help us charge different prices based on different level of quality difference or use cases where the highest possible quality needed, we'll be able to achieve higher gross margins over time. Lastly, we'll continue to reduce bandwidth and server costs through optimization and better reuse of resources. We are actually long-term optimistic about gross margin. In the near term, because all these efforts will take time to be reflected in the accounts, so we do expect gross margin to remain under pressure, but unlikely to decline significantly from what we saw in Q2.
Thank you.
Thank you so much. Once again, ladies and gentlemen, it's star and one if you wish to ask a question. Our next question comes from the line of Emerson Chan from Bank of America. Your line is now open.
Hi. Thank you, management. I have three questions. My first question is regarding the RTE adoption and entry barrier. What RTE use case we found the most challenging to serve in terms of technology, and how should we look at the technological barrier of RTE in the future? Is it going to be higher, let's say, due to more advanced use case emerging or the entry barriers of RTE is actually getting lower? Secondly, our revenue in Q3 grow 80% year-over-year, and we maintain our full year guidance, meaning Q4 revenue to be up about 40% at least. Can you give us more color on the reason behind the potential slowdown in Q4 and what we have seen so far in Q4?
My last question is regarding the COVID-19 impact. I just want to get a rough sense of how much COVID still contributes to our revenue growth in Q3. I suppose it should be coming from overseas market. How much percentage of revenue will still be coming from overseas? Thank you.
Sure. I'll take the RTE use case question. Actually, in general, I think RTE faces many challenges in almost all use cases. We actually think there are many challenges in the three pillars I was talking about just now in opening remarks, like shared context. How do you support all the different context or environment or background use case with how to create? Also, any scale in the interaction, from very small ones to very large ones. Lastly, about ubiquity. Like we need to continue to support all tools and frameworks. Sometimes developers do not really need all of those framework of tools, but they want to have the flexibility and the availability of the support for that platform.
When they choose those tools or platforms to build their RTE use cases, they would consider the ones that have the most support for almost all tools they want, they might potentially use. AR/VR is definitely a hard one, because they definitely demand more computational power, more transmission bandwidth, et cetera. There is, even like in the large classroom experience I was just talking about, we need to support tens of thousands concurrent students, and they can break out in matter of seconds to thousands of small groups and then form meaningful small group discussion or competition within the group and come back in another second.
You can organize all that in a programmable way. All those have some challenges in how we design the technical infrastructure, because you have to imagine that the student and teachers could come from all over the world, and all use different equipment or devices to join the session. They all deserve to have the same high-level quality assurance.
Okay. On the second question about revenue guidance, I would say we are certainly not suggesting a quarter-on-quarter slowdown in Q4. As always, we don't want to be aggressive in guiding the market, and we don't feel it's necessary to adjust the year guidance frequently unless there is a sudden change in expectations. As for Q4, so far, I would say we are encouraged by what we have seen so far in Q4, and hopefully we will be able to deliver another strong quarter. As to the revenue mix and impact from COVID. That's right. In Q3, I would say there was almost zero impact from COVID in China. Outside China, revenue contribution at the time was more than 20%, probably more than 20% in Q3.
Outside of China?
Yes, outside of China.
Yeah.
I wouldn't say that conversion was all due to COVID. Actually, the majority of that conversion was due to our own market expansion efforts there. There's more portion there simply due to the short-term demand from COVID. Can you still hear us?
Okay. It seems that he disconnected his line. Thank you. Yeah, we'll move on to our next question. Comes from the line of Rich Valera from Needham. Your line is now open.
Thank you, and thank you for the opportunity to ask some questions here. First, just wanted to clarify the comments on Q4. I apologize, but the line is a little difficult to hear. Jingbo, did you say that you did not expect a quarter-over-quarter decline in Q4? I just wanted to make sure I understood what your comments were with respect to the Q4 outlook.
Sure. Obviously, Q4 is still only halfway through. It's still too early to say. I guess what I wanted to clarify, we are not guiding a QoQ slowdown in Q4.
Okay. That's perfect. You've had a couple of quarters of really strong new customer growth. I think probably two of the strongest quarters in your history in Q2 and Q3. I'm wondering if you can talk about when you expect to see them starting to generate revenue and, maybe if you've seen any early signs of those added in Q2 starting to generate revenue, and just how we should think about those contributing to future revenue growth.
Sure. In general, new customers will take three to six months to fully ramp up. This has been actually same this year. The difference this year is the fluctuation in the first three to six months is much, much bigger due to COVID. Imagine an education customer who joins the platform in January or February, they will suddenly have very high usage despite all the technical difficulty initially. Given the situation back then, they had to go with it whatever the situation. You have very high initial usage. Then as China reopened in mid-May, early June, the usage would fall. That's why it's a lot harder to tell from the data we've seen so far. Generally, we feel that the trend is the same. These customers will continue to generate revenue, but just the short-term pattern was very heavily distorted.
Understood. Final question from me, in terms of customer concentration, I think last quarter you didn't actually have any 10% customers, but I believe you had two customers that were 9% of revenue. I was wondering if you could comment on your largest customers this quarter, how big they were, and maybe the status of those two 9% customers from last quarter. Thank you.
Sure. This quarter, we had one customer slightly above 10%, top 10 in total was 37%. The 10% customer was one of the two 9% customers last quarter. The other customer was smaller. As I explained, for some of the education use cases, especially for public schools, obviously after the reopening, the usage would fall back.
Got it. That's very helpful. Thank you for taking the questions.
Thank you so much. Once again, ladies and gentlemen, if you wish to ask a question, it's star and one on your telephone. Again, it's star and one if you wish to ask a question. Again, ladies and gentlemen, it's star and one if you would like to ask a question. There are no further questions at this time. Speakers, you may continue.
Thank you, operator. If any investors or research analysts have any questions, please feel free to reach out to us through our IR email or our IR website. Today's webcast, and also the presentation, is available on our IR website as well. Thank you, everybody.
Thank you.
Thank you.
Thank you. That does conclude the conference for today. Thank you for participating. You may all now disconnect.