Applovin Corp (APP) 2025 Q4 法說會逐字稿

(technical difficulty) our GAAP results. Please be sure to review the GAAP results and the reconciliations of our GAAP and non-GAAP financial measures in our earnings release and financial update available on our Investor Relations site. This conference call is being recorded, and a replay will be available for a period of time on our IR website.

（技術難題）我們的 GAAP 結果。請務必查看我們投資者關係網站上發布的盈利報告和財務更新中的 GAAP 結果以及 GAAP 和非 GAAP 財務指標的調整表。本次電話會議正在錄音，錄音回放將在我們投資者關係網站上保留一段時間。

Now, I'll turn it over to Adam and Matt for some opening remarks, then we'll have the moderator take us through Q&A.

現在，我將把發言權交給亞當和馬特，請他們做開場白，然後由主持人帶領我們進入問答環節。

Thanks, everyone, for joining us today. I want to start by addressing what's clearly on many people's minds. While I prefer to ignore short-term fluctuations in the stock price and focus on maximizing value over the long term, the recent volatility warrants addressing.

感謝各位今天蒞臨。我想先談談很多人都關心的問題。雖然我更傾向於忽略股價的短期波動，專注於長期價值最大化，但最近的波動性值得關注。

For the past few weeks, there's been a lot of discussion about how AI and competition will challenge our business. But when I look at our internal dashboards, we are delivering the strongest operating performance in our history. What's fueling that growth is our own AI models. And as research and AI, both internal and external, continues to improve, our business will grow with it. There is a real disconnect between market sentiment and the reality of our business.

過去幾週，關於人工智慧和競爭將如何挑戰我們業務的討論很多。但當我查看我們的內部儀表板時，我們發現我們正處於公司歷史上最強勁的營運業績時期。推動這一成長的是我們自己的人工智慧模型。隨著內部和外部研究及人工智慧技術的不斷進步，我們的業務也將隨之成長。市場情緒與我們業務的實際情況之間存在著嚴重的脫節。

Before I talk about the opportunity ahead, let me explain how we think about competition in AI. First, on competition. Since day one, we've competed against many companies. We've never feared competition because it forces us to innovate to serve our gaming ecosystem even better. We operate a foundational piece of the ecosystem, the MAX auction. It's critical for the ecosystem that the MAX auction improves through more competition, which in turn helps publishers make more money, leading to more user acquisition.

在談論未來的機會之前，讓我先解釋一下我們如何看待人工智慧領域的競爭。首先，談談競爭。從一開始，我們就與許多公司競爭。我們從未懼怕競爭，因為競爭會迫使我們創新，以便更好地服務我們的遊戲生態系統。我們經營生態系統的基礎部分——MAX 拍賣。對於整個生態系統而言，透過增加競爭來改進 MAX 拍賣至關重要，這反過來又可以幫助發行商賺更多錢，從而獲得更多用戶。

Now in a typical zero-sum auction-based market, if one improves another loses. In our case, as bid density goes up, the pie expands. And while our share of the auction may shrink, our economics actually grow. There are impressions our model understands extremely well and it values highly and others where we have less signal and value them less. When competition wins an impression, it's very likely to be the one that we value less. This leads to the publisher making more. And in many cases, we do as well because instead of winning a low-value impression, we get to charge the winning bidder 5%.

在典型的零和拍賣市場中，如果一方獲利，另一方就會虧損。就我們而言，隨著競價密度的增加，蛋糕也會擴大。雖然我們在拍賣中的份額可能會減少，但我們的經濟實際上會成長。有些印像我們的模型理解得非常透徹，並給予高度評價；而有些印像我們感知到的訊號較少，因此評價也較低。當競爭對手贏得曝光機會時，它很可能是我們不太重視的那一個。這會讓出版商賺得更多。而且在很多情況下，我們也能做到這一點，因為我們不是贏得低價值的曝光機會，而是可以向得標者收取 5% 的費用。

When you hear about a start-up coming for our business, you should be asking how their value proposition can be stronger than ours. If our value proposition wasn't strong with our partners, we would have lost those relationships long ago. We have competed very well in mediation against giant companies. The network effects in this business are very real. We scaled our network by providing the best monetization and even more importantly, the best advertising tools to publishers. The combination is not something peers can overcome.

當聽到有新創公司要挑戰我們的業務時，你應該問他們如何才能比我們更有價值。如果我們對合作夥伴的價值主張不夠強，我們早就失去這些關係了。我們在與大型企業的調解中表現出色。這個產業的網路效應是真實存在的。我們透過向發布商提供最佳的獲利模式，更重要的是，提供最佳的廣告工具，擴大了我們的網路規模。這種組合是同儕無法克服的。

Second, let's talk about AI and game creation. The bearish view assumes that if AI makes games easier to build, the value of our ecosystem declines. Well, we believe the exact opposite. AI will dramatically lower the cost of creation, which means content will explode. And when content becomes abundant, discovery becomes a scarce resource. Even in the past, as mobile games were built by human teams, the content was plentiful and, in many cases, commoditized.

其次，我們來談談人工智慧和遊戲創作。悲觀觀點認為，如果人工智慧使遊戲開發變得更容易，我們生態系統的價值就會下降。我們的觀點恰恰相反。人工智慧將大幅降低創作成本，這意味著內容將爆炸性成長。當內容變得過剩時，發現就成了稀缺資源。即使在過去，當手機遊戲是由人類團隊開發時，內容也十分豐富，而且在許多情況下都已商品化。

That is actually what allowed us to deliver such a strong value proposition through our advertising solutions. In a world where anyone can create an app or a game, millions of experiences will compete for attention. The winners will be the platforms that can efficiently match the right user to the right content at the right moment. That is exactly what our models are designed to do. We are not tied to any specific genre or format, casual, mid-core or beyond. Our systems follow engagement and AI only increases the potential of that capability.

正是這一點讓我們能夠透過廣告解決方案提供如此強大的價值主張。在任何人都可以創建應用程式或遊戲的世界裡，數百萬種體驗將爭奪用戶的注意力。能夠有效率地在適當的時機將合適的用戶與合適的內容相符的平台將成為最終的贏家。這正是我們模型的設計初衷。我們不局限於任何特定的類型或形式，無論是休閒、中核或其他類型。我們的系統會追蹤使用者互動，而人工智慧只會增強這種能力的潛力。

Furthermore, we don't see any evidence of a declining mobile gaming. Casual gaming serves a different human need than console, PC gaming, AAA games or any other form of deeply immersive game experience. People will always look for entertainment that fits naturally into their day. What's changing is how well that attention can be monetized. Even today, we convert only a small portion of those impressions that we serve. We view that not as a limitation, but as a large long-term opportunity as our models continue to improve.

此外，我們沒有看到任何行動遊戲市場衰落的跡象。休閒遊戲滿足的人類需求與主機遊戲、PC遊戲、3A遊戲或任何其他形式的深度沉浸式遊戲體驗都不同。人們總是會尋找能夠自然融入日常生活的娛樂方式。正在改變的是，這種關注度能夠以何種方式變現。即使在今天，我們也只能將投放的廣告展示中的一小部分轉化為實際轉換。我們認為這不是一種限制，而是一個巨大的長期機遇，因為我們的模型將不斷改進。

Now that brings me to what we control, performance and culture. Our performance, our business is executing extremely well. We continue to grow very quickly despite the numbers getting much bigger. We delivered strong growth in Q4. And despite typical seasonality where Q1 should be softer than Q4, we are guiding to meaningful sequential growth. That reflects both continued strength in gaming and the scaling of our e-commerce and our self-service customers.

現在，我想談談我們能夠控制的方面：業績和文化。我們的業績和業務執行都非常出色。儘管人數越來越多，但我們仍然保持著非常快速的成長。我們在第四季度實現了強勁成長。儘管通常情況下，第一季會比第四季疲軟，但我們預計第一季將實現顯著的環比成長。這不僅反映了我們在遊戲領域的持續強勁表現，也反映了我們電子商務和自助服務客戶的規模不斷擴大。

On culture, we embrace being underestimated. A skeptical market sharpens our focus and pushes our teams to execute. Our revenue per employee remains among the highest in the world because we build the best and most scalable products in our category. If the market chooses to price our stock based on fear, while we continue to compound revenue, cash flow and product capability, we'll stay focused on execution and let our results speak over time. From where we sit, we are still in the early innings of what this platform can be.

在文化方面，我們欣然接受被低估。充滿懷疑的市場環境會讓我們更加專注，並促使我們的團隊更加努力地執行任務。我們之所以能維持全球最高的人均收入，是因為我們打造了同類產品中最好、最具可擴展性的產品。如果市場選擇基於恐懼來為我們的股票定價，而我們將繼續提升收入、現金流量和產品能力，我們將專注於執行，讓業績隨著時間的推移來證明一切。從我們目前的角度來看，這個平台的發展還處於早期階段。

With that, I'll turn it over to Matt to walk through the financials.

接下來，我將把財務部分交給馬特來講解。

Thanks, Adam. Q4 marked what was not just a strong quarter, but the most exceptional year we've ever delivered and one of the strongest performances in the public markets. At our scale, the combination of growth, profitability, free cash flow and capital returns we are delivering is extraordinarily rare. Revenue in the fourth quarter was $1.66 billion, up 66% year-over-year, driven by continued technology advancements to our core mobile gaming business, seasonal strength and the expanding impact of our e-commerce initiative.

謝謝你，亞當。第四季不僅標誌著一個強勁的季度，而且標誌著我們有史以來最卓越的一年，也是我們在公開市場上表現最強勁的一年之一。以我們目前的規模而言，我們所實現的成長、獲利能力、自由現金流和資本回報的結合是極為罕見的。第四季營收為 16.6 億美元，年成長 66%，這主要得益於我們核心行動遊戲業務的持續技術進步、季節性強勁成長以及電子商務計畫的影響力不斷擴大。

Adjusted EBITDA was $1.4 billion, up 82% year-over-year, representing an 84% margin. Margins expanded over 700 basis points from the same period last year and quarter-over-quarter flow-through to adjusted EBITDA was approximately 95%, again, demonstrating our relentless dedication to execution and how efficiently incremental revenue converts into earnings for our business.

調整後 EBITDA 為 14 億美元，年增 82%，利潤率為 84%。與去年同期相比，利潤率成長超過 700 個基點，季度環比調整後 EBITDA 的轉換率約為 95%，再次證明了我們對執行的不懈追求，以及增量收入如何高效地轉化為我們業務的收益。

Investors often reference the Rule of 40 in software. On that basis, our 66% revenue growth and 84% adjusted EBITDA margins translate to a score of 150. That level of profitability at this growth rate is almost unheard of and reflects the fundamental operating leverage of our model.

投資人經常在軟體領域提及「40法則」。基於此，我們 66% 的營收成長和 84% 的調整後 EBITDA 利潤率相當於 150 分。在這樣的成長速度下達到這樣的獲利水準幾乎是聞所未聞的，這反映了我們模式的基本營運槓桿作用。

Free cash flow for the quarter was $1.31 billion, an 88% increase year-over-year, growing our cash balance to $2.5 billion and reinforcing the strength of our balance sheet. This was a truly remarkable year for AppLovin. Revenue reached $5.48 billion, growing 70% year-over-year. Adjusted EBITDA was $4.51 billion, up 87% year-over-year at an 82% adjusted EBITDA margin, a margin profile that very few companies ever achieve, let alone sustain at the scale. Free cash flow totaled $3.95 billion, up 91% year-over-year, underscoring not just growth, but the exceptional quality and durability of our earnings.

本季自由現金流為 13.1 億美元，年增 88%，使我們的現金餘額增加到 25 億美元，並鞏固了我們資產負債表的實力。對 AppLovin 來說，這真是個意義非凡的一年。營收達 54.8 億美元，年增 70%。調整後 EBITDA 為 45.1 億美元，年增 87%，調整後 EBITDA 利潤率為 82%，這種利潤率水準很少有公司能夠達到，更不用說在這種規模上維持了。自由現金流總計 39.5 億美元，年增 91%，這不僅凸顯了成長，也凸顯了我們獲利的卓越品質和持久性。

Simply put, very few public companies are scaling faster, more profitably and with greater cash generation than we are today. That strength directly translates into shareholder returns. During the quarter, we repurchased and withheld approximately 800,000 shares for $482 million. For the full year, we repurchased and withheld approximately 6.4 million shares for a total of $2.58 billion, funded entirely by free cash flow.

簡而言之，如今很少上市公司能夠比我們更快擴張、更有利潤、產生更多現金流。這種實力可以直接轉化為股東回報。本季度，我們回購並持有約 80 萬股股票，價值 4.82 億美元。全年共回購並持有約 640 萬股股票，總額達 25.8 億美元，全部資金來自自由現金流。

As of the end of the year, we had a remaining share repurchase authorization of approximately $3.28 billion. Over the last four quarters, we reduced our weighted average diluted shares outstanding from 346 million to approximately 340 million, while simultaneously investing in organic growth and maintaining substantial liquidity. Our share repurchase program reflects our conviction in the value and durability of the business.

截至年底，我們剩餘的股票回購授權額度約為 32.8 億美元。在過去的四個季度裡，我們將加權平均稀釋後流通股從 3.46 億股減少到約 3.4 億股，同時投資內生成長並維持充足的流動性。我們的股票回購計劃體現了我們對公司價值和永續發展的信心。

Turning to our outlook for the first quarter of 2026. We expect revenue between $1.745 billion and $1.775 billion, representing 5% to 7% sequential growth. Adjusted EBITDA is expected to be between $1.465 billion and $1.495 billion with an adjusted EBITDA margin of approximately 84%, maintaining best-in-class profitability as we continue to scale.

接下來展望一下我們對 2026 年第一季的展望。我們預計營收將在 17.45 億美元至 17.75 億美元之間，較上季成長 5% 至 7%。調整後 EBITDA 預計在 14.65 億美元至 14.95 億美元之間，調整後 EBITDA 利潤率約為 84%，在我們繼續擴大規模的同時，保持一流的盈利能力。

To close, AppLovin represents a combination that is exceedingly rare, sustained hyper growth, exceptional margins, massive free cash flow generation and disciplined capital returns. We believe this puts us in a category of our own and positions us to continue delivering outsized value for shareholders over the long term.

總而言之，AppLovin 代表了一個極為罕見的組合：持續的超高速成長、卓越的利潤率、巨額自由現金流和嚴謹的資本回報。我們相信這使我們獨樹一幟，並使我們能夠在長期內繼續為股東創造超額價值。

Now with that, let's move to Q&A.

接下來，我們進入問答環節。

(Operator Instructions) Benjamin Black, Deutsche Bank.

（操作員指示）本傑明·布萊克，德意志銀行。

So my first question is on the e-commerce opportunity. So could you perhaps sort of reflect on the self-service launch? What were some of the key learnings, what worked? What didn't? Where is there room for improvement? And to the extent possible, it would be great if you could sort of share or quantify the e-commerce contribution to revenue or gross ad spend this quarter and in the guide? And then I have a quick follow-up.

所以我的第一個問題是關於電子商務的機會。那麼，您能否談談自助服務推出的情況呢？有哪些關鍵經驗教訓？哪些方法行之有效？什麼沒有？還有哪些方面需要改進？如果可能的話，您能否在本季和指南中分享或量化電子商務對收入或廣告總支出的貢獻？然後我還有一個簡短的後續問題。

Yeah. So the e-commerce business obviously has been live with us for 1.5 years. It's doing really well. In Q4, we opened up the self-service platform, referral only. So we're not at the point yet where we're sort of a GA type launch. We'll get there. We said first half of this year, that's still on track.

是的。所以，我們的電子商務業務顯然已經上線一年半了。它表現得非常出色。第四季度，我們開放了自助服務平台，僅限推薦用戶使用。所以，我們目前還沒有到正式發布階段。我們會到達那裡的。我們之前說過，今年上半年的計畫仍然在按計畫進行。

What gets us excited are a couple of things. One is the current customers that had lapped Q4 2024 into Q4 2025 saw material increases in spend as our models just keep getting better. Now remember, this business and the model around this business that drives the value for the advertisers is really in its infancy. It takes us a while to continuously iterate to improve the model. In fact, just a few weeks ago, we had a pretty sizable uplift.

有兩件事讓我們感到興奮。其中之一是那些已經將 2024 年第四季過渡到 2025 年第四季的現有客戶，他們的支出大幅增加，因為我們的模型一直在不斷改進。請記住，這項業務以及圍繞這項業務為廣告商創造價值的商業模式，實際上還處於起步階段。我們需要一段時間不斷迭代改進模型。事實上，就在幾週前，我們經歷了相當大的成長。

So those same customers from the prior year cohort saw big growth. Then you ended up with new customers coming in from the referral program. I mentioned on the last earnings call, we were seeing substantial growth there. We're not going to see anything that's going to impact our overall numbers for a while, but we're seeing great trends. And I'll talk about advertising, leading advertisers into our platform later on the call, but we're just seeing numbers that get us excited.

因此，與前一年相同的客戶群實現了大幅成長。然後，你就透過推薦計畫獲得了新客戶。我在上次財報電話會議上提到過，我們那方面已經實現了顯著成長。短期內不會出現任何影響我們整體數據的事件，但我們看到了良好的發展趨勢。稍後我會在電話會議上談到廣告，以及如何引導廣告商進入我們的平台，但我們現在看到的一些數字讓我們感到興奮。

On the breaking out e-commerce, we're not going to do that because we think of our platform as a unified platform. Let's say, tomorrow, the engineering team improves the gaming model 50%. Well, e-commerce would go down, but the business would be ripping. That wouldn't mean that there's anything wrong with our platform.

關於電子商務的拆分，我們不會這樣做，因為我們認為我們的平台是一個統一的平台。假設明天工程團隊將遊戲模型改進了 50%。電子商務可能會下滑，但企業本身會蓬勃發展。但這並不意味著我們的平台有什麼問題。

Fundamentally, we're one single auction, but getting more diversity will give the model more ways to serve the end consumer and should drive up our overall conversion rate. But we think if we start talking about verticals in a marketplace like ours, you start getting really misleading information that will throw investors off.

從根本上講，我們採用的是單一的拍賣模式，但增加多樣性將為該模式提供更多服務終端消費者的方式，並應能提高我們的整體轉換率。但我們認為，如果我們開始談論像我們這樣的市場中的垂直領域，就會開始出現一些誤導性的信息，從而誤導投資者。

Great. I guess. And then sort of my second question is sort of related to a feature set that you're rolling out. So I think in the past, you mentioned the conversion uplift your partners could see if they have the ability to not only optimize for the best performing creators but also sort of really scale and automate the creation of these video assets. So, I guess, with that backdrop, how far along the automation curve are we now and where could be in the next, call it, 12 to 18 months?

偉大的。我猜也是。我的第二個問題與你們正在推出的一組功能有關。所以我覺得你之前提到過，如果合作夥伴不僅能夠優化表現最佳的創作者，而且還能真正規模化和自動化這些影片素材的創建，那麼轉換率的提升就顯而易見了。所以，我想，在這樣的背景下，我們現在處於自動化曲線的哪個階段，以及在接下來的 12 到 18 個月內，我們可能會處於什麼階段？

Yeah. Great question. We're still pretty early. I pulled some numbers earlier just to compare where e-commerce companies are when they upload ads and how many they upload to our platform versus the gaming companies that have really had a decade plus to optimize for our platform. Top gaming companies run tens of thousands of ads at any given time. The top e-commerce companies are in the hundreds.

是的。問得好。現在還為時過早。我之前提取了一些數據，只是為了比較電子商務公司在我們平台上傳廣告時所處的階段以及他們上傳到我們平台的廣告數量，與遊戲公司相比，後者已經有十多年的時間來優化我們的平台。頂級遊戲公司在任何時候都會投放數萬則廣告。頂尖的電子商務公司有數百家之多。

So you've got a huge discrepancy here. And if you throw more ads in our system, the model does much better. That's just a fact. It gets the chance to diversify what the end user sees and try to find something that's going to eventually convert that user.

所以這裡存在著巨大的差異。如果在我們的系統中投放更多廣告，該模型的表現會更好。這是事實。它有機會豐富最終用戶看到的內容，並嘗試找到最終能夠轉換該用戶的內容。

Now how do we bridge the gap? It's twofold. One is the customers have to get more accustomed to our platform to know what types of creatives work so that they can build up production around the things that work and multiply out the count.

那麼我們該如何彌補這一差距呢？有兩方面原因。一方面，客戶需要更多地了解我們的平台，才能知道哪些類型的創意有效，以便圍繞有效的創意進行創作，並擴大作品數量。

More importantly, though, generative AI tools to build creatives in a really low-cost way in an automated way is on the way. We already have in a pilot with over 100 customers, generative AI-based tools for one part of the ad unit. Our ads are a video plus then an interactive page and then follow-up of a shop preview dynamic product page.

更重要的是，能夠以極低的成本和自動化的方式建立創意內容的生成式人工智慧工具即將問世。我們已經在超過 100 位客戶中開展試點項目，將基於生成式人工智慧的工具應用於廣告單元的某個部分。我們的廣告形式為：一段視頻，然後是一個互動頁面，最後是一個店舖預覽動態產品頁面。

But that middle one, the interactive page is not something that these advertisers are accustomed to building because they don't need them on social or search. We're now generating those automatically for over 100 customers. We'll roll that out soon as it's showing good performance to the broader set of customers. And shortly, we're going to have the video model go live as well. We're going to run the same pilot process. We're going to make sure the video output looks good.

但中間那個互動頁面，是這些廣告商不習慣建構的，因為他們在社群媒體或搜尋上不需要它們。我們現在為超過 100 位客戶自動產生這些文件。鑑於該功能在更廣泛的客戶群中表現出良好的性能，我們將盡快推出。不久之後，我們的視訊模式也將上線。我們將運行相同的試點流程。我們要確保視訊輸出效果良好。

But if we get to a place where the video model can help these customers create new video ads in bulk in cost of dollars versus cost of thousands of dollars, we expect the count of ads for these new customers on our platform is going to go up a lot, and that will make them more competitive against the gaming customers that are on our platform.

但是，如果我們能夠讓視訊模型幫助這些客戶以幾美元的成本（而不是數千美元的成本）大量創建新的視訊廣告，我們預計這些新客戶在我們平台上的廣告數量將會大幅增加，這將使他們與我們平台上的遊戲客戶更具競爭力。

Jason Bazinet, Citi.

Jason Bazinet，花旗集團。

How are you guys doing? I think one of the things that investors have always struggled with a bit is just sort of what they would call the black box nature of the model, like they can't get comfortable sort of doing a P times Q and sort of extrapolating the model out there. And I know you're not going to break out, as you said, e-commerce from mobile. But as investors are sort of looking at these e-commerce accounts that have your pixel on it, what counsel would you provide to the buy side in terms of using those numbers and trying to do some math to get some sense of how well you're doing?

你們最近怎麼樣？我認為投資人一直以來都有些難以接受的一點，就是他們所謂的模型的「黑箱」性質，例如他們無法放心地進行 P 乘以 Q 的計算，然後把模型的結果外推到其他地方。我知道你不會像你所說的那樣，從行動端剝離電子商務。但是，當投資人開始關注這些安裝了你們像素的電子商務帳戶時，你們會向買方提供哪些建議，以便他們能夠利用這些數據進行一些計算，從而了解你們的業績如何？

Yeah. I mean, look, it's really, really hard because we're just getting started. Facebook's probably pixeled over 10 million sites. You've been publishing our pixels in the thousands. So it does correlate to the count of advertisers that we have on the platform, but we're just starting out.

是的。我的意思是，你看，這真的非常非常難，因為我們才剛起步。Facebook可能已經對超過1000萬個網站進行了像素檢測。你們已經發布了成千上萬個我們的像素。所以它確實與我們平台上的廣告商數量有關，但我們才剛起步。

We can't build a P times Q model in a period of time that we expect to have really outsized growth. Once we open up the platform, start running marketing to the platform, start doing more sales, you're going to end up having a ton of advertisers coming in. We can't predict it. It's not going to be stable until we get to a much further into the future point. But what gets us excited, like I mentioned on the last answer is we're already testing in advertising to get customers on the platform.

我們無法在預期會出現超高速成長的時期內建構 P 乘以 Q 的模型。一旦我們開放平台，開始在平台上進行行銷，開始增加銷售額，最終將會有大量的廣告商湧入。我們無法預測。這種情況要到更遙遠的未來才會穩定下來。但正如我在上一個回答中提到的，真正讓我們興奮的是，我們已經在廣告方面進行測試，以吸引客戶加入該平台。

So what you're seeing right now is just referral-based advertiser count. But if we start actually being able to market our platform and get customers to convert through the funnel, that's going to help us really catalyze faster growth to go get advertisers in the absence of a sales team. And one stat that's really cool that I pulled earlier, we're testing in light volume on ads on social and search, and then we're doing referral programs. But right now, we're seeing somewhere around day 30 LTV to cost of user acquisition.

所以你現在看到的只是基於推薦流量的廣告商數量。但是，如果我們能夠真正開始推廣我們的平台，並讓客戶透過銷售漏斗完成轉化，這將真正幫助我們在沒有銷售團隊的情況下更快地吸引廣告商，從而加速成長。我之前看到的一個數據非常酷，我們正在小規模測試社群媒體和搜尋廣告，然後我們也在進行推薦計畫。但目前我們看到的用戶生命週期價值與用戶獲取成本之比約為第 30 天左右。

So if you think about lead gen models and if you know lead gen models and also if you understand the life of value that we create for advertisers, which is in the many years, to be able to break even on the media buy in 30 days is exceptional. We've got arguably one of the best business models the world has ever seen, and we're seeing the ability to market our platform and small testing at that level. That gets us excited.

所以，如果你考慮潛在客戶開發模式，如果你了解潛在客戶開發模式，並且如果你明白我們為廣告商創造的價值生命週期是多年的，那麼能夠在 30 天內收回媒體購買成本是非常難得的。我們擁有可以說是世界上最好的商業模式之一，我們看到了在這個層面上推廣我們的平台和小規模測試的能力。這讓我們很興奮。

Now we're not ready to go to GA yet, and you can ask why. And the why is because we still need to optimize our conversion funnel. We're no different than any social network that's going out and trying to get users. We want to get users into this advertising platform. Right now, we're seeing of qualified leads, 57% of advertisers go live. That's 43% breakage. We think we can get that number much closer to 100% before we open up. And despite that, we're seeing this 30-day breakeven roughly on LTV to CAC.

現在我們還沒準備好去GA，你可能會問為什麼。原因在於我們仍然需要優化轉換漏斗。我們和其他任何努力獲取用戶的社交網路並無不同。我們希望吸引用戶使用這個廣告平台。目前，我們看到，在合格的潛在客戶中，57% 的廣告主會選擇上線投放廣告。破損率高達43%。我們認為在開放之前，我們可以將這個數字提高到接近 100%。儘管如此，我們看到 LTV 與 CAC 的 30 天損益平衡點大致相同。

So we're really excited about where we are. We think we're going to go through this hyper growth phase as we open up and really start pouring advertisers in. Once we get to a stable place where we could start predicting count of advertisers in quarters to come, then we'll give you the P times.

所以我們對目前所取得的成就感到非常興奮。我們認為，隨著我們逐步開放並真正開始吸引大量廣告商，我們將經歷一個高速成長階段。一旦我們達到穩定的狀態，可以開始預測未來幾季的廣告商數量，我們就會告訴你 P 時間。

Omar Dessouky, Bank of America.

奧馬爾·德蘇基，美國銀行。

I wanted to ask a more general question because the market seems to be having an AI moment here. So let's just forward three to five years in the future to potentially a world where consumers interface with the Internet through natural language-based agents. LLMs are different than performance advertising neural networks. But how would your business be affected by potentially a change in how consumers interface with the Internet. So that's the first part of it.

我想問一個更普遍的問題，因為市場似乎正在經歷人工智慧時代。那麼，讓我們展望未來三到五年，看看在那個時代，消費者可能會透過基於自然語言的代理商與網路互動。LLM 與效果廣告神經網路不同。但是，如果消費者與網路的互動方式發生潛在變化，您的業務會受到怎樣的影響？這是第一部分。

And then the second is, how do you expect -- again, over the long term, three to five years, how would you expect mobile game developers to casual mobile game developers to adapt their content to a world where they compete for time with these sophisticated chatbots?

第二個問題是，從長遠來看，三到五年內，你認為行動遊戲開發者，尤其是休閒行動遊戲開發者，應該如何調整他們的內容，以適應與這些複雜的聊天機器人爭奪用戶時間的世界？

Yeah. So great question. The way I see the LLM is that it's going to enable anyone to be able to type out ideas for a game and get a game created. That doesn't mean you're going to get users to play your game, but that means you can create content in a very low-cost way, not being an engineer. That will accelerate content production for studios that are developed, that are sophisticated.

是的。問得好。我認為LLM（遊戲法碩士）將使任何人都能寫下遊戲創意並製作出遊戲。這並不意味著你就能吸引用戶玩你的遊戲，但這確實意味著你可以以非常低成本的方式創建內容，而無需成為工程師。這將加快成熟、成熟的工作室的內容製作速度。

It will also create a flood of content that's more personalized across people who could just become game developers with no engineering background. That's a good thing for us. In the talk track I mentioned, as you get more content, that content becomes commoditized, the discovery platforms, of which there are very few become the true value because those customers are going to have to come to MAX to monetize and they're going to have to come to our discovery platform to run advertising to get users to their space.

它也將催生大量更個人化的內容，讓那些沒有工程背景的人也能成為遊戲開發者。這對我們來說是件好事。在我提到的對話中，隨著內容的增多，內容會變得商品化，而發現平台（目前數量很少）則成為真正的價值所在，因為這些客戶必須來到 MAX 才能實現盈利，他們必須來到我們的發現平台投放廣告才能吸引用戶訪問他們的空間。

Now if humans just want to talk to a chatbot, do nothing else. Of course, while there's nothing else that they're going to do, we would lose time spent in games. that seems far-fetched. The games that people play today in our domain are not very immersive games. They're quick-to-play relaxing games and people play things like a Solitaire, people play things like a Mahjong. This audience skews older, it skews female. This is an audience that's very unlikely to stop playing crosswords in 10 years because they're talking to a chatbot.

如果人類只是想和聊天機器人對話，那就什麼都不用做。當然，如果他們沒有其他計劃，我們就會失去玩遊戲的時間。這似乎有點牽強。人們現在在我們這個領域玩的遊戲，沉浸感並不強。這些都是節奏快、輕鬆休閒的遊戲，人們會玩像紙牌接龍這樣的遊戲，也會玩像麻將這樣的遊戲。這個受眾群體年齡偏大，女性比例較高。這群人不太可能因為在和聊天機器人聊天就在 10 年後停止玩填字遊戲。

So not only do we think that the base audience is going to keep playing games, you're going to need relaxing outlets. If LLMs increase productivity, people likely have more time to go to outlets like games. And if LLMs increase production quality and production capability of games, you're going to have more content, which leads right into a business model like ours.

所以我們認為，不僅核心受眾會繼續玩遊戲，他們還需要放鬆的途徑。如果LLM（學習管理軟體）提高了生產力，人們可能會有更多的時間去玩遊戲等娛樂活動。如果 LLM 能夠提升遊戲的製作品質和製作能力，你就會有更多的內容，這正好符合我們這樣的商業模式。

Bernard McTernan, Needham.

伯納德·麥克特南，尼德姆。

I wanted to follow up on e-commerce. Given the self-service launch, are you seeing any changes in the type size of customers that are entering because of self-service versus the prior managed service? And then also as a follow-up, we've seen some examples in our own tracking of apps that aren't e-commerce with the pixel. Is this a one-off or a new growth vector that you're pursuing within this?

我想進一步了解電子商務的情況。鑑於自助服務的推出，您是否觀察到，透過自助服務進入的客戶類型規模與先前透過託管服務進入的客戶類型規模相比有任何變化？此外，作為後續研究，我們在自己對非電子商務應用程式的追蹤中也看到了一些使用像素的例子。這是一次性的嘗試，還是您正在探索的新的成長方向？

Yeah. I mean, obviously, with opening up self-service, we're not gating to a minimum GMV anymore. So in the prior year, we had pretty high minimum standards for businesses. Now you're getting companies that have a few hundred thousand dollars of GMV, a couple of million dollars of GMV a year buying on our platform.

是的。我的意思是，很顯然，隨著自助服務的開放，我們不再限制最低商品交易額 (GMV) 了。所以去年，我們對企業的最低標準要求相當高。現在，一些年交易額達幾十萬美元、數百萬美元的公司也開始在我們平台上購買。

What's nice about that is not only do we see the ability to track performance very clearly on a brand that's small, but they start spending money, the money translates to revenue. And it's really obvious in their numbers that their pre-existing revenue was a few hundred thousand dollars. I'll give you an example. We had, a year ago, this Israeli cookware company come live that did $4 million of revenue. And these types of stories make us really proud.

這樣做的好處不僅在於我們能夠非常清楚地追蹤小品牌的業績，而且當他們開始花錢時，這些錢就會轉化為收入。從他們的數據中可以明顯看出，他們之前的收入只有幾十萬美元。我給你舉個例子。一年前，一家以色列廚具公司成立，營收達 400 萬美元。這些故事讓我們感到非常自豪。

Last year, they scaled on our platform. 65% of all their user acquisition spend was on our platform. They scaled to $16 million of revenue and are profitable in doing so. And we can see the results directly translated from the spend on our platform to their growth. This year, they're ramping so quickly, again, putting most of their UA spend on our platform. It's looking like they're projecting that $80 million of revenue.

去年，他們透過我們的平台實現了規模化發展。他們所有用戶獲取支出中有 65% 都花在了我們的平台上。他們的收入規模擴大到 1600 萬美元，並且實現了盈利。我們可以直接看到用戶在我們平台上的消費與其業務成長之間的轉換效果。今年，他們的成長速度非常快，再次將大部分用戶獲取支出投入我們的平台。他們似乎確實預測收入將達到 8000 萬美元。

So you start seeing that kind of ramp-up in certain specific customers, and we know the product works and it works exceptionally well. That makes us proud. We think what we really like to do at this company is service the smaller businesses, help those smaller businesses become big businesses.

所以，你會看到某些特定客戶的需求開始成長，我們知道這款產品有效，而且效果非常好。這讓我們感到自豪。我們認為，我們公司真正想做的就是服務小型企業，幫助這些小型企業發展成為大企業。

That's how we built into gaming. We help the indices first. Now every gaming company in the world that's in mobile that's substantial is probably working with our platform in a major way. But in e-commerce and these other categories, we feel the same way, help the smaller businesses, help them scale their business, and then we'll get to all the brands over the coming years.

這就是我們進軍遊戲產業的方式。我們先幫助指數。現在，世界上所有規模較大的行動遊戲公司可能都在以某種重要方式與我們的平台合作。但在電子商務和其他這些領域，我們的想法也一樣，幫助小型企業，幫助他們擴大業務規模，然後在未來幾年裡，我們將惠及所有品牌。

In new categories, again, we're not gating anything. So if you end up in auto insurance and you sign up for our platform today, you can go live. We're still early in these other categories because for each one of these cases, we have to tune the model somewhat, and we're focused right now on anything that's transactional-based business versus a lead-based business. So think if you're a transactional-based business, you're not e-commerce, but you're fintech or something. That could work really well out of the box.

在新類別中，我們同樣不會設定任何門檻。所以，如果你最終選擇了汽車保險，並且今天註冊了我們的平台，你就可以上線了。在其他這些類別中，我們仍處於早期階段，因為對於每種情況，我們都必須對模型進行一些調整，而且我們目前專注於基於交易的業務，而不是基於線索的業務。所以想想看，如果你是一家以交易為基礎的企業，那你不是電子商務，而是金融科技之類的企業。這或許能直接奏效。

That's another category of growth. But then there's a huge sector of lead gen businesses. You want to collect the user's information and sell them in a call center. That's going to come too for us in the coming months where we really focus on a category like that. It's our job to service every transactional category that buys on a performance basis and do it well so that the 1 billion-plus daily actives we see on our platform can get a diverse set of content that they can enjoy.

這是另一個成長類別。但除此之外，還有龐大的潛在客戶開發業務領域。你想收集用戶信息，然後透過呼叫中心出售這些資訊。在接下來的幾個月裡，我們也會專注於這類產品類別。我們的職責是服務每一個按效果付費的交易類別，並且做好服務，以便我們平台上超過 10 億的日活躍用戶能夠獲得各種各樣的內容，讓他們盡情享受。

Matthew Cost, Morgan Stanley.

馬修‧科斯特，摩根士丹利。

I guess starting just to put a finer point on some of the comments from your prepared remarks. because the data points from the channel were a little unusual and thin, did you see a test from Meta doing more advertising in the in-game ad environment? And whether or not you did, how should the market think about what potential impact a scale player like Meta, for example, getting more aggressive in that market would mean? And then a second question just around MAX. You've been in the business of convincing people to switch mediation platforms for a long time. I guess in light of that experience, how would you characterize the moat around MAX?

我想先對您準備好的發言稿中的一些評論進行更細緻的闡述。因為頻道的數據點有些異常且不足，您是否注意到 Meta 在遊戲內廣告環境中進行更多廣告投放的測試？無論你是否真的這麼做了，市場應該如何看待像 Meta 這樣的大型企業在該市場採取更激進的策略可能帶來的潛在影響？然後是關於 MAX 的第二個問題。你從事說服人們更換調解平台的工作已經很久了。根據那次經歷，你會如何描述 MAX 周圍的護城河？

Yeah. So, two good questions. So the Meta thing, Meta was a launch partner of MAX. They're a good partner. They've been in the MAX platform for a long time. They are a bidder on anything that has an ID today, and they're not a bidder on anything that does not have an ID. So think IDFA off. That's about two thirds of the full-screen ad units that they currently bid on. They started bidding more on that, as you saw on LinkedIn. They did not start bidding on no ID traffic, and they might very well bid on no ID traffic in the next couple of quarters.

是的。兩個好問題。所以，Meta是MAX的創始合作夥伴。他們是很好的合作夥伴。他們已經在MAX平台上運作很久了。今天，只要有身分證明，他們就會參與競標；只要沒有身分證明，他們就不會參與競標。所以，別考慮IDFA了。這大約佔他們目前競標的全螢幕廣告單元的三分之二。正如你在領英上看到的那樣，他們開始加強競標。他們還沒有開始競標無 ID 流量，而且他們很可能會在接下來的幾季開始競標無 ID 流量。

As you know from our business over the last couple of years that you've been following, there have been numerous times where we've added more competition in the MAX auction. Unity has grown with Vector, Liftoff rolled out Cortex and it's growing really quickly. Moloco has been growing and we turned them into a bidder in the last year. You had Google switch to bidding. Every time this has happened, there's been this confusion around, well, in an auction dynamic-based system, competition should decay AppLovin's edge. You've never seen that happen. And this is what I tried to cover in the talk track.

正如您過去幾年一直關注的我們的業務一樣，我們曾多次在 MAX 拍賣中增加競爭。Unity 借助 Vector 發展壯大，Liftoff 推出了 Cortex，並且發展非常迅速。Moloco 一直在發展壯大，去年我們促使他們成為競標者之一。你讓谷歌改用競價模式。每次發生這種情況時，都會出現這樣的困惑：在基於拍賣動態的系統中，競爭應該會削弱 AppLovin 的優勢。你從未見過這種情況。這就是我在演講環節中試圖涵蓋的內容。

The reason is that, one, we're really big and really good at what we do. But two, every impression is not worth the same thing. Our model is exceptionally good at valuing the impression for what it is. for our data and our customers. Sometimes it values it really high where we make a lot of money.

原因有二：一是我們規模龐大，二是我們非常擅長自己所做的事情。但第二，並非每一次印像都具有相同的價值。我們的模型非常擅長準確評估印象的價值，這對我們的數據和客戶都至關重要。有時它的價值很高，我們能從中賺很多錢。

Sometimes it values it low where we make less money or potentially lose a lot of money. And what happens when you get these bidders getting better is that they take some of those impressions that we value low where we might not have made any money or any material money and now we get a 5% amount that we can tax the bidder in our ecosystem.

有時它的價值被低估，導致我們賺的錢減少，甚至可能損失很多錢。當這些競標者水準提高時，他們會拿走一些我們原本估值較低、可能根本賺不到錢或任何實質收益的展示機會，而現在我們可以從中抽取 5% 的佣金，並向我們生態系統中的競標者徵稅。

So growth in the ecosystem allows the MAX market economics to improve, and we've never seen growth in the ecosystem cause harm to our AppLovin bidding environment since we got so good. Now this is post AXON 2, where we became the market's dominant player. So I do think it's an important point that people are getting lost on. So I want to see if you have another question on that before I go on to the mediation discussion.

因此，生態系統的成長使得 MAX 市場經濟得以改善，而且自從我們發展得如此出色以來，我們從未看到生態系統的成長對我們的 AppLovin 競價環境造成損害。現在是 AXON 2 發布之後，我們成為了市場上的主導企業。所以我認為這是一個人們容易忽略的重要點。所以，在繼續討論調解問題之前，我想看看你是否有其他相關問題。

Okay. So on mediation, look, we got in the market with MAX when we bought it and built that technology. We didn't buy technology. We just bought a way into the market. That was in 2018. The MAX platform before we bought MoPub, had already grown to become #1 in the space. When we bought MoPub, we put the two together. As everyone knows, we're very dominant in the sector. We took the MoPub technology in sub-90 days, threw it out, migrated everyone that we could over to our platform, rebuilt the best features of MoPub, put them on top of the best features of MAX and ended up with a very dominant position.

好的。所以，關於調解，你看，我們收購 MAX 並建立這項技術後就進入了市場。我們沒有購買技術。我們剛剛買了進入市場的途徑。那是2018年的事了。在我們收購 MoPub 之前，MAX 平台已經發展成為該領域的領頭羊。收購 MoPub 時，我們將這兩家公司合併了。眾所周知，我們在該領域佔據絕對優勢。我們在不到 90 天的時間內就掌握了 MoPub 的技術，然後將其拋棄，將我們能遷移到的所有用戶都遷移到我們的平台上，重建了 MoPub 的最佳功能，並將它們與 MAX 的最佳功能相結合，最終獲得了非常強大的市場地位。

At this point in mediation, there's been a few players that are sort of out of business, and then there's Google, Level Play and us. And then obviously, you have the start-up that people have been discussing. So in a world where Cloud X becomes a start-up that comes into the space, you have to talk about like what are they walking into? How is the world different today versus what it was?

在調解進行到這個階段，已經有幾家公司倒閉了，剩下的還有Google、Level Play 和我們。當然，還有大家一直在討論的那家新創公司。所以，在 Cloud X 成為一家進入該領域的新創公司之後，你必須討論他們究竟要面對什麼？如今的世界與過去有何不同？

The moat around our mediation is not because of the mediation. We're very good. We've got the most bid density. In any mediation A/B test, if you talk to publishers, you'll hear MAX does better. But we don't blow it out of the water. We're a few percentage points better than other mediations. If someone wanted to pay a bonus to cover that, they could potentially pay a bonus to cover that.

我們調解過程中所存在的障礙並非源自於調解本身。我們非常優秀。我們的競標密度最高。在任何媒體 A/B 測試中，如果你與出版商交談，你會聽到 MAX 的表現更好。但我們並沒有把它徹底擊敗。我們比其他調解機構好幾個百分點。如果有人願意支付獎金來彌補這部分損失，他們或許可以支付獎金來彌補這部分損失。

Where it gets really expensive for the publisher and where we're really locked in is that we have the best advertising solutions on the market. In fact, for a lot of these publishers, we're over 50% of all their user acquisition spend. They can't go get that anywhere else. If they go off MAX, that decays. And so they're left in a world where they have the best buying tools. They have the best monetization tools. It becomes a really strong 360 solution and their growth depends on it.

對出版商來說，真正昂貴的地方在於，我們擁有市場上最好的廣告解決方案，而我們也因此被牢牢鎖定。事實上，對於許多這樣的出版商來說，我們的投入佔其用戶獲取總支出的 50% 以上。他們在其他地方都得不到那種東西。如果他們停止使用 MAX，效果就會衰減。因此，他們身處在一個擁有最佳購買工具的世界。他們擁有最好的獲利工具。它成為一個非常強大的全方位解決方案，他們的發展也取決於此。

And then the MAX ecosystem is not growing slow, as we've talked about in prior calls, this is a double-digit, very strong growing category where these publishers are seeing their businesses improve because of the improvement in our technology. When you've got that in the foundation and you've got a really strong moat with technologies that no one else can replicate or have, then you end up with a sticky solution, and we're very confident our solution is just that.

而且，正如我們在之前的電話會議中討論的那樣，MAX 生態系統的成長速度並不慢，這是一個兩位數的強勁成長類別，這些出版商的業務由於我們技術的改進而得到改善。當你擁有這樣的基礎，並且憑藉其他人無法複製或擁有的技術構築起一道非常強大的護城河時，你最終會得到一個具有粘性的解決方案，我們非常有信心我們的解決方案正是如此。

James Heaney, Jefferies.

James Heaney，傑富瑞集團。

How are you thinking about the AXON marketing investment in 2026? I know you talked late last year about a goal of maybe $1 million of spend per day. Is that still the right way to think about it? And is ultimately GA kind of the biggest prohibiting factor to you guys doubling down on that investment?

您如何看待 AXON 在 2026 年的行銷投資？我知道你去年底曾談到每天消費100萬美元的目標。這種思考方式現在還正確嗎？最終，GA 是否是你們加強投資力度的最大阻礙因素？

Yeah. I mean, look, right now, we're testing. So numbers aren't that big. You didn't see anything reflected in our EBITDA margins. In fact, you saw them go up to 84%.

是的。我的意思是，你看，現在我們正在進行測試。所以數字並沒有那麼大。你並沒有在我們的 EBITDA 利潤率中看到任何反映。事實上，你看到它們最高達到了 84%。

We think the growth in the business is so high and our LTV to CAC is going to look so juicy that you're not going to see much decay in our EBITDA margins from ramping up marketing. If you do, then that would be a great thing. It just means we're ripping on the marketing side, and we see a ton of customers that we can bring in to accelerate growth. But it's much more likely that we're going to go controlled on this because we do want to take our time to build out tools.

我們認為業務成長潛力巨大，而且我們的 LTV 與 CAC 比值非常可觀，因此加強行銷投入不會對我們的 EBITDA 利潤率造成太大影響。如果你能做到，那將會是一件非常好的事。這只是意味著我們在行銷方面取得了巨大成功，我們看到了大量可以吸引來加速成長的客戶。但我們更有可能採取管控措施，因為我們確實想花時間開發工具。

When we start buying and go GA and bring customers in, if we don't have good content creation tools for them, create the video with generative AI tools, create the interactive pages with generative AI tools and do it in an exceptionally strong way, they won't have as much success on our platform. So that's a long way of saying we're in no rush. We're seeing 30-day LTV to CAC roughly. If we see that at scale, we're going to scale, but we're not in a rush because we want the tools to catch up to the opportunity.

當我們開始購買內容、使用 GA 並吸引客戶時，如果我們沒有好的內容創作工具，無法使用生成式 AI 工具創建影片、創建互動式頁面，並且做得非常出色，那麼他們在我們的平台上就不會取得太大的成功。所以說，我的意思是，我們不急。我們看到 30 天 LTV 與 CAC 大致相當。如果我們看到這種模式具有規模性，我們就會擴大規模，但我們並不著急，因為我們希望工具能夠跟上這種機會。

Alec Brondolo, Wells Fargo.

Alec Brondolo，富國銀行。

I appreciate it. I want to double click on the Meta thing because I kind of think that's the most -- out of all the noise in the market, I think Meta is by far the most important. Right now, probabilistic targeting is the basis of competition between all the mobile game and app networks. I think there's no doubt that you can definitely compete effectively against Meta on a probabilistic basis.

謝謝。我想重點說 Meta，因為我覺得在市場上所有的喧囂中，Meta 無疑是最重要的。目前，機率性定向是所有行動遊戲和應用網路之間競爭的基礎。我認為毫無疑問，從機率上講，你完全可以有效地與 Meta 競爭。

When I talk to investors, I think the concern in the market is that Meta could find a way to bid deterministically and opt out ATC traffic with their audience graph, which would give them an advantage over the vertical ad networks. Do you think that's possible from a tech perspective? And if it was, would you still be confident that market expansion can offset the loss of market share?

當我與投資者交談時，我認為市場擔心的是 Meta 可能會找到一種方法，利用其受眾圖譜進行確定性競價並選擇退出 ATC 流量，這將使他們比垂直廣告網絡更具優勢。你認為從技術角度來看，這有可能實現嗎？如果真是如此，您是否仍然相信市場擴張能夠彌補市場份額的損失？

Yeah. I mean, look, I think it's possible from a technology perspective. I think it's bluntly against Apple's terms. I don't think the space is big enough for Meta to say they want to violate the platform they depend on terms. So I'm skeptical what you said is going to happen.

是的。我的意思是，從技術角度來看，我認為這是有可能的。我認為這完全違反了蘋果公司的條款。我不認為 Meta 有足夠的理由公開表示他們想要違反他們所依賴的平台條款。所以我懷疑你說的事情是否會發生。

It makes no logical sense. in a world where they're bidding deterministically or probabilistically on no IDFA, they're still competing against the AXON 2 model. Five years ago, when Meta was really big in the space, and I think this is what's throwing people off. People recall a time Meta was half the space. They think it's going to be half the space again.

這完全沒有邏輯意義。在一個他們以確定性或機率性方式競標，且不使用IDFA的世界裡，他們仍然在與AXON 2模型競爭。五年前，Meta 在這個領域非常火爆，我認為這就是讓人們感到困惑的原因。人們還記得 Meta 曾經只佔現在一半的空間。他們認為空間會再次縮減一半。

Meta has been on IDFA-based and Google ad ID-based traffic since that no IDFA change. Nothing has changed for them. What's changed in the marketplace is that the other ad platforms that are built for this category, Unity, Liftoff, Moloco, etc, have gotten better.

Meta 一直基於 IDFA 和 Google 廣告 ID 進行流量匹配，IDFA 沒有任何變化。他們的生活沒有任何改變。市場的變化在於，其他專為此類廣告而建立的廣告平台，如 Unity、Liftoff、Moloco 等，都得到了改進。

Now we've gotten the best. AXON 2 was the biggest breakthrough in a model in this category period, and we were able to end up becoming the number one by a lot. AXON 2 didn't exist five years ago. So there's no world where Meta is going to end up becoming that kind of a dominant player in the face of this competition. In fact, I don't see a world anyone else can because they're going up against that dominance. And these models, as they build more data, it's a closed-loop model that's continuously reinforcing itself and getting smarter.

現在我們得到了最好的。AXON 2 是該類別中最大的突破，我們最終以絕對優勢成為第一名。五年前，AXON 2 還不存在。所以，Meta 絕對不可能在這場競爭中成為那種佔據主導地位的玩家。事實上，我認為其他人看不到這樣的世界，因為他們都在與這種主導地位對抗。隨著這些模型累積更多數據，它們會形成一個閉環模型，不斷自我強化，變得越來越聰明。

Our model is so far into getting smart for this niche. The niche isn't that small, and we've got such a strong position. It's highly unlikely that someone else is going to come in and materially disrupt it. So a long way of saying, again, no, we don't see what people are so afraid of. We think psychologically, people just index on numbers from five years ago and think, oh, Meta is going to ramp to that. But just ask the customers you talk to, what's the share of wallet between us to them and to everyone else on IDFA-based traffic, and that will give you an indicator of how good we are.

我們的模型目前已經朝著這個細分市場變得智慧化邁出了重要一步。這個細分市場其實並不小，而且我們已經佔據了非常有利的地位。不太可能有人會介入並對其造成實質的干擾。所以說了這麼多，其實就是想再次強調：不，我們不明白人們到底在害怕什麼。我們認為，從心理學角度來看，人們只是以五年前的數字作為參考，然後認為，哦，Meta 將會成長到那個層次。但只要問問你接觸的客戶，在基於 IDFA 的流量中，我們佔他們的錢包份額，與其他所有客戶相比，這就能讓你了解我們有多優秀。

Perfect. That's super helpful. And maybe just one additional question. On the marketing dollars that you're spending against AXON today, where you're earning the 30-day LTV to CAC breakeven, what are the most effective channels? Like I see the ads on Google Search, I see the ads on Facebook and Instagram. Where are you finding the most success? Which platforms are you finding the most success in leveraging to acquire customers today?

完美的。這太有幫助了。或許我還有一個問題。對於您目前在 AXON 上投入的行銷資金，在實現 30 天 LTV 與 CAC 收支平衡的情況下，哪些管道最有效？就像我在谷歌搜尋上看到廣告一樣，我也在Facebook和Instagram上看到廣告。你覺得在哪些方面最容易成功？目前，您認為哪些平台在獲取客戶方面最有效？

We're -- it's a great question. I mean we're too early in testing to really assess or give anything directional because we don't want to be misleading. But I'll say some of the coolest yields that we've seen and coolest partnerships are these partnerships with the measurement companies. So if you talk to triple, they've got a sponsorship from us, TBPN, the podcast branded their Gong with our brand, and that's driven a bunch of business.

這是一個很好的問題。我的意思是，現在測試還處於早期階段，無法真正評估或給出任何方向性建議，因為我們不想誤導他人。但我要說的是，我們所看到的最酷的收益和最酷的合作關係，就是與測量公司的這些合作關係。所以如果你和 Triple 談談，他們得到了我們 TBPN 的贊助，播客用我們的品牌給他們的 Gong 打上了烙印，這帶來了很多生意。

So stuff like that, we're at such an early point. Part of the direct response channel here is actually just building the brand. So when we get placements like that, people start hearing about this Axon Ads platform. Then they go to Google, and we don't have any SEO yet. Once we get history and it put more content out on our blog, which we're committed to do and get more link backs, we'll also have SEO. But today, in the absence of Google search words are also great. We're starting to get brand recognition. People know our platform is critical.

所以諸如此類的事情，我們還處於非常早期的階段。直接回應管道的一部分其實就是品牌建立。所以當我們獲得這樣的廣告位時，人們就開始聽到 Axon Ads 這個平台。然後他們會去谷歌搜索，而我們當時還沒有任何SEO優化。一旦我們掌握了歷史數據，並在我們的部落格上發布更多內容（我們致力於這樣做並獲得更多反向連結），我們也會擁有 SEO。但如今，在沒有Google搜尋的情況下，文字搜尋也同樣重要。我們的品牌知名度開始提升了。人們都知道我們的平台至關重要。

They go search Google and we're the first ad word now. We're not the first organic link because we don't have that history. So all this stuff is going to improve over time. And we think our job is to combo brand advertising alongside this direct response to really unlock growth in this thing.

他們去谷歌搜索，現在我們的廣告詞排在第一位。我們不是第一個有機鏈接，因為我們沒有那段歷史。所以隨著時間的推移，所有這些都會得到改善。我們認為我們的工作是將品牌廣告與這種直接回應相結合，從而真正釋放這一領域的成長潛力。

Clark Lampen, BTIG.

克拉克·蘭彭，BTIG。

At the risk of, I guess, just sort of overdoing it around this Meta point, I wanted to see if you guys can just -- that's lapping. So I think I am overdoing it right now. But just remind us, I guess, the sort of scale of bidding that you guys are doing right now. I think the $11 billion or over $11 billion number that you provided in the blog post is the most recent example.

我猜，我可能有點離題了，但我還是想看看你們能不能──這算是跑圈吧。所以我覺得我現在有點用力過頭了。不過，我想請你們提醒我們一下，你們現在進行的競標規模有多大。我認為你在部落格文章中提到的 110 億美元或超過 110 億美元的數字是最新的例子。

Is that something that really needs to be replicated in this segment of the market for probabilistic bidding to really drive signal or capture a signal that would allow somebody to replicate the same sort of degree of efficacy. I think it would just be helpful for people to understand maybe even if somebody wanted to move in this direction, what are the sort of capital constraints of doing something like that?

在這個細分市場中，機率競價是否真的需要被複製，才能真正驅動訊號或捕捉訊號，從而使其他人能夠複製相同的效果？我認為，如果有人想朝這個方向發展，了解做這件事的資金限制是什麼，這對人們來說會很有幫助。

Yeah. I mean, look, first of all, like at the base, there are different approaches to modeling these businesses. Meta has a model that's fantastic, super sophisticated. They're built for social. Most of their training data is on social. We've got a model that's cutting edge, very sophisticated, built for our ecosystem. Our training data is predicated on that ecosystem. Our customers are perfectly tuned for that ecosystem. We started in gaming, where the biggest there is in game UA dollars. So those dollars aren't just going to shift because someone else comes into the space.

是的。我的意思是，首先，從根本上來說，對這些企業進行建模有不同的方法。Meta 的模型非常棒，極為複雜。它們專為社交而設計。他們的大部分訓練數據都來自社群媒體。我們擁有一個尖端、非常複雜的模型，專為我們的生態系統而設計。我們的訓練資料是基於此生態系統的。我們的客戶與該生態系統完美契合。我們從遊戲產業起家，因為遊戲領域的用戶獲取資金規模最大。所以，即使有其他人進入這個領域，這些資金也不會輕易轉移。

The dollars are already locked in on us, and our model is very smart. When there's a high-value user, sometimes our model bids thousands of dollars on a CPM basis. The model knows what works for these customers. So it's just not something that's conceivable in a marketplace where you have budgets that are already given to us. and we're on cutting-edge technologies for someone else to come in.

美元已經鎖定在我們身上了，而且我們的模式非常巧妙。當遇到高價值用戶時，我們的模型有時會以 CPM 出價數千美元。該模型了解哪些方法對這些客戶有效。所以，在一個預算已經分配給我們，而且我們採用的是尖端技術的市場中，讓別人來搶佔先機，這根本不可能。

Now, if we stop innovating and we decay on the technology side and we stop providing the service to the customers where they start looking around, yes, those things could change. But that's not going to happen overnight. And the one thing we can say confidently 14 years into being an engineering product-led org is you've seen us move incredibly quick. We are innovators. Engineering team is top-notch, second to none in this field. And so we're very much focused on continuing to push our models forward, and we're doing it from a very strong leadership position.

如果我們停止創新，在技術方面衰落，停止為客戶提供服務，以至於他們開始四處尋找其他選擇，那麼是的，這些事情可能會改變。但這不會一蹴可幾。在過去 14 年裡，作為一個以工程產品為主導的組織，我們可以自信地說，你們已經見證了我們驚人的發展速度。我們是創新者。工程團隊一流，在這個領域無人能及。因此，我們非常專注於繼續推進我們的模型，並且我們正處於非常強大的領導地位。

Okay. And if I may just ask sort of a very quick follow-up. prospecting campaigns. This was a product launch that sort of came about recently. There was a lot of very strong feedback in the channel throughout the quarter around the efficacy for some customers seeing full shift in sort of repeat versus new customer mix. Could you talk about the way that, that product -- I know you give sort of specifics around your non-gaming business versus the gaming one, but would it be possible to sort of contextualize the impact for us or maybe what you're seeing with advertiser behavior heading into 2026?

好的。如果可以的話，我想快速問一下後續問題。關於客戶開發活動。這是最近才推出的產品。本季度，通路內收到了許多非常強烈的回饋，一些客戶表示，他們的滿意度發生了顯著變化，回頭客與新客戶的比例也發生了徹底的轉變。您能否談談該產品——我知道您在非遊戲業務和遊戲業務方面給出了一些具體細節，但能否為我們解釋一下這對我們的影響，或者您認為到 2026 年廣告商的行為會有哪些變化？

Yeah. I mean, you've probably talked to some of the advertisers, so you hear positive commentary on it, but it's really hard for advertisers to understand the incremental value of retargeting. It's really easy to understand the incremental value of a new customer. And so it wasn't lost on us for 1 year, 1.5 years when we got into the market that just a universal campaign that had a split of new customers in retargeting wasn't the final answer. We needed more targeting for these customers to map to their goals.

是的。我的意思是，你可能已經和一些廣告商談過了，所以你會聽到一些正面的評價，但廣告商真的很難理解重新導向的增量價值。理解新客戶的增量價值其實很容易。因此，在我們進入市場的1到1.5年裡，我們始終明白，僅僅開展一個將新客戶分成兩部分進行再行銷的通用行銷活動並不是最終的解決方案。我們需要更精準地定位這些客戶，以便幫助他們達成目標。

So we rolled out this product in Q4. We let them upload their historical purchases. Using that data, our model can start tuning away from purchasers from the past and towards users that they've never seen before. Results were fantastic. It takes a while in advertising marketplace for every new product to get adoption, but we've seen really quick adoption on this one because when they flip the switch, they instantly start seeing many more new customers.

因此，我們在第四季推出了這款產品。我們允許他們上傳歷史購買記錄。利用這些數據，我們的模型可以開始從過去的購買者轉向他們以前從未見過的用​​戶。結果非常棒。在廣告市場中，任何新產品都需要一段時間才能被市場接受，但我們看到這款產品很快就被市場接受了，因為當他們啟用該產品時，他們立即開始看到更多的新客戶。

Robert Coolbrith, Evercore.

Robert Coolbrith，Evercore。

Adam, you mentioned a couple of times now the connection between ad ROAS and MAX. Just wondering if we could maybe put a finer point on it. If a publisher were to give another platform first look at inventories something like that, would that significantly degrade that connectivity between A ROAS and MAX from a UA perspective for that publisher. The I'll just have a quick follow-up.

Adam，你已經多次提到廣告 ROAS 和 MAX 之間的連結。我只是想知道我們是否可以把它說得更清楚一些。如果出版商將類似這樣的庫存優先提供給另一個平台，從用戶獲取的角度來看，這是否會顯著降低該出版商的 A ROAS 和 MAX 之間的連接性？我再簡單跟進一下。

Yeah. I mean, our terms are that they can't do that. So that's not part of the equation.

是的。我的意思是，我們的條件是他們不能這樣做。所以這不屬於考慮因素。

Got it. Perfect. And then I think there's also a notion out there that demand partners might bid differently in a different environment or something along those lines. Do you have any thoughts on that just --

知道了。完美的。而我認為還有一種觀點認為，需求夥伴在不同的環境下可能會採取不同的出價方式，或是類似這樣的情況。你對此有什麼想法嗎？--

I mean, look, MAX is fully fair, transparent. We get audited by the bidding partners. It's the vast majority of the marketplace. Wouldn't it be a foolish business decision to bid differently in a very small platform versus the platform that owns most of the space doesn't make any sense. So if someone is going to be a competitive bidder, they're almost certainly going to want to be a competitive bidder in the marketplace leading market.

我的意思是，你看，MAX 完全公平、透明。我們會接受競標夥伴的審計。它佔據了市場絕大部分份額。在一個規模很小的平台上採取與佔據大部分市場份額的平台不同的競價策略，這難道不是一個愚蠢的商業決定嗎？這根本說不通。因此，如果有人想成為競標者，他們幾乎肯定會想在市場領先的市場中成為競標者。

Got it. And then just one last quick follow-up on the breakage that you mentioned in terms of the qualified leads to go live or launch. Just wondering if you could talk about maybe some of the inhibiting factors today and how you plan on solving some of those.

知道了。最後，關於您提到的合格潛在客戶數量不足的問題，我想快速跟進。我想問您能否談談目前存在的一些阻礙因素，以及您打算如何解決這些問題。

Yeah. The biggest one is that they just don't have video ads for our platform made to the type of format that we need. So those generative AI tools should be able to raise that to a much higher percentage of customers going live.

是的。最大的問題是，他們沒有為我們平台製作符合我們需求的影片廣告格式。因此，這些生成式人工智慧工具應該能夠將上線客戶的比例提高到更高的水平。

Jim Callahan, Piper Sandler.

吉姆·卡拉漢，派珀·桑德勒。

Just a follow-up on the market growth. You've kind of talked about this double-digit framework through the back half of '25. Can we expect that through 2026? Is that like a step function change that we think can continue?

關於市場成長的後續報導。你已經在2025年下半年討論過這個兩位數成長框架了。我們能預期這種情況會持續到2026年嗎？這是否類似我們認為可以持續發生的階躍變化？

Look, MAX is growing really quickly. A lot of it's driven by the strength of our platform. So when we're growing 70% plus year-over-year, a lot of the dollars spent on our platform are the publishers buying customers. So it fuels growth in the ecosystem. so long as we're doing a good job and the other marketing platforms in gaming are doing a good job, the MAX market is likely to be growing really quickly.

你看，MAX發展得真快。這很大程度上得益於我們平台的強大實力。因此，當我們的年增長率達到 70% 以上時，我們平台上花費的許多錢都是出版商購買客戶的。因此，它能促進整個生態系的發展。只要我們和其他遊戲行銷平台都做得好，MAX市場就很有可能快速成長。

And this is now no longer a case where you just don't have other data points. Unity Vector is growing really quickly. You've got a company like Moloco that's private, but talking about going public that's growing quickly. Liftoff did testing the waters in a road show. So those numbers were out there that are growing really quickly. When you've got all these market players growing quickly, that MAX market is growing in a way that people I don't think understood or would have expected, and we're not seeing any slowdowns there.

現在的情況已經不再是缺乏其他數據點了。Unity Vector發展非常迅速。像 Moloco 這樣的公司，雖然是私人企業，但正在考慮上市，而且發展迅速。Liftoff公司透過巡迴演出試水溫。所以，這些數字確實存在，而且成長速度非常快。當所有這些市場參與者都在快速成長時，MAX市場的成長方式是人們沒有理解或預料到的，我們沒有看到任何放緩的跡象。

Got it. That's helpful. And then just you've talked about this reinforcement learning framework for gaming. Now that we're, I guess, like 1.5 years into e-commerce, how does e-commerce reinforcement learning kind of compare? Is it faster, slower? Does it take longer?

知道了。那很有幫助。然後你剛才也談到了用於遊戲的強化學習框架。現在我們進入電子商務領域大概已經一年半了，電子商務強化學習與傳統學習相比如何？是更快還是更慢？需要更長時間嗎？

I mean, what we've talked about is that the gaming model based on its own results can improve. So it creates its own memories, it gets smarter. The e-commerce model is the same thing, but I do just remind people that we're much earlier in e-commerce. We have very little data in the marketplace. So if you think about transactional volume that we have in gaming, we've got most of the market inside our system.

我的意思是，我們討論過的是，基於自身結果的遊戲模型可以改進。所以它會創造自己的記憶，變得越來越聰明。電子商務模式也是如此，但我只是想提醒大家，我們目前還處於電子商務發展的早期階段。我們掌握的市場數據非常少。所以，如果你看看我們在遊戲領域的交易量，你會發現我們系統內已經佔據了大部分市場份額。

So the model is pretty complete from a data perspective. And then also, we've had now, I think it's roughly three years to continuously improve that model. And as you've seen, we've continuously improved it many times over the last 12 to 14 quarters.

因此，從數據角度來看，模型相當完善。而且，我認為我們已經有大約三年的時間來不斷改進這個模型。正如你所看到的，在過去的 12 到 14 個季度裡，我們已經多次對其進行了改進。

E-commerce is newer. It's starting from an earlier place. We have much less data penetration. So we can't go and track like, okay, in a stable world, how much is the model improving because the model is going to be improving much more so from our team improving the model with changes and then just customers coming on. Every new customer that pixels their site passes data to us, both engagement data and transactional data. So as we get customer ramp up, our data penetration in the market is going to go from very little to much greater. And as that happens, that's going to be a big catalyst for improvement in that category.

電子商務是比較新的事物。它是從更早的地方開始的。我們的數據普及率要低得多。所以我們不能去追踪，在一個穩定的世界裡，模型改進了多少，因為模型會隨著我們團隊不斷改進模型以及客戶的加入而得到更大的改進。每個在其網站上新增像素的新客戶都會向我們傳遞數據，包括互動數據和交易數據。隨著客戶數量的成長，我們在市場上的數據滲透率將從非常低大幅提高。隨著這種情況的發生，這將極大地促進該領域的進步。

Stephen Ju, UBS.

Stephen Ju，瑞銀集團。

I think you've previously talked about AppLovin being demand constrained versus supply constrained. But can you help clarify particularly, yes, there's yourself and the private company you just called out as well as Unity all out there getting better as well. And I think investors are so accustomed to thinking about all of this being zero sum. So can you help us think about how much more supply there can be with perhaps new publishers and what existing publishers who already accept that can think about doing?

我認為你之前討論過 AppLovin 是受需求限製而不是受供給限制。但您能否幫忙澄清一下，特別是，您自己、您剛才提到的那家私人公司以及 Unity 都在不斷進步。我認為投資人已經習慣把這一切都當作零和遊戲。那麼，您能否幫助我們思考一下，透過引進新的出版商，還能增加多少供應量？以及已經接受這種供應的現有出版商可以考慮採取哪些措施？

Yeah. I think we're a long way from needing new publishers. I mean we've talked about, one, I think at this point, people should believe it's not a zero-sum market. This isn't rideshare where one person takes a ride and another loses. So if that was the case, with our growth and our market position being this dominant, there'd be no conceivable way these smaller gaming ad networks could be growing the way that they are.

是的。我認為我們離需要新的出版商還很遠。我的意思是，我們已經討論過，第一，我認為在這一點上，人們應該相信這不是一個零和市場。這跟拼車不一樣，拼車不是一個人搭車，另一個人吃虧。所以如果真是那樣的話，以我們如此快速的成長和如此強大的市場地位，這些規模較小的遊戲廣告網路根本不可能發展到現在這樣的程度。

So if we set aside the notion of zero sum, then you have to ask, well, how many transactions can we drive to this 1 billion-plus users. 1 billion users playing mobile games, casual games every single day, shopping users because they're adults, they skew female, are not done at this number of dollars. I mean if you just translate to Meta's users, it's 3 times more users, a little bit more time spent. But the revenue that they're driving to advertisers over what we're driving and the whole space is driving to advertisers is probably 8 times.

所以，如果我們拋開零和博弈的概念，那麼你就得問，我們能為這超過10億的用戶帶來多少筆交易？這10億用戶每天都在玩手機遊戲、休閒遊戲，他們也是購物用戶，因為他們是成年人，而且女性用戶居多，他們的消費遠不止這些。我的意思是，如果只看 Meta 的用戶數量，用戶數量是原來的 3 倍，花費的時間也略多一些。但他們為廣告商帶來的收入，比我們以及整個產業為廣告商帶來的收入可能高出 8 倍。

So there's a lot of room to go on monetizing this audience. We've also given you historically that our conversion rates on 1,000 impressions were about 1%. They're obviously higher now. We talked about that a year or two ago. The conversion rate is higher, but we think that could go as high as 5%. That's what we see when we're serving an ad where the model is confident that the user is going to take an action.

因此，在利用這部分受眾群體獲利方面還有很大的提升空間。我們也曾提供歷史數據，即每 1000 次展示的轉換率約為 1%。現在它們的價格顯然更高了。我們一兩年前就討論過這個問題。轉換率更高了，但我們認為最高可達 5%。這就是我們在投放廣告時所看到的，因為模型確信用戶將會採取行動。

Now, why isn't our whole business converting at 5%? Why aren't we doing $50 billion of revenue on the system? Well, we don't have enough advertisers yet to know what the user is going to be into at that moment. So we constantly serve them gaming ads, some users are into a new game, they're converting at 5%, 50 over 1,000. Other users aren't and the conversion rate is atrocious. It might be 2 over 1,000, and it dilutes you to 1%.

那麼，為什麼我們整個業務的轉換率沒有達到 5% 呢？為什麼我們沒能從這個系統中獲得 500 億美元的收入？嗯，我們目前還沒有足夠的廣告商來了解用戶在那個時刻會喜歡什麼。所以我們不斷地向他們投放遊戲廣告，有些用戶喜歡上了一款新遊戲，轉換率達到 5%，也就是 1000 人中有 50 人轉換。其他用戶則不然，轉換率非常糟糕。可能是 2/1000，這會將你的比例稀釋到 1%。

E-commerce has given us a path to diversity, but we only have a few customers. Once we get deeply penetrated into the space and we've got really diverse content to show the customer, we think that conversion rate is going to keep rising, and it's going to keep rising really quickly.

電子商務為我們提供了實現多元化的途徑，但我們的客戶仍然很少。一旦我們深入滲透到這個領域，並且擁有真正多樣化的內容可以向客戶展示，我們認為轉換率將會持續上升，而且會上升得非常快。

Got you. And secondarily, I think you've articulated a desire to go after advertisers of a certain size. So first, before targeting the advertisers with perhaps larger budgets, but maybe more upper funnel. So is that part of the market in addressable right now? Or is that something that might be in the roadmap in the future?

抓到你了。其次，我認為你已經明確表達了想要爭取一定規模的廣告商的意願。所以首先，在瞄準那些預算可能更大，但可能更注重行銷漏斗上層的廣告商之前。那麼，這部分市場目前是否可以開拓？或者說，這會成為未來發展規劃的一部分？

I think we're going to be going after anything that's not brand dollars. So by brand, I mean, companies that aren't optimizing to point of transaction, whether they optimize point of transaction or something performance-based like a lead, we're going to go after them.

我認為我們將追求品牌資金以外的一切。所以，我所說的品牌，指的是那些沒有在交易點進行優化的公司，無論他們優化的是交易點還是基於績效的方面，例如潛在客戶，我們都會爭取他們。

Now, we're not going to build out the sales force to go after the -- even in the performance space, advertisers then need you to go to their agency that need you to take a lot of time to go get a lead to go get an advertiser live. When I talk about this Israeli cookware company, it was a business that was probably spending a couple of thousand dollars a day ramped up to many tens of thousands, possibly hundreds of thousands of dollars a day on our platform. They ramped up really quickly. They scaled their business from near $0 to $16 million to now hoping for $80 million this year.

現在，我們不會組成銷售團隊去爭取——即使在效果廣告領域，廣告商也需要你去他們的代理機構，而代理機構需要你花費大量時間去獲取線索，才能讓廣告商上線。當我談到這家以色列廚具公司時，它最初可能每天在我們的平台上花費數千美元，後來逐漸增加到每天數萬美元，甚至數十萬美元。他們的擴張速度非常快。他們的業務規模從幾乎為零發展到 1,600 萬美元，現在希望今年能達到 8,000 萬美元。

That's an example of taking a customer that didn't have much of a business, plugging them into our tool set, letting our technology do the job and creating a huge business for them and a great relationship for us. That's the type of stuff that really cascades into much more value creation over time because it happens quickly, they become dependent on our platform. We love those stories because we're helping them build the business.

這是一個將原本業務規模不大的客戶納入我們的工具集，利用我們的技術為他們創造巨大業務，並為我們建立良好合作關係的例子。這種事情隨著時間的推移會真正帶來更大的價值創造，因為它發生得很快，他們就會依賴我們的平台。我們喜歡這些故事，因為我們正在幫助他們建立事業。

If we help Coca-Cola place more dollars in advertising, we're not moving the needle of anything in the economy. Coca-Cola is sort of blindly putting dollars out there. But if we can help customers spend the dollars, see the direct correlation of that revenue, hire people, grow their business. It's a fantastic story for us because it shows that we're growing the world's economy, we're creating jobs, and we're really creating this dependency on our platform.

如果我們幫助可口可樂公司增加廣告投入，對經濟的任何方面都不會產生影響。可口可樂公司有點像是盲目地投入資金。但如果我們能夠幫助客戶花錢，看到收入與招募員工、發展業務之間的直接關聯，那麼他們就能實現目標。這對我們來說是一個非常棒的故事，因為它表明我們正在促進世界經濟成長，我們正在創造就業機會，而且我們正在真正讓人們依賴我們的平台。

So really, that's what we're focused on. And in games, that was indie game developers and brands that's going to be these performance D2C companies, much more of the Shopify merchants, much less of those big brands that people talk about that buy through Madison Avenue agencies.

所以，這才是我們真正關注的重點。在遊戲領域，這指的是獨立遊戲開發者和品牌，這些品牌將成為以效果為導向的 D2C 公司，更多的是 Shopify 商家，而不是人們常說的那些透過麥迪遜大道廣告公司購買廣告的大品牌。

Cory Carpenter, JPMorgan.

科里·卡彭特，摩根大通。

Matt, I got to give you a question before the call ends. So I'll ask about the 1Q guide, the 5% to 7% sequential growth. It's above what you typically guide to in 1Q. Of course, this year, you have the added headwind, if you will, from the e-commerce business seasonality. Could you just talk about some of the assumptions you're making in that outlook and the trends you're seeing so far in the year in e-commerce and gaming?

馬特，通話結束前我得問你個問題。所以我將詢問第一季業績指引，即 5% 至 7% 的環比成長率。這高於您通常在第一季的指導值。當然，今年，電子商務業務的季節性波動也帶來了額外的阻力。您能否談談您在展望未來時所做的一些假設，以及您今年迄今為止在電子商務和遊戲領域觀察到的趨勢？

Yeah. So, consistent with kind of our normal practice, Cory, we're guiding to the level where we have a very high level of confidence. For Q1, obviously, coming from Q4, we had a very strong exit rate. So given the factors that Adam was talking about before, the performance of the mobile gaming business, the e-commerce launch as well as the prospecting model, we had a lot of growth in Q4. So the exit rate was quite good.

是的。所以，科里，按照我們通常的做法，我們會引導到我們非常有信心的水平。很明顯，第一季度，由於第四季度的情況，我們的退出率非常高。因此，考慮到亞當之前提到的那些因素，行動遊戲業務的表現、電子商務的推出以及潛在客戶開發模式，我們在第四季度取得了巨大的成長。所以退出率相當不錯。

And then that's partially offset by just seasonality going from Q4 into Q1 normally being a weaker seasonal period. And then we also had a couple of days less in Q1 versus Q4. So it's partially offset against that. So that's how we kind of landed in that 5% to 7% sequential.

然後，由於第四季到第一季通常是淡季，季節性因素也會部分抵消這種影響。此外，第一季的工作天數也比第四季少了幾天。所以這在一定程度上抵消了上述影響。所以，這就是我們最終確定5%到7%這個順序的原因。

And then, Adam, you alluded to, I think, an unlock that you saw a couple of weeks ago in the e-commerce business. Anything you can elaborate on what that was would be helpful.

然後，亞當，我想你提到了幾週前你在電子商務領域看到的突破。如果您能詳細說明一下那是什麼，那就太好了。

Yeah. Look, the team is constantly improving the model. We do a lot of testing. We saw a material lift in the model. We rolled it out. Advertisers saw a huge improvement in return on ad spend. They started putting more budget into our system quickly. So those are the types of things that really catalyze growth for us. When we talk to advertisers and say, test the product in our system, it's test now, but be patient because understand performance today is not going to be indicative of performance in six months.

是的。你看，團隊一直在不斷改進模型。我們做了大量的測試。我們看到模型中出現了實質的提升。我們已經把它推出來了。廣告商的廣告支出報酬率大幅提升。他們很快就開始加大對我們系統的預算投入。所以，這些正是真正能促進我們發展的因素。當我們與廣告商溝通並表示「在我們的系統中測試產品」時，現在就可以進行測試，但請耐心等待，因為今天的測試結果並不能代表六個月後的效果。

Our system is constantly improving because our team is improving these models. They do internal research, they use external research published in the AI field. These techniques can apply to what we do. And when they apply well, these advertisers see gains.

我們的系統不斷改進，因為我們的團隊一直在改進這些模型。他們進行內部研究，也使用人工智慧領域已發表的外部研究成果。這些技巧可以應用在我們的工作上。如果運用得當，這些廣告商就能獲得收益。

In gaming, they've seen gains very consistently since we launched AXON 2. In e-commerce, again, we're starting from a lower point because you need to get more data and you need to get more time to make the models more complex to use that data and create a better value for the customer. We're now doing our job. We're improving the model. They're seeing the result. They put more money into our platform.

在遊戲領域，自從我們推出 AXON 2 以來，他們的業績一直穩定成長。在電子商務領域，我們同樣是從較低的起點出發，因為你需要獲取更多的數據，並且需要更多的時間來建立更複雜的模型，以便利用這些數據為客戶創造更大的價值。我們現在正在執行任務。我們正在改進模型。他們看到了結果。他們向我們的平台投入了更多資金。

Ralph Schackart, William Blair.

拉爾夫·沙卡特，威廉·布萊爾。

Adam, maybe just staying on that, the model unlock for a second. Maybe kind of frame it. Was it sort of like consistent with other unlocks that you've seen before? Was it different maybe in order of magnitude? And then I'm not sure if it's for you or Matt, but can you give us a sense once self-serve rolls out and rolls out to GA, how meaningful of an impact can this have on the business? Will this be a slow build? Could this be something that could potentially be additive to second half growth? Just any way you can kind of frame that for me, that would be great.

亞當，或許就先停留在那裡，讓模型解鎖一秒鐘。或許可以把它框起來。它和您之前見過的其他解鎖方式有相似之處嗎？或許數量級有所不同？然後，我不確定這是問你還是問 Matt，但你能否告訴我們，一旦自助服務推出並推廣到 GA，這對業務會產生多大的影響？這將是一個緩慢的建設過程嗎？這是否有可能對下半年的成長起到促進作用？如果你能幫我大致解釋一下，那就太好了。

Yeah. I mean, on the second one, look, our growth rates are really fast right now. And our business is really big. So taking a self-service platform and opening it up, I wouldn't imagine day one or month one or month two is really going to move the needle on the overall numbers. it's going to build over time.

是的。我的意思是，關於第二點，你看，我們目前的成長率真的很快。我們的業務規模非常大。所以，即使推出自助服務平台，我認為第一天、第一個月或第二個月也不會對整體數據產生太大影響，這需要時間慢慢累積。

Now, certainly, if you're bringing in dollars that you just didn't have before, it's going to add something. So I guess it depends on what our overall growth rates are, but the scale of the business at the growth rate that we have is getting to very, very large numbers quickly. So I would expect that to build over time. I wouldn't expect it to be immediately impactful as a major growth catalyst. It will be noticeable in our numbers.

當然，如果你能帶來以前沒有的資金，那肯定會有好處。所以我覺得這取決於我們的整體成長率，但以我們目前的成長率來看，業務規模正在迅速擴大到非常非常龐大的規模。所以我認為這種情況會隨著時間的推移而逐漸形成。我不認為它會立即成為主要的成長催化劑。這一點會在我們的數據中反映出來。

When it comes to improving the models, we don't call it out anymore. We're constantly improving the gaming models. It happens every single quarter that we find something. And then the e-commerce side is starting from a worse place. It's just earlier stage, less data like we've talked about. So the model uplifts from the team improving the model can be much more substantial.

至於如何改進模型，我們不再公開指出了。我們一直在不斷改進遊戲模型。每個季度我們都會發現一些東西。而且，電子商務方面的起點更差。這只是早期階段，數據還比較少，就像我們之前討論的那樣。因此，團隊改進模型所帶來的模型提升可能會更加顯著。

However, the e-commerce business, as you know, even if you took disclosure from last Q1, last Q1, it was roughly 10% of our business. So if you improve 10% of your business, 40%, you're still only getting a 4% uplift on the whole, right? So it's a much smaller potential today.

但是，如您所知，即使按照上個季度第一季的揭露數據來看，電子商務業務也只占我們業務的約 10%。所以，即使你改善了 10% 的業務，也就是 40%，整體上也只能獲得 4% 的提升，對吧？所以如今它的潛在價值就小得多。

Now what we know about scaling a business is that we've got to compound these gains so that our performance is unquestionably the best in the market to these customers. If they see that, they're going to invest more and more into us. Most of these customers have scaled businesses. Most of these customers already buy on social and search. We're a new entrant. Most new entrants in the field over the last few years have had done a bad job improving performance. So we know that our performance has to be top notch. We're working towards that, and these types of lifts can compound to get us to that answer.

現在我們對業務規模化所了解到的一點是，我們必須不斷累積這些收益，才能確保我們的業績毫無疑問地成為市場上最好的，從而贏得客戶的信賴。如果他們看到這一點，他們就會加大對我們投資。這些客戶大多擁有規模化企業。這些客戶中的大多數已經透過社群媒體和搜尋引擎進行購物。我們是新入行者。在過去幾年裡，該領域的大多數新進業者在提升業績方面做得都很糟糕。所以我們知道，我們的表現必須一流。我們正在朝著這個方向努力，而這些類型的提升可以累積起來，幫助我們找到答案。

Martin Yang, OpCo.

Martin Yang，OpCo.

In the past, you referenced expanding supply sources. Where do you rank among the growth drivers? Where do you rank that among the growth drivers for your overall performance now?

過去，您曾提到擴大供應來源。在成長驅動因素中，你排名如何？您認為這在目前推動公司整體業績成長的因素中處於什麼位置？

I mean, look, our supply is growing very quickly because MAX is growing very quickly, right? So like I do like fix A B on the marketplace that we have. It's a really large marketplace. If we go get a big publisher tomorrow, it's not going to move the needle in the business because the percentage growth from that publisher to the whole of our market isn't all that much. I mean, you're talking about a marketplace, if you know the scale of our business and the ad dollars that we gave you and then we're not the whole market, the MAX marketplace is well over $10 billion a year.

我的意思是，你看，我們的供應量成長得非常快，因為 MAX 的成長速度非常快，對吧？所以，就像我確實想在我們現有的市場上修復 A B 一樣。這是一個非常大的市場。即使我們明天就簽下一家大型出版社，也不會對業務產生太大影響，因為這家出版社為整個市場帶來的成長百分比並不高。我的意思是，你談論的是一個市場，如果你了解我們業務的規模以及我們給你的廣告收入，而且我們並不是整個市場，MAX 市場的年收入就超過 100 億美元。

There's not a lot of publishers that could put a dent in those overall numbers. We will eventually go out to new publishers. We get calls all the time. Every publisher wants a monetization platform like ours to help them monetize their business unless they're Facebook, Google or Amazon. And so there's a lot of opportunity out there for us to expand supply. But right now, we're focused on the demand side because that conversion rate can go up so much more than where it is today. We've got a lot of growth in front of us just by doing the demand generation well and improving our core models.

能夠撼動整體銷售紀錄的出版商並不多。我們最終會尋找新的出版商。我們常接到電話。除了 Facebook、Google 或 Amazon 之外，每個出版商都希望擁有像我們這樣的獲利平台來幫助他們實現業務盈利。因此，我們有很多機會可以擴大供應。但目前，我們專注於需求端，因為轉換率還有很大的提升空間。只要我們做好需求創造並改善核心模式，我們就有很大的成長空間。

Got you. A follow-up on the creative side. Do you view AXON for e-commerce, the creative format and the overall flow with the end cars and elements as an area of differentiation?

抓到你了。創意方面的後續報道。您認為AXON在電子商務領域的差異化優勢體現在哪裡？其創意形式以及與最終產品和元素的整體流程是否構成差異化優勢？

Yeah. I mean, look, our ads force attention. So it's a lot different than ads that people are used to. You'd say the closest to our ads is television because you get a 30-second clip. But as we know, most people are ADD and don't really watch the ad on TV these days. Our ads are over 30 seconds of engagement. The user can't do anything else. They're already on their phone. It's a full screen lockup, and it gives the advertiser somewhere between 30 to 60 seconds plus to engage the consumer with their content. They can't get an experience like that anywhere else.

是的。我的意思是，你看，我們的廣告能夠吸引註意力。所以它和人們習以為常的廣告大不相同。可以說，最接近我們廣告形式的是電視廣告，因為電視廣告只有 30 秒。但我們都知道，現在大多數人都有註意力缺陷障礙，根本不會認真看電視廣告。我們的廣告互動時間超過 30 秒。使用者無法進行其他任何操作。他們已經在玩手機了。它採用全螢幕鎖定模式，讓廣告主有 30 到 60 秒以上的時間來吸引消費者關注其內容。他們在其他地方是體驗不到這樣的經驗的。

So I'd call like the starting point in our business in terms of ad quality, the best that an advertiser can access anywhere in the world.

因此，我認為我們業務的起點，就廣告品質而言，是廣告商在全世界範圍內能夠獲得的最佳水平。

Vasily Karasyov, Cannonball.

瓦西里‧卡拉肖夫，《砲彈》。

Just to follow up on what you said earlier, Adam, about e-commerce model being at a stage where it can't learn as fast now. And I think in the blog post you made recently, you mentioned that the e-commerce targeting is different because there are many more parameters and they're different, right? There is a different LTV calculation. There is no advertising component.

亞當，我只是想補充你之前提到的，電子商務模式目前處於一個學習速度無法像以前那樣快的階段。我認為在你最近發表的部落格文章中，你提到電子商務定位有所不同，因為參數更多，而且這些參數也不同，對嗎？生命週期價值（LTV）的計算方法有所不同。不包含任何廣告。

So I guess it's unfair to ask you to compare where the e-commerce model is at this point in its existence compared to where AXON 2 was, right, because there isn't just enough data. But what gives you confidence that once you get enough data, it will work as well or competitive.

所以，我想，要求你將電子商務模式在當前發展階段與 AXON 2 的發展階段進行比較是不公平的，對吧，因為數據還不夠。但是是什麼讓你確信，一旦獲得足夠的數據，它就能發揮同樣出色或具有競爭力的作用？

Yeah. I mean, look, if you talk to the customer base, a lot of the customers are seeing equal performance on us to any other top-of-funnel discovery channel, including the largest social platforms today. So it's not like we're starting behind trying to catch up. We're starting in a competitive place trying to become the best in the world.

是的。我的意思是，你看，如果你和客戶群談談，很多客戶都認為我們的產品/服務與其他任何漏斗頂端發現管道（包括當今最大的社交平台）的效果一樣好。所以，這並不是說我們一開始就落後，然後努力追趕。我們起步階段競爭非常激烈，目標是成為世界第一。

Now, we're starting with very little data penetration. This is an important point. If you think about the gaming model, I made this point earlier, all the transactions in the space, say, all the IP in the space, all the ad impressions in the space, our model sees the vast majority of everything in the world of mobile gaming today. That allows our engineers to take that data and translate it into exceptional predictions on the other side.

現在，我們的數據滲透率非常低。這一點很重要。如果你考慮一下遊戲模式，我之前也提到過這一點，比如說，這個領域的所有交易，這個領域的所有知識產權，這個領域的所有廣告曝光，我們的模式涵蓋了當今行動遊戲領域的大部分內容。這使得我們的工程師能夠利用這些數據，並將其轉化為卓越的預測結果。

In e-commerce, we're starting with -- I mean, the reports show thousands of sites, right? So we're starting with thousands of sites in a world where we can pixel 10 million plus. We have a long ways to go to get to the same data place in e-commerce as we are in gaming.

在電子商務領域，我們是從——我的意思是，報告顯示有成千上萬個網站，對吧？所以，我們一開始只能處理數千個網站，而我們最終可以處理超過 1000 萬個像素。電子商務領域的數據水平要達到遊戲領域的水平，我們還有很長的路要走。

Now the good news is these models are exceptionally smart, and our team that's putting them together are exceptionally smart. So we don't need to have 100% market penetration or anything close to gaming to really make an impact.

好消息是，這些模型非常智能，我們負責組裝這些模型的團隊也非常有智慧。所以，我們不需要100%的市場滲透率，也不需要接近遊戲領域才能真正產生影響力。

We're already proving it works with very little market data penetration. Once that number starts growing quickly, you're going to start seeing this model just improving itself because it gets more data. And off of that incremental data, the same model is going to make better predictions, both for gaming and non, and that's really going to help catalyze growth in our business as we bring on advertisers. We think about every new advertiser as dollars, but just as importantly, data into the model.

我們已經證明，即使市場數據滲透率很低，這種方法也行之有效。一旦這個數字開始快速成長，你就會看到這個模型不斷自我改進，因為它獲得了更多的數據。基於這些增量數據，同樣的模型將能夠做出更好的預測，無論是遊戲領域還是非遊戲領域，這都將真正有助於促進我們業務的成長，因為我們將吸引廣告商。我們把每個新廣告商都看作是金錢，但同樣重要的是，他們也為模型帶來了數據。

So I guess it's more fair to say that what you have now in terms of the e-commerce model is already competitive. It's not like you're behind and need more.

所以我覺得比較準確的說法是，你們現在的電子商務模式已經很有競爭力了。這並不是說你落後了，需要更多。

Yeah. I mean like we're not an e-commerce brand, but if you talk to them, if you talk to 10, at least five are going to tell you our performance on our platform is really good.

是的。我的意思是，我們雖然不是電商企業，但如果你和他們交談，如果你和 10 個人交談，至少有 5 個人會告訴你，我們在平台上的表現非常好。

Tim Nollen, SSR.

蒂姆·諾倫，SSR。

Tail end of the discussion. So just a few tidying up questions actually. For Adam, first off, we're talking about e-commerce all this conversation, but there are other sectors beyond e-commerce that you're servicing. Can you just clarify, I guess, e-commerce is, by far, the largest nongaming sector, but what other sectors are you servicing? And how meaningful are they?

討論接近尾聲。所以其實只有幾個收尾問題。亞當，首先，我們一直在談論電子商務，但你服務的對像不僅限於電子商務，還有其他領域。您能否澄清一下，我想說，電子商務是迄今為止最大的非遊戲產業，但您還服務哪些其他產業？它們究竟有多大意義？

Yeah. Look, the other sectors are still early. We started calling it e-commerce, I call it web advertising now. Anyone with the website with a transactional business model should be able to work on our platform. However, we're much earlier in the model evolution for other businesses outside of e-com.

是的。你看，其他產業還處於早期階段。我們一開始稱之為電子商務，我現在稱之為網路廣告。任何擁有採用交易型商業模式的網站的人都應該能夠使用我們的平台。然而，對於電子商務以外的其他產業而言，我們目前仍處於模式演進的早期階段。

Yeah. And then a couple of bits for Matt actually. You've done a great job bringing costs down quite a lot. A lot of that is SBC, I think. You had a slight step-up, I guess, sequentially in your R&D number in the cost lines in Q4. I guess the question is with such a high adjusted EBITDA margin, you seem very confident in maintaining that. What do we need to know in terms of other cost items that could rise? How much could they rise in the coming several periods, not just Q1?

是的。然後還有幾句是給馬特的。你們在降低成本方面做得非常出色。我覺得其中很多都是SBC的功勞。我猜想，你們第四季研發成本項目上的數字較上季略有成長。我想問的是，在調整後的 EBITDA 利潤率如此之高的情況下，您似乎非常有信心維持這一水平。還有哪些成本項目可能會上漲，我們需要了解這些情況？未來幾個季度（不僅僅是第一季）它們可能會上漲多少？

And then last question is regarding cash. You've got now $2.8 billion of cash with $3.5 billion of debt on your balance sheet. Just wondering what your capital structure thoughts and your use of cash priorities would be from here?

最後一個問題是關於現金的。你現在擁有 28 億美元的現金，但資產負債表上卻有 35 億美元的債務。想了解您對目前的資本結構有何看法，以及您未來的現金使用優先事項是什麼？

Yeah. In terms of the margin, I mean, we feel very confident in the margin level that we're at today. The kind of X factors that could potentially change that in the near term or in the short term would be potentially performance marketing that we were talking about before. So to the extent that we see really great performance with some of the campaigns that we're running and we scale those up pretty significantly, we could see kind of a shorter-term impact in margin.

是的。就利潤率而言，我們對目前的利潤率水準非常有信心。可能在短期內改變這種情況的X因素，可能是我們之前討論過的效果行銷。因此，如果我們看到一些正在進行的行銷活動取得了非常好的效果，並且我們大幅擴大了這些活動的規模，我們可能會在短期內看到利潤率受到一定影響。

Obviously, we're going to be doing that in the way that we spend, generally speaking, which is very disciplined and then obviously looking at the returns. So we'll run return-based campaigns. So we'll get the spend back in a relatively short period of time. Adam mentioned 30-day return from those. So, overall, that's why we can feel so confident in the overall margin level that it shouldn't change materially from here.

顯然，我們會按照我們一貫的消費方式來做到這一點，我們的消費方式非常有紀律，然後當然也會關注回報。所以我們會進行以回報為導向的行銷活動。因此，我們可以在相對較短的時間內收回成本。Adam提到這些產品30天內即可回收成本。所以總的來說，這就是為什麼我們對整體利潤率水準充滿信心，認為它不會從目前的情況發生實質變化。

And then in terms of use of cash, obviously, our first priority is spending on the organic growth initiatives. So that's ensuring that we're continuing to retain our talent, compensating people very well. hiring to continue to support our growth initiatives like e-commerce and our engineering team, et cetera. But that really hasn't moved the needle, obviously, in terms of the cash and the cash balance growing. So then it's really what do we do with the cash after that. And we've been very active with our repurchase program, and we continue to plan to be.

至於現金的使用，顯然，我們的首要任務是投資有機成長計畫。這可以確保我們繼續留住人才，為員工提供優厚的薪酬，並繼續招聘以支持我們的成長計劃，例如電子商務和工程團隊等等。但顯然，就現金和現金餘額的成長而言，這並沒有真正產生任何影響。那麼接下來的問題就是，我們該如何處理這筆錢呢？我們一直積極推行股票回購計劃，並計劃繼續這樣做。

Okay. Thanks. I'm right at 60 minutes exactly. So good timing. Thanks.

好的。謝謝。我剛好花了60分鐘。時機真好。謝謝。

Perfect. Thank you.

完美的。謝謝。

And that concludes the question-and-answer session for this quarter. We thank you all for joining us today. Have a good afternoon.

本季問答環節到此結束。感謝各位今天蒞臨。午安.

Thanks, everyone.

謝謝大家。

Portions of this transcript marked (technical difficulty) indicate audio problems. The missing text will be supplied if a replay becomes available.

本文字稿中標示（技術難題）的部分錶示有音訊問題。如果有重播視頻，我們會補充缺少的文本。