Applovin Corp (APP) 2025 Q1 法說會逐字稿

Welcome to AppLovin's earnings call for the first quarter ended March 31, 2025. I'm David Hsiao, Head of Investor Relations. Joining me today to discuss our results are Adam Foroughi, our Co-Founder, CEO and Chairperson; and Matt Stumpf, our CFO. Please note, our SEC filings to date as well as our financial update and press release discussing our first quarter performance are available at investors.applovin.com.

歡迎參加 AppLovin 截至 2025 年 3 月 31 日的第一季財報電話會議。我是投資者關係主管 David Hsiao。今天與我一起討論我們成果的有我們的共同創辦人、執行長兼董事長 Adam Foroughi；以及我們的財務長 Matt Stumpf。請注意，我們迄今向美國證券交易委員會 (SEC) 提交的文件以及我們的財務更新和討論第一季度業績的新聞稿均可在 investors.applovin.com 上查閱。

During today's call, we will be making forward-looking statements, including, but not limited to, the future development and reach of our platform, our expected growth opportunities, the result in timing of our proposed sale of our games business, the efficiency of our operations, the expected future financial performance of the company and other future events.

在今天的電話會議中，我們將做出前瞻性陳述，包括但不限於我們平台的未來發展和影響力、我們預期的成長機會、我們提議出售遊戲業務的時機、我們營運的效率、公司預期的未來財務表現以及其他未來事件。

These statements are based on our current assumptions and beliefs, and we assume no obligation to update them except as required by law. Our actual results may differ materially from the results predicted. We encourage you to review the risk factors in our most recently filed Form 10-K for the year ended December 31, 2024. Additional information may also be found in our quarterly report on Form 10-Q for the fiscal quarter ended March 31, 2025, which will be filed today.

這些聲明是基於我們目前的假設和信念，除非法律要求，否則我們不承擔更新它們的義務。我們的實際結果可能與預測的結果有重大差異。我們鼓勵您查看我們最近提交的截至 2024 年 12 月 31 日的 10-K 表中的風險因素。您也可以在我們今天提交的截至 2025 年 3 月 31 日的財政季度的 10-Q 表季度報告中找到更多資訊。

We will also be discussing non-GAAP financial measures. These non-GAAP measures are not intended to be superior to or a substitute for 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.

我們也將討論非公認會計準則財務指標。這些非公認會計準則 (non-GAAP) 指標並非旨在優於或取代我們的公認會計準則 (GAAP) 結果。請務必查看我們投資者關係網站上的收益報告和財務更新中的 GAAP 結果以及 GAAP 與非 GAAP 財務指標的對帳。

This conference call is being recorded, and a replay will be available for a period of time on our IR website. 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.

本次電話會議正在錄製中，重播將在一段時間內在我們的 IR 網站上提供。現在我將把時間交給亞當和馬特來致開場白。然後主持人將帶領我們進行問答。

Thank you all for joining us today. Q1 2025 was another fantastic quarter, marked by resilience and robust growth. After seeing the stock price rise roughly 50 times in two years, we face short seller scrutiny, which we've addressed comprehensively and won't revisit here. Our mission remains clear, helping advertisers reach new customers profitably. It's important to remember, in our business, our only financial incentive is to drive measurable revenue and profitability to our advertisers.

感謝大家今天的參與。2025 年第一季又是一個出色的季度，具有韌性和強勁增長的特點。在看到股價在兩年內上漲了約 50 倍之後，我們面臨著賣空者的審查，我們已經全面解決了這個問題，這裡不再贅述。我們的使命始終很明確，那就是幫助廣告商以獲利的方式接觸新客戶。重要的是要記住，在我們的業務中，我們唯一的財務激勵是為我們的廣告商帶來可衡量的收入和盈利能力。

Without that, we could not scale our business nor would we get paid. As a leading performance marketing platform, our technological innovations have catalyzed the return to growth in the gaming ecosystem, reviving an industry that would otherwise be struggling without our advancements over the past two years.

如果沒有這些，我們就無法擴大業務，也無法獲得報酬。作為領先的績效行銷平台，我們的技術創新催化了遊戲生態系統的恢復成長，使這個在過去兩年中如果沒有我們的進步就會陷入困境的行業煥然一新。

We've empowered sophisticated media buyers investing over $10 billion annually with us, driving strong returns and generating significant impact for their businesses. We're now expanding into broader categories, confident in our ability to fuel their growth as we did for gaming.

我們為成熟的媒體購買者提供支持，他們每年與我們一起投資超過 100 億美元，為他們的業務帶來豐厚的回報並產生重大影響。我們現在正在向更廣泛的類別擴展，相信我們有能力推動它們的成長，就像我們對遊戲所做的那樣。

Few platforms operate at our scale, and we're proud of our role in driving economic growth. Our partners' vocal support this quarter was inspiring, and we're doing more business than ever.

很少有平台能像我們這樣規模運營，我們為自己在推動經濟成長方面所扮演的角色感到自豪。本季我們合作夥伴的大力支持令人鼓舞，我們的業務比以往任何時候都多。

Let's dive into our outstanding Q1 performance. The first quarter is typically challenging for advertising due to seasonality in fewer days compared to Q4, yet we achieved remarkable growth. How? We further refined our machine learning models, enabling mobile gaming companies to scale their campaigns on our platform. Less significant but impactful was the full quarter contribution from web advertisers.

讓我們深入了解我們出色的第一季業績。由於季節性因素的影響，第一季的廣告業務比第四季度少，因此通常面臨較大的挑戰，但我們仍然取得了顯著的成長。如何？我們進一步完善了機器學習模型，使行動遊戲公司能夠在我們的平台上擴大他們的活動。網路廣告商對整個季度的貢獻雖然不那麼重要，但影響卻很大。

This diversification is transformative and fuels our excitement for what's ahead. Today, we're thrilled to announce the signing of the definitive agreement to sell our games business in its entirety. This strategic move sharpens our focus on advertising. To our studio teams, you've been instrumental in building the technology that powers our AXON platform. As you transition to a company dedicated to game development, we're grateful for your contributions and excited for your future.

這種多樣化具有變革性，激發了我們對未來的興奮。今天，我們很高興地宣布簽署最終協議，出售我們的整個遊戲業務。這項策略舉措使我們更加重視廣告。對於我們的工作室團隊來說，你們在建立支援我們 AXON 平台的技術方面發揮了重要作用。當您轉型為致力於遊戲開發的公司時，我們感謝您的貢獻並對您的未來感到興奮。

Now looking ahead, we're focusing all resources on three key priorities for 2025. First, we're relentlessly improving our machine learning models. Our research science team is leveraging rapid AI advancements to deliver even greater value to our partners, ensuring our platform remains a performance leader. Second, we're advancing our e-commerce and web advertising solutions with three areas of focus. We're continuously refining our models.

展望未來，我們將集中所有資源實現 2025 年的三個關鍵目標。首先，我們正在不懈地改進我們的機器學習模型。我們的研究科學團隊正在利用快速的人工智慧進步為我們的合作夥伴提供更大的價值，確保我們的平台保持效能領先地位。其次，我們將從三個重點領域推進我們的電子商務和網路廣告解決方案。我們正在不斷完善我們的模型。

While our product already works well for many advertisers, it's still early days, and we believe it can be significantly better. Each iteration brings us closer to that goal.

雖然我們的產品已經為許多廣告商提供了良好的服務，但它仍處於早期階段，我們相信它可以變得更好。每次迭代都使我們更接近該目標。

We're enhancing integrations with third-party platforms and attribution vendors to provide advertisers with a seamless measurement experience. The web advertising space is more fragmented than apps. So this will take time, but it's a straightforward task. We're also developing a self-service dashboard, and we'll be launching it this quarter for select customers. Once fully rolled out, this tool will enable seamless automation, allowing new advertisers to set objectives, budgets, upload ads and let our system deliver results.

我們正在加強與第三方平台和歸因供應商的整合，為廣告主提供無縫的測量體驗。網路廣告空間比應用程式更加分散。所以這需要時間，但這是一項簡單的任務。我們也正在開發自助服務儀表板，並將於本季向部分客戶推出。一旦全面推出，該工具將實現無縫自動化，讓新廣告商設定目標、預算、上傳廣告並讓我們的系統產生效果。

While we've seen great performance so far in our web advertising pilot, we're currently less than 0.1% of the potential market of total advertisers. Each new partner adds to our growth. It will take a few quarters to refine these tools for a broader release, but when we launch self-service globally, we expect it to unlock a massive opportunity. Third, we're enhancing ad testing and automated ad creation. Improving the creative experience is a priority this year, enabling advertisers to better optimize campaigns effortlessly.

儘管我們的網路廣告試點迄今取得了出色的成績，但目前我們在整個廣告商潛在市場的份額還不到 0.1%。每一個新的合作夥伴都會促進我們的成長。需要幾個季度的時間來完善這些工具以供更廣泛的發布，但當我們在全球範圍內推出自助服務時，我們預計它將帶來巨大的機會。第三，我們正在加強廣告測試和自動化廣告製作。改善創意體驗是今年的首要任務，讓廣告主更輕鬆地優化廣告活動。

These initiatives are both immediate and long term, positioning us for sustained success. Now let me address the potential impact of tariffs. 90% plus of our revenue in advertising comes from mobile games, which aren't directly impacted by tariffs. In web-based advertising, some assume we rely heavily on large Chinese e-commerce businesses, which are impacted by the de minimis tariff exemption changes. In reality, we focus on mid-market web advertisers and aren't yet working with the largest players as we're refining our tools first.

這些舉措既是短期的，也是長期的，為我們持續成功奠定了基礎。現在讓我來談談關稅的潛在影響。我們的廣告收入 90% 以上來自手機遊戲，而手機遊戲並不受關稅的直接影響。在網路廣告領域，有些人認為我們嚴重依賴中國大型電子商務企業，而這些企業受到了最低限度關稅豁免變化的影響。實際上，我們專注於中型網路廣告商，尚未與最大的廣告商合作，因為我們正在首先完善我們的工具。

It is absolutely possible some of the merchants we do work with will have their businesses impacted by tariffs. However, with such low market penetration, we're well positioned to grow through macroeconomic changes like tariffs without any visible impact on our business trajectory. I also want to address a few topics in the news. On competition, we embrace it. It drives innovation and pushes us all to improve.

與我們合作的一些商家的業務絕對有可能受到關稅的影響。然而，由於市場滲透率如此之低，我們完全有能力透過關稅等宏觀經濟變化來實現成長，而不會對我們的業務軌跡產生任何明顯的影響。我還想談談新聞中的幾個主題。對於競爭，我們欣然接受。它推動創新並促使我們所有人進步。

With our growing data moat and AI expertise, we're confident in our leadership and ability to grow rapidly regardless of others' advancements. Regarding potential alternative payment systems in the App Store, we see this as a positive. Lower fees for content developers, our customers, means they can invest more in user acquisition, which benefits our platform. For context on our TikTok bid, please see my blog post we published a few minutes ago. Our lean team continues to impress, showcasing a model for how AI-based businesses can operate efficiently.

隨著我們不斷增長的數據護城河和人工智慧專業知識，無論其他人如何進步，我們都對我們的領導地位和快速發展的能力充滿信心。對於 App Store 中潛在的替代支付系統，我們認為這是一個積極的方面。對於內容開發者（我們的客戶）來說，降低費用意味著他們可以在用戶獲取方面投入更多，這對我們的平台有利。有關我們競標 TikTok 的背景信息，請參閱我幾分鐘前發布的部落格文章。我們的精實團隊繼續給人留下深刻印象，展示了基於人工智慧的企業如何高效運作的典範。

Our run rate adjusted EBITDA per employee in our advertising business has risen to approximately $4 million annually, reflecting our commitment to operational excellence and robust economics. Thank you for your continued support as we execute our vision to build one of the world's most influential marketing platforms. With each quarter, I grow more confident in our ability to deliver incremental value to our partners.

我們的廣告業務每位員工的調整後 EBITDA 已升至每年約 400 萬美元，這反映了我們對卓越營運和強勁經濟的承諾。感謝您一直以來的支持，我們將努力實現我們的願景，打造世界上最具影響力的行銷平台之一。每個季度，我對我們為合作夥伴提供增量價值的能力越來越有信心。

With that, I'll turn it over to Matt for a deeper dive on our financials.

說完這些，我會把話題交給馬特，讓他更深入地了解我們的財務狀況。

Thanks, Adam. And thanks, everyone, for joining us today. We're thrilled to share that Q1 was another outstanding quarter for us, showcasing the strength and efficiency of our business model. Total revenue soared 40% from the same period last year to $1.5 billion, and adjusted EBITDA increased a remarkable 83% to an impressive $1 billion, achieving a fantastic 68% adjusted EBITDA margin. We've driven a 600 basis points increase in our EBITDA margin over the previous quarter, a testament to our ability to leverage our primarily fixed cost base while fueling revenue growth through cutting-edge technology.

謝謝，亞當。感謝大家今天的參與。我們很高興地告訴大家，第一季對我們來說又是一個出色的季度，展現了我們商業模式的優勢和效率。總營收較去年同期飆升 40% 至 15 億美元，調整後 EBITDA 大幅成長 83% 至令人印象深刻的 10 億美元，調整後 EBITDA 利潤率達到驚人的 68%。我們的 EBITDA 利潤率比上一季提高了 600 個基點，這證明了我們有能力利用主要固定成本基礎，同時透過尖端技術推動營收成長。

In the first quarter, we generated $826 million in free cash flow, up a staggering 113% year-over-year. Quarter-over-quarter, our free cash flow grew 19%, representing an impressive 82% flow-through from adjusted EBITDA to free cash flow. At the end of the first quarter, we had $551 million in cash and cash equivalents.

第一季度，我們產生了 8.26 億美元的自由現金流，比去年同期成長了驚人的 113%。與上一季相比，我們的自由現金流成長了 19%，這意味著從調整後的 EBITDA 到自由現金流的轉換率高達 82%。截至第一季末，我們擁有 5.51 億美元的現金和現金等價物。

This quarter, we repurchased and withheld a total of 3.4 million shares for a total cost of $1.2 billion, primarily funded through our free cash flow as well as a temporary draw on our revolving credit facility, which we've now repaid. As a result of our strategic share management activities, we were able to reduce the total outstanding shares net of share issuances to employees, demonstrating our commitment to delivering value to shareholders and ended the quarter with 338 million shares outstanding.

本季度，我們回購並扣留了總計 340 萬股股票，總成本為 12 億美元，主要透過我們的自由現金流以及循環信貸額度的臨時提取提供資金，我們現已償還。透過實施策略性股份管理活動，我們能夠減少扣除向員工發行的股票後的總流通股數，這表明我們致力於為股東創造價值，本季末流通股數為 3.38 億股。

Shifting to the advertising business. We generated $1.16 billion in revenue and $943 million in adjusted EBITDA, achieving an incredible 81% margin. Our revenue growth in the quarter was driven by a combination of factors, including continued enhancements in our AI-driven technology which has delivered even better performance for the advertisers using AppDiscovery and the full quarter impact of our web-based advertising solution, which continues to perform strongly while coming off a seasonally high spend period for e-commerce.

轉向廣告業務。我們創造了 11.6 億美元的收入和 9.43 億美元的調整後 EBITDA，實現了驚人的 81% 的利潤率。本季我們的營收成長受到多種因素的推動，包括我們人工智慧技術的持續增強，這為使用 AppDiscovery 的廣告商帶來了更佳的業績，以及我們基於網路的廣告解決方案對整個季度的影響，該解決方案在擺脫電子商務季節性高支出期的同時繼續表現強勁。

Quarter-over-quarter flow-through from revenue to adjusted EBITDA was an exceptional 104%, which is slightly higher than our normal levels due to certain nonrecurring costs last quarter. However, after adjusting for these costs, our flow-through was still a robust 100% despite the step-up in data center costs I mentioned last quarter, highlighting our dedication to operating lean while scaling smartly.

收入與調整後 EBITDA 之間的環比流轉率達到了驚人的 104%，由於上個季度的某些非經常性成本，這一數字略高於我們的正常水平。然而，在調整這些成本之後，儘管我上個季度提到資料中心成本有所上升，但我們的流通量仍然保持了 100% 的強勁增長，這凸顯了我們在精益運營和智慧擴展方面的奉獻精神。

As Adam mentioned, we're excited to share that we've signed a definitive agreement with Tripledot Studios to divest our Apps business. Consideration includes $400 million in cash and a 20% ownership stake in the combined business. Subject to regulatory clearance, we anticipate closing this transaction in the second quarter, and we're confident in the success this business will achieve under new leadership, while we sharpen our focus on advertising. Finally, turning to our financial guidance for next quarter. In light of the Apps sale and our strategic focus on the Advertising business, we will only provide guidance for our Advertising segment.

正如亞當所提到的，我們很高興地告訴大家，我們已經與 Tripledot Studios 簽署了最終協議，剝離我們的應用程式業務。對價包括 4 億美元現金和合併後業務 20% 的所有權股份。在獲得監管部門批准後，我們預計將在第二季完成這筆交易，並且我們相信這項業務將在新的領導下取得成功，同時我們將更加專注於廣告業務。最後，談談我們對下個季度的財務指引。鑑於應用程式的銷售以及我們對廣告業務的策略重點，我們將僅為我們的廣告部門提供指導。

In the second quarter of 2025, for the Advertising business, we anticipate delivering between $1.195 billion and $1.215 billion in revenue with adjusted EBITDA between $970 million and $990 million, targeting an adjusted EBITDA margin of 81%. We're confident these targets position us to continue driving strong growth and value for our partners and shareholders.

2025 年第二季度，對於廣告業務，我們預計營收將在 11.95 億美元至 12.15 億美元之間，調整後的 EBITDA 將在 9.7 億美元至 9.9 億美元之間，調整後的 EBITDA 利潤率目標為 81%。我們相信，這些目標將使我們能夠繼續為我們的合作夥伴和股東帶來強勁的成長和價值。

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

現在，讓我們進入問答環節。

(Operator Instructions) Jason Bazinet, Citi.

（操作員指示）花旗銀行 Jason Bazinet。

So I just had one quick question, maybe two parts. On the guidance for next quarter, I think it implies something, maybe I'm doing the math wrong, 3% to 5% sequential growth, which seems a little bit slower than, I think, your aspiration just on the mobile-only part of the business.

所以我只想問一個簡單的問題，大概分成兩個部分。關於下個季度的預期，我認為它暗示了一些東西，也許我的計算是錯誤的，3％到5％的連續增長，這似乎比你對行動業務部分的期望要慢一點。

But maybe I'm misinterpreting something or maybe something else is going on. That's the first part of my question. And then the second one is I was just looking back towards your historic financials, and there have been some quarters where there were like sequential declines in ads revenue, but it ended up not being indicative at all of how much you were going to grow.

但也許我誤解了某些事情，或者也許發生了其他事情。這是我的問題的第一部分。第二個問題是，我只是回顧了你們的歷史財務狀況，發現有些季度的廣告收入出現連續下降，但最終根本無法顯示你們的成長幅度。

And so do you think investors should sort of brace for maybe, somewhere down the future, there could be a sequential decline in ads revenue, but it doesn't -- it's not really thesis changing? I guess that's my question. I think it was 2022, was the year. Those are my two questions.

那麼，您是否認為投資人應該做好準備，也許在未來的某個時候，廣告收入可能會持續下降，但事實並非如此——這實際上並沒有改變論點？我想這就是我的問題。我認為那是 2022 年。這是我的兩個問題。

So I can answer the second part. Matt can jump in if you want some guidance. But Jason, 2022 was a lot different. It was pre-AXON model, and since we've released it, we've obviously seen an immense amount of growth. This thing continues to get better as it scales.

所以我可以回答第二部分。如果您需要一些指導，馬特可以提供幫助。但傑森，2022 年的情況大不相同。它是 AXON 之前的型號，自從我們發布它以來，我們顯然看到了巨大的成長。隨著規模的擴大，這個東西會變得越來越好。

And you get a really strong flywheel embedded into this type of a machine learning model. As we get more impressions, more engagements with the ads, more conversions, the thing just get -- continues to get smarter and smarter. And then you pair that self-learning with the technological gains that our team continues to layer on top, and the growth of the business has obviously been phenomenal since. So I wouldn't look back at '22 and before.

您將獲得一個嵌入到這種機器學習模型中的非常強大的飛輪。隨著我們獲得更多的印象、更多的廣告互動、更多的轉化，事情就會變得越來越聰明。然後，你將這種自學與我們的團隊不斷累積的技術進步結合起來，業務的成長顯然從那時起就變得非常驚人。所以我不會回顧 22 年及之前的事。

Then to your first question, the other thing you got in our business is, in terms of seasonality, we have a bit of a unique Advertising business because you typically expect Q1 to be your worst season. And our seasons are tied to time spent on mobile device. And Q1 gets the benefit of a whole bunch of holiday days at the beginning of the quarter, and then you've got other things like Ramadan and spring break to take effect in Q1. Q2, you actually don't have almost anything as only the tail end gets into summer. Q3, you get summer.

然後回答您的第一個問題，您在我們的業務中發現的另一件事是，就季節性而言，我們的廣告業務有點獨特，因為您通常預計第一季是您最糟糕的季節。我們的季節與花在行動裝置上的時間息息相關。第一季在季度初享受了大量假期，然後齋月和春假等其他假期也會在第一季生效。Q2，其實你幾乎什麼都沒有，因為只是到了夏天的尾聲。Q3，你迎來了夏天。

Q4, you get the holidays. So if you look last year at our quarter-over-quarter sequential growth, the only single-digit quarter was Q2. Now we've said we're looking to grow 20% to 30% but there's a whole bunch of unpredictable reasons why, every year, we think the growth could actually be materially higher. We almost grew 20% quarter-over-quarter in a single quarter in Q1 over Q4 if you normalize by day. So the business is growing really quickly.

Q4，你有假日。因此，如果你看看去年我們季度間的連續成長，你會發現唯一一個個位數成長的季度是第二季。現在我們說過，我們希望實現 20% 到 30% 的成長，但由於一系列不可預測的原因，我們每年都認為實際成長率可能會更高。如果以天計算，第一季的季增幾乎達到 20%。因此業務成長非常快。

We continue to see a lot of excitement from advertisers on the platform. They're spending more than they've ever spent before on the platform, both across games and the web advertising initiative. And then even more exciting, I touched on this in the talk script. We're finally going to be releasing our new dashboard to some select advertisers for feedback this quarter. That's a huge catalyzing effect.

我們持續看到平台上廣告商的熱情高漲。他們在該平台上的投入比以往任何時候都多，包括遊戲和網路廣告。更令人興奮的是，我在演講稿中提到了這一點。我們最終將在本季度向一些選定的廣告商發布我們的新儀表板以徵求反饋。這是一個巨大的催化作用。

When we do go to a full self-service state, we're going to open up our platform from a very small amount of advertiser penetration to the entirety of the world being able to come on to our platform. Now there's a lot to do between here and there, but once we do that, again, we'll go through another one of those transformational moments where the business will just change a lot. So the past in our business isn't really indicative of the future growth trends because we just haven't been in a stable place yet.

當我們真正進入完全自助服務狀態時，我們將開放我們的平台，從只有極少量的廣告商可以使用，到全世界都可以使用我們的平台。現在我們還有很多事情要做，但是一旦我們完成了這些事情，我們就會再次經歷一次轉型時刻，業務將會發生很大的變化。因此，我們業務的過去情況並不能真正預示未來的成長趨勢，因為我們尚未達到穩定的狀態。

Matthew Cost, Morgan Stanley.

摩根士丹利的馬修‧科斯特。

Thank you for taking my qiestion. Maybe I'll start off just talking about category exposure. I think last quarter, you highlighted a couple of new categories including fintech and health care and insurance that you're having success moving into. So I guess, are you continuing to push into new categories beyond the ones that you've talked about?

感謝您回答我的問題。也許我首先要談的是類別曝光。我認為上個季度您重點介紹了幾個新類別，包括金融科技、醫療保健和保險，您已成功進入這些領域。那麼我想，除了您談論的類別之外，您是否還會繼續拓展新的類別？

And then when you're having those conversations with new types of advertisers what are the pain points that they're looking to have solved? Is it about onboarding those attribution partners? Is it about self-serve? And then I have one follow-up.

那麼，當您與新類型的廣告商進行對話時，他們希望解決的痛點是什麼？是否與加入這些歸因合作夥伴有關？是自助服務嗎？然後我還有一個後續問題。

Yes. So I would say, Matt, we're not looking to push into any category of web advertising right now. We've got a line out the door of customers that have been waiting to come on to the platform. And then we've onboarded, I think I said in a blog, hundreds of advertisers. But our team is small.

是的。所以我想說，馬特，我們現在不打算進軍任何類別的網路廣告。門口已經排起了長隊，等候上月台的顧客們已經排好隊了。然後我們就加入了，我想我在部落格裡說過，數百個廣告商。但我們的團隊很小。

So when I say we're not looking to push, we're looking to push over time, but we need to get the self-service tools into the market so that we can pair that with the team to automate a lot more of the processes. What we're excited about today is the model is still in infancy when it comes to web advertising. We think it can get a lot better.

所以當我說我們不想推動時，我們想隨著時間的推移而推動，但我們需要將自助服務工具推向市場，以便我們可以將其與團隊配對，以實現更多流程的自動化。今天令我們興奮的是，就網路廣告而言，該模式仍處於起步階段。我們認為情況會變得更好。

And then you layer on those attribution integrations and platform integrations and make the integration with advertisers more seamless. Both those two things, the automated tools, better integrations paired with a much better model, which our model is inevitably going to get better as the engineering team has more time to evolve it, we think we're going to get to a place in web advertising where we are in games.

然後，您可以對這些歸因整合和平台整合進行分層，使與廣告商的整合更加無縫。這兩件事，自動化工具、更好的整合以及更好的模型，隨著工程團隊有更多的時間來改進它，我們的模型必然會變得更好，我們認為我們將在網路廣告領域達到與遊戲領域一樣的地位。

Today, if you're a game advertiser, we're the best destination in the world to spend money. Every type of game regardless of whether it launches today or launched years ago in any category plugs into our system, puts in a return on ad spend goal and scales at their goals. We're not there yet on web across categories, but we're seeing better performance than we expected this early. We're going to get there, and when we get there and we pair that with the self-service dashboard and launch that product globally, you're going to have an effect that catalyzes a ton of long-term growth.

今天，如果你是遊戲廣告商，我們就是全世界最好的花錢目的地。每種類型的遊戲，無論是今天推出的還是幾年前推出的，都屬於任何類別，都可以接入我們的系統，設定廣告支出回報目標，並根據目標進行擴展。我們尚未在網路上實現跨類別的功能，但我們看到的效能比我們早期預期的要好。我們會實現這一目標，當我們實現這一目標並將其與自助服務儀表板配對並在全球範圍內推出該產品時，你將會產生催化大量長期增長的效果。

Great. And then I think you mentioned in the prepared remarks something about automated ad creation. So is that essentially tailored creative? And if so, what is the cost-benefit analysis given the cost of generating creative for advertisers in real time?

偉大的。然後我認為您在準備好的評論中提到了一些有關自動廣告創建的內容。那麼這本質上是量身訂製的創意嗎？如果是的話，考慮到即時為廣告主產生創意的成本，成本效益分析是什麼？

Yes. I mean, to answer the second part first, we don't know the cost-benefit until we do it. But what we want to do is generative AI-based ad creative. And if you think about the platform today, we serve over 1 billion daily active users. We're serving a lot of impressions.

是的。我的意思是，先回答第二部分，我們只做了才知道成本效益。但我們想要做的是基於人工智慧的生成廣告創意。如果你考慮一下今天的平台，我們每天為超過 10 億活躍用戶提供服務。我們提供了很多印象。

But if you take a single advertiser, say, Activision with Candy Crush, they upload 20 static videos, let's say, for their campaign. And that's what their team can go create to run on our platform. And those videos go and serve, and they drive a certain response rate. Well, in the world of large language models and customization, we're not going to get to the extreme where every single end user can see a differentiated video, but we can certainly get to a place where you can take the best videos that an advertiser uploads, run them through a large language model, can now put up more videos that are dynamically generated, run those videos through our platform, let the model then personalize the ad to the end user at a greater level. When you're able to test ad creative, it's one of the biggest levers that advertisers have to move the needle on growth on response rate.

但如果你選擇一個廣告商，比如說 Activision 的 Candy Crush，他們會上傳 20 個靜態視頻，作為他們的廣告活動。這就是他們的團隊可以在我們的平台上創建並運行的東西。這些影片會進行播放，並產生一定的回應率。好吧，在大型語言模型和定制的世界中，我們不會達到每個最終用戶都可以看到差異化視頻的極端，但我們肯定可以達到這樣的程度：你可以獲得廣告商上傳的最佳視頻，通過大型語言模型運行它們，現在可以上傳更多動態生成的視頻，通過我們的平台運行這些視頻，然後讓模型在更高層次上為最終用戶個性化廣告。當您能夠測試廣告創意時，它就是廣告商推動回應率成長的最大槓桿之一。

It's free effectively. It's a very low cost amount of production to get a much higher response rate on advertising. If we can systematically do it and we're certain we can, it's just taking time to get this to market, we're going to be able to extract a lot higher response rate from the audience who sees the advertisements. And again, every time we have one of those events, it creates both a short-term step-up in growth, and then that compounds over time because the system then starts getting more transactions through it. It learns better, and it continues to build on itself.

它實際上是免費的。以非常低的製作成本獲得更高的廣告回應率。如果我們能夠有系統地做到這一點，並且我們確信我們可以做到，只是需要時間將其推向市場，我們將能夠從看到廣告的受眾那裡獲得更高的回應率。再說一次，每次我們遇到這樣的事件，它都會在短期內帶來成長，隨著時間的推移，這種成長會越來越強烈，因為系統會開始透過它獲得更多的交易。它學得更好，並且能夠不斷自我提升。

So that's another one of those events that we think are going to be very impactful and it's something that's a priority for us to work on across the rest of the year.

所以，我們認為這是另一個將會產生巨大影響的事件，也是我們在今年剩餘時間內優先進行的工作。

Omar Dessouky, Bank of America.

美國銀行的奧馬爾‧德蘇基 (Omar Dessouky)。

Thanks taking my question. Now that you've been in the pilot -- or excuse me, that e-commerce has scaled for a couple of quarters, I was wondering whether you had any updated points of view on churn among your advertisers, whether you're seeing any churn. And what would your go-forward assumptions on churn be? That's one. And then whether you're seeing the spend per advertiser increase as the months go by or they may -- or if that spend per advertiser is just affected by seasonal trends.

謝謝您回答我的問題。現在您已經參與了試點——或者不好意思，電子商務已經擴大了幾個季度，我想知道您對廣告商的流失是否有任何最新的看法，是否看到了任何流失。您對客戶流失的未來假設是什麼？那是一個。然後，您是否看到每個廣告商的支出隨著時間的推移而增加，或者每個廣告商的支出是否僅受到季節性趨勢的影響。

Yes. So we disclosed to you, Omar, the -- I think I said 600-some-odd advertisers in the blog, $1 billion run rate. In terms of churn and web spend growth quarter-over-quarter, comp against Q4 versus Q1 isn't particularly fair in shopping, so I'm not going to disclose growth by advertiser. But when we think about churn with where we are in the advertising product for web, we're early. So inevitably, we're going to have some customers where it doesn't work for.

是的。因此，奧馬爾，我們向您披露了——我想我在部落格中說過 600 多個廣告商，10 億美元的運行率。就客戶流失率和網路支出的環比增長而言，第四季度與第一季的比較在購物方面並不是特別公平，因此我不會透露廣告商的成長情況。但是，當我們考慮到網路廣告產品的客戶流失率時，我們還處於早期階段。因此，我們不可避免地會遇到一些客戶無法使用我們的服務的情況。

Now actually, right before this call, we pulled the number. For advertisers that spent run rate $250,000 a year, we had sub-3% churn, so very little. Now that's not an acceptable number for us. In gaming, we have basically no churn. Unless a game is not marketable and it's going to go out of business, companies in gaming do not drop off our platform.

實際上，就在這通通話之前，我們撥通了這個號碼。對於每年花費 25 萬美元的廣告客戶，我們的客戶流失率低於 3%，非常低。對我們來說這不是一個可以接受的數字。在遊戲領域，我們基本上沒有客戶流失。除非某個遊戲沒有市場並且即將倒閉，否則遊戲公司不會放棄我們的平台。

We're basically a requirement to their success. We want to be that in web advertising as well. But again, remember, this product is months old. As the model gets better, we think we're going to be a destination for any type of advertiser that has a website or an app or both to come market themselves and have success where we become a required marketing destination for them in their business.

我們基本上是他們成功的必要條件。我們希望在網路廣告領域也能如此。但再次提醒，請記住，該產品已有幾個月的歷史了。隨著模型變得越來越好，我們認為我們將成為任何擁有網站或應用程式或兩者的廣告商的行銷目的地，並取得成功，成為他們業務中必需的行銷目的地。

Okay. And if I could just ask a quick follow-up on a different topic. You disclosed 49% year-on-year growth in net revenue per install. However, your e-commerce advertisers don't drive installs. They drive actions. So is that metric inclusive of the actions that are driven by e-commerce?

好的。我是否可以就另一個話題快速詢問一下。您揭露了每次安裝淨收入年增 49%。但是，您的電子商務廣告商不會推動安裝。它們推動行動。那麼該指標是否包括由電子商務推動的行動？

No, it's not, Omar. So it's inclusive of the revenue associated with those web-based advertisers, but that's driving then an increase in the net revenue per install because the installation volume in that metric is staying stagnant, right? It's only based on the CPI-based advertisers.

不，不是這樣的，奧馬爾。因此，它包括與那些基於網路的廣告商相關的收入，但這會推動每次安裝的淨收入增加，因為該指標中的安裝量保持停滯，對嗎？它僅基於基於 CPI 的廣告商。

Chris Kuntarich, UBS.

瑞銀的 Chris Kuntarich。

Adam, you mentioned that advertising could be better for some of the advertisers on the web-based experience. Curious if there's kind of a common thread that you would identify across those advertisers that we should be thinking about and some of the progress you could be making as we look forward into 2Q and later this year were just kind of road signs of progress for those advertisers.

亞當，您提到，對於一些基於網路的廣告商來說，廣告可以帶來更好的體驗。我很好奇，您是否能在那些我們應該考慮的廣告商中發現某種共同點，而當我們展望第二季度和今年晚些時候時，您可能取得的一些進展只是那些廣告商取得進展的路標。

Yes. I wouldn't say it's trends or anything specific because it's just not the way the models work. But if you think back a couple of -- on the game advertising -- pre-AXON 2, and what happened when we released the new version of the technology and what's built, the customers get better ROAS and more scale. Every time the model becomes more predictive, that's what happens. So I'd sort of rate us as like a B-plus right now with where we are.

是的。我不會說這是趨勢或任何具體的東西，因為這不是模型的工作方式。但是，如果你回想一下 AXON 2 之前的遊戲廣告，以及我們發布新版本技術時發生的事情和建構的內容，客戶會獲得更好的 ROAS 和更大的規模。每當模型變得更具預測性時，就會發生這種情況。因此，就我們現在的狀況而言，我會給我們 B+ 的分數。

For how little time we've had to market with the web-based model, it's performing really well. But we think it could perform a lot better. Now games have grown a ton over the last couple of years. I don't know exact number, but I think I put in the blog, ad spend on the platform has roughly quadrupled since we rolled out AXON 2. Now that happens because you have better ROAS and better scale because the model evolves and gets better.

儘管我們利用基於網路的模式進行行銷的時間很短，但它的表現卻非常好。但我們認為它的表現可以更好。在過去的幾年裡，遊戲已經取得了長足的發展。我不知道確切的數字，但我想我在部落格中提到過，自從我們推出 AXON 2 以來，該平台上的廣告支出大約增加了四倍。現在這種情況發生了，因為你擁有更好的 ROAS 和更好的規模，因為模型不斷發展並且變得更好。

So if we get to a place where, tomorrow, one of our engineers releases a new version of the web model that's 30%, 40% better, I wouldn't be surprised. That doesn't mean that ROAS is 30%, 40% better because usually we achieve in advertisers' ROAS goals, but it means at the comparable ROAS, they get 30%, 40% more scale. Then that starts compounding, but that ends up universal on the platform.

因此，如果明天我們的一位工程師發布了一個比現有版本好 30% 或 40% 的網路模型，我不會感到驚訝。這並不意味著 ROAS 會好 30% 或 40%，因為通常我們會實現廣告商的 ROAS 目標，但這意味著在可比的 ROAS 下，他們的規模會大 30% 或 40%。然後它開始復合，但最終在平台上變得普遍。

Got it. And maybe just one follow-up. As we talk to advertisers, one of the common feedbacks we hear is the desire to have exclusionary audiences. Curious kind of if you're thinking around offering that to your advertisers specifically on the web-based side has evolved since we've last spoken.

知道了。也許只需要一個後續行動。當我們與廣告商交談時，我們聽到的一個常見回饋是希望擁有排他性受眾。我很好奇，自從我們上次談話以來，您是否考慮向您的廣告商提供這種服務，特別是在基於網路的方面，這已經發生了變化。

Yes. So this is an interesting offering because it's related to how they're used to buying on Meta's properties. And across Meta, this is just common. Advertisers upload an exclusion list and say, I want to target new audiences at this percentage of dollars spent versus retargeting. The way we look at our product, again, is it's early. So our best gauge of success is how we can scale at the ROAS that they want, not by divvying up audiences but by just making the model better at matching.

是的。所以這是一個有趣的產品，因為它與他們習慣在 Meta 的資產上購買的方式有關。在 Meta 中，這很常見。廣告商上傳一份排除清單並說，我想以此百分比的花費來瞄準新受眾，而不是重新定位。再次強調，我們看待我們產品的方式是，現在還為時過早。因此，我們衡量成功的最佳標準是如何按照他們想要的 ROAS 進行擴展，不是透過劃分受眾，而是透過使模型更好地匹配。

Now over time, we may introduce things like exclusionary targeting. And we played around with some of it over the last few months as well, but it's not a focus for us because we're still too early in the product. If our product's performance continues to improve, we ramped up on hundreds of advertisers to $1 billion run rate, and I mentioned we're sub-0.1% penetration in the market.

現在，隨著時間的推移，我們可能會引入諸如排他性定位之類的措施。在過去的幾個月裡，我們也嘗試了其中的一些功能，但這不是我們的重點，因為我們對該產品還處於早期階段。如果我們產品的表現持續提高，我們將向數百個廣告商投放廣告，使運行率達到 10 億美元，而我之前提到，我們的市場滲透率不到 0.1%。

So despite being really early and despite having some bumps with some advertisers where we can look across the advertiser base and no, we're not making it work for everyone, if we just open up the platform today, you're going to see a ton of growth and we're going to be working for a lot of advertisers. Now we want to make it work for every single one, but we don't judge ourselves by saying we want to build products that are similar to what others do.

因此，儘管我們處於早期階段，儘管與一些廣告商之間存在一些障礙，我們可以縱觀整個廣告商群體，但我們並沒有讓它為所有人服務，如果我們今天開放平台，你就會看到大量的增長，我們將會為很多廣告商服務。現在，我們希望它能為每一個人服務，但我們不會評判自己是否想打造與其他人類似的產品。

We want to build products that are improved to what others do for our advertisers. And so we take these points of feedback. We think about them, but we're focused on our own path, which is improve the model and give the advertiser what we think is the best outcome.

我們希望為我們的廣告商打造比其他產品更優質的產品。因此我們採納了這些回饋意見。我們會考慮這些問題，但我們專注於自己的道路，即改進模型並為廣告商提供我們認為最好的結果。

James Heaney, Jefferies.

詹姆斯希尼，傑富瑞集團。

Great, thanks. Just be great to get an update on the velocity of new web advertiser additions in the quarter. Just interested to hear how the pace has changed since you last reported 600 customers in December. And anything around how to think about that going into Q2 and second half of the year? And then I had a follow-up.

太好了，謝謝。很高興能了解到本季新增網路廣告商的速度。我只是想知道自從您上次報告 12 月有 600 名客戶以來，速度發生了怎樣的變化。您如何看待第二季和下半年的情況？然後我進行了後續跟進。

Yes. So pace has slowed down because we just don't have the resources. So team is still about 20 people. So if you think about like what a team consists of, there's sales engineering, integration people, BD, analysts, sales, so there's not a lot of people in any one of those groups across this team. The objective right now is to give them the self-service dashboard, the automated tooling so that then they could start onboarding at the pace that they were able to do to get up to the 600 and change.

是的。由於我們沒有足夠的資源，所以步伐已經放慢。所以團隊仍然有大約20人。因此，如果你想想一個團隊是由什麼組成的，那麼就會有銷售工程師、整合人員、業務發展、分析師、銷售，因此這個團隊中任何一個群體中的人都不多。現在的目標是為他們提供自助服務儀表板和自動化工具，以便他們能夠以自己的速度開始入職，達到 600 並進行變革。

And I'm sure we're above that at this point, but the reality is we don't have the manual resources to onboard the line out the door. But we're there now with a dashboard that can really help the team. We're going to be in testing with it. We'll get early feedback. It looks quite good to us, and we believe it will be quite good for the advertisers.

我確信我們現在已經超過了這個目標，但現實是，我們沒有足夠的人力資源來支持生產線的運作。但我們現在有了可以真正幫助團隊的儀表板。我們將對其進行測試。我們會儘早得到回饋。我們認為這非常好，我們相信這對廣告商來說也會非常好。

Once we get that in the hands of advertisers where they can manually automate however they work with us, then the team is going to be able to onboard a lot more quickly. And so you'll start seeing phases of how we roll this out. You'll hear about it on Twitter, I'm sure, with people playing around with the new dash and then us getting it into the hands of the advertiser base this quarter.

一旦我們將這項技術交到廣告商手中，讓他們能夠手動自動化與我們合作的方式，那麼團隊就能更快加入。因此，您將開始看到我們如何實施此計劃的各個階段。我確信，您會在 Twitter 上聽到這個消息，人們正在試用這個新的 dash，然後我們會在本季度將它交到廣告商手中。

And then in the coming quarters, we'll start opening it up more and more, and that will really help onboarding of advertisers to the point where, eventually, we will completely open it up, go global with the product and then start really onboarding at a quick rate.

在接下來的幾個季度裡，我們將開始越來越多地開放它，這將真正有助於廣告商的入職，最終，我們將完全開放它，讓產品走向全球，然後開始真正快速地入職。

And then just as another question. We've heard from a lot of your e-com advertisers that the web product works really well for advertisers that optimize for those 24-hour conversion windows. I'd love to hear about any progress you're making with helping advertisers that have longer consideration windows and higher repeat purchases?

然後就像另一個問題一樣。我們從許多電子商務廣告商那裡聽說，對於針對 24 小時轉換視窗進行最佳化的廣告商來說，該網路產品確實非常有效。我很想聽聽您在幫助廣告主延長考慮窗口和提高重複購買率方面取得了哪些進展？

Well, look, we don't have a conversion API. We don't have what other companies have with e-mail address and phone number, so we can't attribute back anything outside of a cookie window. And the cookie window in this day and age, especially on Safari, is a very short window. What that means is that our model has to drive actions quickly. That's great for most advertisers.

嗯，看，我們沒有轉換 API。我們沒有其他公司所擁有的電子郵件地址和電話號碼，因此我們無法將任何資訊歸因於 cookie 視窗之外。而當今時代的 Cookie 窗口，尤其是在 Safari 上，是一個非常短的窗口。這意味著我們的模型必須快速推動行動。對大多數廣告商來說，這是一件好事。

Most advertisers aren't selling a $20,000 product. But if you're selling a $20,000 product, you're probably not going to sell it in a matter of a few hours. So on the one hand, because of our limited view on attribution, the model has to be really good at going ad to transaction in a very short window. If you can do that for a $200 shirt, that's a great outcome because the advertiser knows ad created actual intent and outcome. If that sale had happened two weeks later, it'd be very hard for the advertiser to say, I can justify the whole value of that sale back to the AppLovin channel, because there's been a whole -- a bunch of other things that have happened in the middle there.

大多數廣告商不會銷售價值 20,000 美元的產品。但如果你銷售的是價值 2 萬美元的產品，你可能不會在幾個小時內就賣掉。因此，一方面，由於我們對歸因的看法有限，該模型必須非常擅長在很短的時間內將廣告轉化為交易。如果您能為一件 200 美元的襯衫做到這一點，那將是一個很好的結果，因為廣告商知道廣告創造的實際意圖和結果。如果該銷售發生在兩週後，廣告商就很難說，我可以將該銷售的全部價值歸還給 AppLovin 管道，因為在這中間還發生了一系列其他事情。

What that does force us to do though is have a challenge when it comes to an advertiser that's selling a much more expensive product. We probably have to go and optimize for that advertiser to something earlier in funnel, maybe a phone call, maybe an e-mail registration but not all the way to the point of transaction if those transactions are basically impossible to drive in a matter of a few hours. Again, I remind you, we're hundreds of advertisers into this thing. So the world of advertisers here is so big. Most sell products pretty quickly.

然而，當廣告商銷售價格高得多的產品時，我們確實面臨挑戰。我們可能必須針對廣告主進行最佳化，使其在漏斗的早期階段就實現某些功能，例如電話呼叫、電子郵件註冊，但不會一直優化到交易階段，因為這些交易在幾個小時內基本上不可能完成。我再次提醒您，我們有數百家廣告商參與此事。所以這裡的廣告商的世界是如此之大。大多數產品都賣得很快。

So this honestly, isn't something that we're even thinking about right now because we have a lot of room to go get advertisers and grow this business before we get to the constraint of, okay, now we have to make the product sell that are in the thousands or tens of thousands of dollars per product.

所以說實話，這不是我們現在正在考慮的事情，因為我們有很大的空間去吸引廣告商並發展這項業務，然後我們才會受到限制，好吧，現在我們必須讓產品賣到每件數千或數萬美元。

Ralph Schackart, William Blair.

拉爾夫·沙卡特、威廉·布萊爾。

Just a question on self-service model. When you go live with that, obviously, you're testing right now, how do you expect the platform to sort of respond? Or maybe a better question is the advertisers, do you expect them to come with substantial budgets that you described a line out the door. Do you think it would be more like a dimmer switch where they would test and then add more budgets? Just trying to understand your best thoughts on what happens when you go live there.

我只是想問一下關於自助服務模式的問題。當您上線該功能時，顯然您現在正在測試，您預計平台會如何回應？或者更好的問題是，對於廣告商，您是否希望他們拿出您所描述的大量預算？您是否認為它更像是一個調光開關，他們會進行測試然後增加更多預算？只是想了解一下當你去那裡生活時你對會發生什麼的最佳想法。

Yes. Look, advertisers need to prove to themselves that the dollars are worth spending. So no one just comes and floods a new system. They usually build up budget over the weeks and months that they're live once they see that a campaign is actually achieving the goals that they have. Now remember, we've constrained the audience in web advertising to US only. That's by design. Our business itself is probably, I want to say off the top of my head, half-half US versus international. We don't operate inside China.

是的。瞧，廣告商需要向自己證明這些錢是值得花的。因此沒有人會突然闖入並淹沒一個新系統。一旦他們發現某項活動確實實現了他們的目標，他們通常會在活動的幾週和幾個月內增加預算。現在請記住，我們已將網路廣告的受眾限制在美國境內。這是設計使然。我想說的是，我們的業務本身可能一半是美國業務，一半是國際業務。我們不在中國境內開展業務。

So there's a lot of opportunity to just flip that switch and go global. Now we'll pair the self-service dashboard with that. That will allow us to really quickly dramatically increase the audience that sees these kinds of ads and pair it with the types of advertisers and count of advertisers that are onboarding into the system. Then it's up to them to find their own way. The good news is we're the last big advertising company that has a self-service dashboard.

因此，只要輕輕一按，就有很多機會走向全球。現在我們將自助服務儀表板與其配對。這將使我們能夠快速大幅增加看到此類廣告的受眾，並將其與加入系統的廣告商類型和廣告商數量相匹配。然後就由他們自己去尋找道路了。好消息是，我們是最後一家擁有自助服務儀表板的大型廣告公司。

All the other ones, the social companies, search, any display channel that's of any sort of scale has this type of ads manager product. Once we release it, agencies and advertisers will know what to do.

所有其他公司、社交公司、搜尋以及任何規模的展示管道都有這種類型的廣告管理產品。一旦我們發布它，代理商和廣告商就會知道該怎麼做。

Great. Maybe a question for Matt. Matt, I think, historically, you've talked about the growth algorithm being around 20% to 30%. Obviously, the business is growing very robustly. Is that still the right sort of framing for the model? Just kind of curious on your thoughts there.

偉大的。這可能是問馬特的一個問題。馬特，我認為，從歷史上看，你談到的成長演算法大約是 20% 到 30%。顯然，業務成長非常強勁。這仍然是該模型的正確框架嗎？只是有點好奇你的想法。

Yes. I mean we continue to believe that 20% to 30% is the right long-term growth rate for the company. As we've talked about before, that's comprised of two components. One is the ongoing reinforcement learning, right, from the model just transacting and then teaching itself and improving over time. And that component, we think, is about kind of 3% to 5%, which is how we're guiding on a quarterly basis.

是的。我的意思是，我們仍然相信 20% 到 30% 是公司合適的長期成長率。正如我們之前討論過的，它由兩個部分組成。一個是持續的強化學習，對吧，從模型開始進行交易，然後自我學習並隨著時間的推移而改進。我們認為，這個部分大約佔 3% 到 5%，這是我們按季度指導的。

That's the component that we know is very stable, very predictable. And so that's what we've included within our guidance. Above and beyond that are directed enhancements, right, models going from ChatGPT-three to four. The other engineering team are working on all the time. And when they launch a new model, those provide step function increases.

我們知道這個組件非常穩定，非常可預測。這就是我們在指導中包含的內容。除此之外還有定向增強，對吧，模型從 ChatGPT-3 到 4。其他工程團隊一直在努力工作。當他們推出新車型時，這些車型會提供階躍功能的成長。

And so those increase the annual growth rate in addition. Roughly around 10% is what we've kind of seen thus far, but we've had a lot of those quarters luckily where the engineers have launched those model enhancements, and we've seen those step function increases. So we think that we should be able to do at least one of those step function increases per year going forward. And it's a relatively conservative assumption from our perspective given where we've performed thus far over the past couple of years since launching.

因此，這些還會增加年增長率。到目前為止，我們看到的增幅大約在 10% 左右，但幸運的是，我們已經有很多季度在工程師推出了這些模型增強功能之後，我們看到了這些階躍函數的成長。因此我們認為，未來我們每年應該能夠至少實現一次階躍函數的成長。從我們的角度來看，考慮到我們自推出以來過去幾年的表現，這是一個相對保守的假設。

Matthew Thornton, FBN Securities.

馬修桑頓，FBN證券。

Maybe two if I could. I guess, first one, maybe for Adam. If we think about some of the next layer of initiatives and what I'm getting at there is really the nongaming app-to-app user acquisition as well as the dynamic personalized ad creative that you alluded to earlier. Should we think of those as initiatives that could actually start to contribute in 2025? Or should we think of those as more of 2026 type of initiatives?

如果可以的話也許兩個。我想，第一個，也許是給亞當的。如果我們考慮下一層級的舉措以及我所得到的，實際上是非遊戲應用程式到應用程式的用戶獲取以及您之前提到的動態個人化廣告創意。我們是否應該將這些視為真正能夠在 2025 年開始做出貢獻的舉措？或者我們應該把它們看作是 2026 年類型的舉措？

And then just one quick housekeeping follow-up. Matt, I think you guys have talked about the nongaming audience plus business maybe being 10% plus of total advertising revenue this year. Is that still how you're thinking about that business for 2025?

然後只需進行一次快速的後續跟進。馬特，我想你們已經討論過非遊戲受眾加上業務可能佔今年總廣告收入的 10% 以上。您對 2025 年的業務仍抱持這樣的想法嗎？

Yes. On the question on the new stuff that we're really prioritizing, I don't know if it's new, just expansion of what we're working on. I said in the talk script that focuses for this year to execute on that list. If we do that well, we're set up to have a fantastic 2026. We're already on the way to have a fantastic 2025.

是的。關於我們真正優先考慮的新內容的問題，我不知道它是否是新的，只是我們正在研究的內容的擴展。我在演講稿裡說過，今年的重點是執行那份清單。如果我們做得好，我們將擁有一個輝煌的 2026 年。我們已經踏上了實現美好 2025 年的旅程。

I don't know of any other tech company with the financial profile that we have and scale growing the way we are. I think it's on a rule of 150 or something. And what we're focused on when we talk about priorities is how is 2026 going to be. How is '27 going to be? If we execute on self-service dashboard, put it out in the market, automate media buying on our platform, make it simple for advertisers of any kind, let them come in, put in their goals, off they go, the business is going to have many quarters and years ahead of it of immense growth.

我不知道還有哪家科技公司擁有像我們這樣的財務狀況和規模成長速度。我認為這是 150 左右的規則。當我們談論優先事項時，我們關注的是 2026 年將會如何。27 年將會怎樣？如果我們執行自助服務儀表板，將其推向市場，在我們的平台上實現媒體購買自動化，讓任何類型的廣告商都可以輕鬆參與，讓他們加入，設定他們的目標，然後他們就可以開始，那麼業務將在未來許多季度和幾年內實現巨大的增長。

At some point, we're going to become a marketing-based business. We can run AppLovin ads first, any sort of small business to come in and buy on us, and you'll have an LTV to CAC model because the value of every one of those advertisers, given where we're starting in advertiser penetration to the millions that we can go after, would be immense. And then pair that with dynamic ad creative, continued model improvement, the self-learning in the system. And we've got a lot of vectors of growth that we're really excited about for the rest of the year and then beyond.

在某種程度上，我們將成為一家以行銷為基礎的企業。我們可以先執行 AppLovin 廣告，任何類型的小型企業都可以加入我們併購買我們的產品，您將擁有一個 LTV 到 CAC 模型，因為考慮到我們在廣告商滲透率方面的起步，我們可以追求的數百萬廣告商，每個廣告商的價值都是巨大的。然後將其與動態廣告創意、持續的模型改進和系統的自我學習相結合。我們擁有許多成長動力，我們對今年剩餘時間及以後的成長感到非常興奮。

And in terms of web-based advertising and contribution, Matt -- and we had mentioned previously, right, that we thought that they would become kind of 10% of the overall contribution in revenue -- it's very difficult to guess where that might go within the year because we've got a lot of factors, right, on both the existing mobile gaming business, how that develops over time based on these model enhancements that we're launching.

至於網路廣告和貢獻，馬特——我們之前提到過，對吧，我們認為它們將占到總收入的 10% 左右——很難猜測這一數字在一年內會達到什麼程度，因為我們有很多因素需要考慮，對吧，既包括現有的手機遊戲業務，也包括基於我們推出的這些模型增強功能，它將如何隨著時間的推移而發展。

To the extent that we launched significant model enhancements, that business could grow at a faster pace than e-commerce, but we're very optimistic about the e-commerce business after we launched the self-serve model. That business could grow quite significantly and outpace that 10% metric that we provided previously. So it's quite likely that it could represent a larger than 10% portion of the revenue this year.

在我們推出重大模型增強功能的範圍內，該業務的成長速度可能會比電子商務更快，但我們對推出自助服務模式後的電子商務業務非常樂觀。該業務可能會大幅成長，並超過我們之前提供的 10% 指標。因此，它很有可能佔今年收入的 10% 以上。

Matt, to follow up on what you said a couple of questions ago, I think last year, we had two quarters where there was, I think, the term is human-directed enhancements, right? So I think it was Q1 and Q3. And do I understand correctly that you're saying that, that can happen even like three, four, five years into AXON 2? It's not a function of the life stage of the machine. It's a function of it's constantly learning.

馬特，為了跟進你幾個之前提到的問題，我想去年我們有兩個季度出現了「人類主導的增強」這個術語，對嗎？所以我認為是 Q1 和 Q3。我是否理解正確？您說的是，即使在 AXON 2 推出三、四、五年後，這種情況仍有可能發生？這不是機器生命階段的功能。這是它不斷學習的功能。

And then a related question. So help us understand, well, at least me, is it the same learning that's applied to gaming and e-commerce. Or are those two different processes that can happen, let's say, in one quarter, like e-commerce, learn something and stepped up and gaming did not?

然後是一個相關問題。所以請幫助我們理解，至少對我來說，這是適用於遊戲和電子商務的相同學習。或者說，這兩個不同的過程可能發生在一個季度內，例如電子商務學到了一些東西並取得了進步，而遊戲卻沒有？

Yes. Yes, those are great questions. I'll address the second one first. So there are two different -- two separate models. So both models are learning at the same time.

是的。是的，這些都是很好的問題。我先解決第二個問題。因此，有兩種不同的——兩種獨立的模型。因此兩個模型同時在學習。

So you've got reinforcement learning occurring to both models, and then you also have directed enhancements that the team could launch on either model. Obviously, we've got teams -- we have a team that's working on both models at the same time, similar to what Adam mentioned about our business development team. We run lean, and so we have a very small team that's working across both.

因此，你可以在兩個模型上進行強化學習，然後你也可以指導團隊在任一模型上啟動增強功能。顯然，我們有團隊——我們有一個團隊同時研究這兩種模式，類似於亞當提到的我們的業務開發團隊。我們實行精益運營，因此我們有一個非常小的團隊來同時開展工作。

But they're constantly testing potential changes to the model that we could launch on either to provide these step function increases at any point in time. And there's a laundry list of these things that they're testing at any given point in time to assess whether or not they could potentially have an impact.

但他們一直在測試模型的潛在變化，我們可以在任何時間點啟動這些階躍函數的增加。他們隨時都會對一系列事項進行測試，以評估它們是否可能產生影響。

And then we launch those model enhancements. So you could have the mobile gaming, the existing mobile gaming model improve in one quarter and the e-commerce model only have reinforcement learning and that more kind of stable growth, plus adding new advertisers there, which is obviously going to increase at a faster pace than the mobile gaming side of the business where we basically have all mobile gaming advertisers at this point. So those two things are moving independently.

然後我們推出這些模型增強功能。因此，您可以擁有行動遊戲，現有的行動遊戲模式在一個季度內得到改善，而電子商務模式只有強化學習和更穩定的成長，再加上增加新的廣告商，這顯然會比行動遊戲業務的成長速度更快，目前我們基本上擁有所有的行動遊戲廣告商。所以這兩件事是獨立進行的。

Yes. Vasily, I would add a couple of things, too, is, let's not forget this form of understanding how to work with neural nets is really new. I mean you've seen it like we all can play with it with ChatGPT. If you look at how quickly they're releasing models, there's a lot of research in the space that is guiding engineers across companies that have gotten very good with models to build new versions of the models. And as you build new versions of the models, you train them on the data, if they're more predictive, they'll be improved.

是的。瓦西里，我還要補充幾點，那就是，我們不要忘記這種理解如何使用神經網路的形式確實是新穎的。我的意思是，您已經看到它了，就像我們都可以使用 ChatGPT 來玩它一樣。如果你看看他們發布模型的速度，你會發現該領域的大量研究正在指導那些已經非常擅長模型的公司的工程師建立新版本的模型。當你建立新版本的模型時，你會根據資料對它們進行訓練，如果它們的預測能力更強，它們就會得到改進。

The beauty of our business model is that, one, we're cutting edge. We're on a lot of the latest research. We built an exceptional implementation of this kind of technology. We may be one of the best examples of how software can unlock value from this kind of AI movement, but the reality is it's really, really early. So there's plenty of research out there that our team still needs to parse through.

我們的商業模式的優點在於：第一，我們處於領先地位。我們正在進行許多最新研究。我們建構了此類技術的卓越實現。我們可能是軟體如何從這種人工智慧運動中釋放價值的最佳例子之一，但事實上，這還處於非常非常早期的階段。因此，我們的團隊仍需要分析大量的研究成果。

There's internal research, so we pair with that. And this is not branding. I don't see any end in sight for us as human beings to be working on these types of AI technologies. The opportunity is just too great. And so we probably shouldn't call AXON 2, AXON 2.

有內部研究，所以我們與之配對。這並不是品牌推廣。我認為人類對這些類型的人工智慧技術的研究不會有任何終點。這個機會實在太好了。所以我們可能不應該把 AXON 2 稱為 AXON 2。

Maybe it's just the AXON model. There's going to be many different iterations of this thing that we go through in the coming quarters and years. And every single time when we have a lift, something that's material, where we can step it up, where you would say like AXON 3 -- not AXON 3, [AXON 2.2, AXON 2.3, AXON 2.4], those incremental changes have shown to be double-digit growth in a single quarter quarter-over-quarter.

也許只是 AXON 模型。在未來的幾個季度和幾年裡，我們將經歷許多不同的迭代。每當我們有提升，有實質的進步，我們就可以加大力度，就像 AXON 3——不是 AXON 3，[AXON 2.2、AXON 2.3、AXON 2.4]，這些漸進式變化已證明單季環比增長達到兩位數。

If we have a less incremental lift but a nominal enhancement where the team is still working on improving the model but it doesn't step up as much, it's still very impactful to our business at the scale that we operate at. So we have a long pipeline of things to work on, and we expect that the research science team to land a time is going to be constantly improving this model.

如果我們的增量提升較少，但有名義上的增強，即團隊仍在努力改進模型，但提升幅度不大，那麼對於我們營運規模的業務來說，它仍然會非常有影響。因此，我們有很多工作要做，我們希望研究科學團隊能夠找到時間並不斷改進這個模型。

So a very quick follow-up. So in simple terms, what Matt was saying that these are two different models, right, so is it possible that you have AXON 2 running and then there is a '22 level of event that happens with the e-commerce model, right? And it steps up in efficiency. And they will be not synchronized but improving at a step function at different -- is that what you're saying?

因此，後續行動非常迅速。所以簡單來說，馬特說這是兩種不同的模式，對吧，那麼有沒有可能你運行 AXON 2，然後在電子商務模式下發生 '22 級別的事件，對吧？且其效率也不斷提高。而且它們不會同步，而是以不同的階躍函數進行改進——這是您所說的嗎？

Yes. It's the reason why I keep saying e-com model is early, is exactly that. It's like when we launched AXON two for gaming. We're multiple iterations in now. So the teams had a lot more time to evolve the gaming model.

是的。這就是我一直說電子商務模式還處於早期階段的原因。這就像我們為遊戲推出的 AXON two 一樣。我們現在已經進行了多次迭代。因此團隊有更多的時間來改進遊戲模型。

The e-commerce model is a couple of quarters live now. It's not -- it hasn't been around for that long. It also doesn't have as much of a data feedback loop because it doesn't have as much impressions, transactions, scale to then retrain itself off of as well. So you have two very strong levers there that are coming.

電子商務模式目前已實施了幾個季度。它並不是──它存在的時間並不長。它也沒有那麼多的資料回饋循環，因為它沒有那麼多的印象、交易和規模來重新訓練自己。因此，您擁有兩個非常強大的槓桿。

Clark Lampen, BTIG.

克拉克·蘭彭（Clark Lampen），BTIG。

Adam, as we are sifting through sort of Q1 performance for the Advertising segment within the sort of 20%-ish sequential growth that you referenced, I'm curious if that -- sort of as Matt put it before, did that include some benefit from step function rather than sort of reinforcement learning type improvements, i.e. sequential growth for gaming was probably ahead of the typical low to mid-single digits rate of growth?

亞當，我們正在篩選廣告部門第一季的表現，正如您提到的 20% 左右的連續增長，我很好奇——就像馬特之前說的那樣，這是否包括階躍函數帶來的好處，而不是強化學習類型的改進，也就是說，遊戲的連續增長可能領先於典型的低到中等個位數的增長率？

Yes, for sure. So look, you have two less days. Q1 even though it's a good season with some of those holiday days, it's not, compared to Q4, a good season. So it was a huge step-up quarter. I mean, again, we might be the fastest growing tech company anywhere at the scale that we operate at, so there have to be multiple drivers there.

是的，當然。所以你看，你少了兩天。儘管第一季度由於有一些假期而表現不錯，但與第四季度相比，這並不是一個好季度。因此這是一個巨大的成長季度。我的意思是，就我們營運的規模而言，我們可能是成長最快的科技公司，因此必須有多種驅動因素​​。

And I mentioned on the talk script that there was better efficiency in the model, so the team is constantly improving the model. Now we don't call it like with self-directed lift or a model change. This wasn't a material change to the model. But the team found wins in the model, and it became more effective. Self-learning is an aspect of it.

而且我在演講稿裡提到，這個模型的效率更高，所以團隊不斷改進這個模型。現在我們不稱之為自我導向提升或模型改變。這並不是對模型的實質改變。但研究團隊發現該模型有其優勢，變得更加有效。自學是其中的一個面向。

And I did say the advertiser contribution from e-com getting a full quarter of the run rate that it had gotten to in Q4 was also impactful. And I think I said it was less impactful than the gaming side.

我確實說過，電子商務廣告商的貢獻達到了第四季度的整整四分之一的運作率，這也是有影響的。我認為我說過它的影響比遊戲方面要小。

So with that, you can infer the greater than 50% of the quarter's growth came from the gaming side. There's still a lot of room to go there. I mean these advertisers, we become the channel that they depend on. Both the advertisers and the publishers need us to do our job. We're doing it very effectively, but we're still at a very low conversion rate per 1,000 impressions to where we think we can get to even in the gaming category.

因此，你可以推斷本季超過 50% 的成長來自遊戲方面。那裡還有很大的發展空間。我的意思是，對於這些廣告商來說，我們成為了他們所依賴的管道。廣告商和出版商都需要我們做好自己的工作。我們做得非常有效，但每千次展示的轉換率仍然很低，我們認為即使在遊戲類別中我們也無法達到這個水平。

And so the technology is still early. We're going to find more lifts. The team's got a long pipeline. So we're really excited about both the opportunity for the model to keep getting better in that flywheel effect because the scale has gotten so big paired with what the team is going to be able to unlock as they make the model better.

因此這項技術還處於早期階段。我們會尋找更多的電梯。團隊擁有很長的人才輸送管道。因此，我們對模型在飛輪效應中不斷改進的機會感到非常興奮，因為規模已經變得如此之大，而團隊在改進模型時將能夠解鎖新的功能。

Okay. That makes a lot of sense, and it's helpful color. Maybe for a second question, just a quick housekeeping one. Does -- maybe this is for Matt. Does the guidance for the second quarter include any of the upside or recognition from legacy studio spend?

好的。這非常有意義，而且很有幫助。也許還有第二個問題，只是一個簡單的問題。也許這是給馬特的。第二季的預期是否包括了傳統工作室支出的任何好處或認可？

I believe way back when you guys disclosed a sort of total software transaction value metric and for accounting purposes, you weren't recording revenue from first-party studios. Now that they're third party, I would imagine that there probably will be some transition so long as they're still active. Is that in the numbers? Or is it material at this point?

我相信很久以前，當你們披露某種總軟體交易價值指標並且出於會計目的時，你們並沒有記錄來自第一方工作室的收入。既然他們是第三方，我想只要他們仍然活躍，就可能會有一些轉變。這是數字嗎？或者說，目前為止，它已經是實質的了？

Yes. So after signing the agreement today with Tripledot to divest the Apps business, our expectation is that the transaction's going to close towards the end of the quarter, actually, Clark. So within the guidance, we have not assumed any incremental uplift from having those studios be external parties and the premium rate on the user acquisition cost that we would normally charge.

是的。因此，在今天與 Tripledot 簽署剝離應用程式業務的協議後，我們預計交易將在本季末完成，克拉克。因此，在指導範圍內，我們並未假設這些工作室作為外部方會帶來任何增量提升，也未假設我們通常會收取的用戶獲取成本的溢價率。

That being said, it will obviously depend on how Tripledot runs those businesses as to whether or not they will continue to spend at the same level. And it's not a material impact on the business one way or the other. It's a nice premium to have, but we don't expect that it will move the needle.

話雖如此，但這顯然取決於 Tripledot 如何經營這些業務，以及他們是否會繼續維持相同的支出水準。無論如何，它都不會對業務造成實質影響。這是一個不錯的溢價，但我們不認為它會起到什麼作用。

Alec Brondolo, Wells Fargo.

富國銀行的亞歷克·布朗多洛。

Adam, if you can maybe elaborate on the App Store regulatory news, some of the fee relief that mobile games might enjoy as we move through the year, I guess, one, how do you think about the potential impact on the business from an ad spend perspective? And then, two, how do you think about positioning the business to maybe assist in that transition, position yourself best to win some of that incremental spend to the extent that it materializes?

亞當，您能否詳細說明 App Store 監管新聞，以及今年手機遊戲可能會享受到的一些費用減免，我想，首先，您如何看待從廣告支出角度對業務的潛在影響？然後，第二，您認為如何定位業務以協助實現這一轉變，如何使自己處於最佳位置以贏得部分增量支出？

Yes. So look, we're the largest channel there is today for mobile gaming customers to spend their money, and I think we're the best destination for it. So maybe we're likely to be the biggest beneficiary of this change. And just in simple terms, if you assume the fee isn't going to go to zero, so let's just, for my example, assume you go to half, 15% fee, you take $1 that the customer was charging -- the mobile game developer was charging a consumer, before, they were getting $0.70. And then they would plug that into a ROAS model with us to go spend dollars.

是的。所以，看，我們是當今行動遊戲客戶消費的最大管道，我認為我們是最好的目的地。所以也許我們可能會成為這項變化的最大受益者。簡單來說，如果你假設費用不會降為零，那麼，舉個例子，假設費用減半，即 15%，你從客戶那裡收取 1 美元——之前，手機遊戲開發商向消費者收取的費用是 0.70 美元。然後他們會將其插入到我們的 ROAS 模型中去花錢。

If that $0.70 and a $0.15 fee structure becomes $0.85, they make 20% more. It's an immediate 20% uplift to their revenues. Now 20% in some portion or in total is going to come back to a marketing platform that's as scaled as ours. And remember, we run a dynamic auction. So if one customer says, "Hey, I want to bid 20% more because I bank 20% more.

如果 0.70 美元加上 0.15 美元的費用結構變成 0.85 美元，他們的利潤就會增加 20%。這使他們的收入立即增加了 20%。現在，其中 20% 的部分或全部將會回到像我們這樣規模的行銷平台。請記住，我們進行的是動態拍賣。因此，如果一位顧客說：「嘿，我想多出價 20%，因為我的存款多 20%。

I like the profit margins I run at. I want to get more growth," then every other customer has to do the same thing because it's competitive auction. So over time, what's going to happen is it's really good for the ecosystem. It will create more growth, and it's exceptionally good for the advertising businesses because the dollars will come into the marketplace and allow the advertisers to bid more competitively, which benefits us, and that feeds back to the publisher. It also feeds back obviously to our bottom line and benefits our shareholders.

我喜歡我所經營的利潤率。我希望獲得更多成長”，那麼其他每個客戶都必須做同樣的事情，因為這是競爭性拍賣。因此，隨著時間的推移，這將對生態系統產生非常大的好處。它將創造更多的成長，這對廣告業務來說尤其有利，因為資金將進入市場，使廣告商能夠更有競爭力地競標，這對我們有利，並且會回饋給出版商。這也明顯地反饋到我們的底線並使我們股東受益。

Jim Callahan, Piper Sandler.

吉姆·卡拉漢、派珀·桑德勒。

Big investor question we get is just your ability to sort of monetize relatively like static base of inventory. Now that we're a couple of quarters into sort of launching e-com, would be curious your sort of thought or if that's changed at all.

我們收到的一個投資者大問題是，您是否有能力將庫存相對靜態地貨幣化。現在我們已經進入電子商務的幾個季度了，我很好奇您的想法，或者這是否有所改變。

Well, look, we have a very large-scale static base of inventory. I'd say, if you look at Meta's business or YouTube, these are static bases of inventory. It's not like there's new users pouring in, especially in the States. But what's happening across the companies that are doing well in performance advertising is the matching and the technology algorithm is getting stronger. We're starting at a very, very low place.

嗯，你看，我們有一個非常大規模的靜態庫存基礎。我想說，如果你看看 Meta 的業務或 YouTube，這些都是靜態的庫存基礎。並不是說有新用戶湧入，尤其是在美國。但在效果廣告方面表現良好的公司正在發生的事情是匹配和技術演算法正在變得越來越強大。我們的起點非常非常低。

I think in the past, we've said, look, we're driving a 1% transaction right now. We've grown a lot, so that's probably gone up a little bit. It doesn't have to go up a lot for our business to expand dramatically. We think that can go up to become 2%, 3%, 4%, 5% over time. The odds are very, very impactful.

我想，過去我們曾說過，看，我們現在正在推動 1% 的交易。我們已經成長了很多，所以這可能已經有所上升。我們的業務並不需要大幅成長就能大幅擴張。我們認為隨著時間的推移，這一比例可以上升到 2%、3%、4%、5%。這個幾率非常非常大。

It's full-screen video. It captures the users' attention. And the more advertiser demand we get, as we open up our self-service platform and hopefully bring on thousands, tens of thousands eventually, hundreds and thousands of millions of customers, we're going to have more content to show the end consumer. We're going to pair that with personalized ad creative. The consumer response is going to go up.

這是全螢幕影片。它吸引了用戶的注意。隨著我們開放自助服務平台，並希望最終帶來數千、數萬、數十億的客戶，我們獲得的廣告商需求越多，我們將有更多的內容展示給最終消費者。我們將把它與個人化廣告創意結合。消費者的反應將會上升。

That's going to help us. It's going to help our publishers. It's going to help our partners, and we think it's going to catalyze potentially years and decade plus of growth.

這對我們有幫助。這將幫助我們的出版商。它將幫助我們的合作夥伴，我們認為它將催化未來數年甚至數十年的成長。

Got it. That is helpful. And then just a quick one on gaming macro. Obviously, not really seeing any signs of weakness in the results. But anything to call out with regard to like studio launches or geographies that are worth noting?

知道了。這很有幫助。然後簡單介紹一下遊戲巨集。顯然，從結果中並沒有看到任何疲軟的跡象。但是，關於工作室的推出或地理位置，有什麼值得注意的嗎？

I mean our business is really diverse at this point. One game was never going to move the needle at the scale that we operate at, so there are -- I guess, there have been some bigger launches or marketing in the ecosystem around a couple of games but nothing dissimilar to the past history of the pace of game launches. I don't think the tariffs impact the digital economy here. And in terms of the economy and slowdowns, free-to-play gaming is a very cheap and accessible form of entertainment. So we've always felt that it's quite insulated from economic change or distress.

我的意思是，目前我們的業務確實非常多元化。一款遊戲永遠不可能在我們營運的規模上產生影響，所以——我想，在生態系統中圍繞幾款遊戲進行過一些更大規模的發布或行銷，但與過去的遊戲發布速度沒有什麼不同。我認為關稅不會影響這裡的數位經濟。就經濟和經濟放緩而言，免費遊戲是一種非常便宜且易於獲得的娛樂形式。因此，我們一直認為它不太容易受到經濟變化或困境的影響。

And so we're confident with the position we have in the market.

因此，我們對自己在市場上的地位充滿信心。

Martin Yang, OpCo.

馬丁楊，OpCo。

A couple of questions on the self-serve dashboard. First, has your view changed on who can have access first to the self-serve dashboard? Is it your existing customers primarily? Or are you applying this to anyone regardless of their annual spend budget?

關於自助服務儀表板的幾個問題。首先，您對誰可以先存取自助服務儀表板的看法是否發生了變化？主要是您的現有客戶嗎？或者您將此應用於任何人，無論他們的年度支出預算是多少？

Yes. We're going to roll it out in stages, Martin. So right now, it's current customers. We'll have a feedback group. Once it gets past feedback group, it will open up to all current customers.

是的。我們將分階段推出，馬丁。所以現在，它是現有客戶。我們將有一個回饋小組。一旦通過反饋小組，它將向所有現有客戶開放。

The goal with that is to just help our team. The team right now is do a bunch of manual labor, and we're capped -- we're resource limited. So that's going to happen pretty quickly here over the coming weeks. I mean I mentioned it in my talk script. We're getting this in the hands of clients now.

這樣做的目的只是為了幫助我們的團隊。團隊現在要做大量的體力勞動，而且我們的資源有限。所以這將在未來幾週內很快發生。我的意思是我在我的演講稿中提到過它。我們現在將其交到客戶手中。

So that's going to be short term. In the medium to long term, we'll go through stages of opening it up, where eventually long term, and this isn't too long term but several quarters, we'll go completely open. But in between now and there, you'll have stages of opening it up to new types of clients so that we ensure the quality of client is high. The platform can be bug free. It can work under all scenarios.

所以這只是短期的。從中長期來看，我們將經歷開放的階段，最終從長期來看，雖然不是太長期，但經過幾個季度，我們將完全開放。但從現在到那時，我們將逐步向新類型的客戶開放，以確保客戶的高品質。該平台可以無錯誤。它可以在所有場景下工作。

The models have time to improve. We open it up and off we go.

模型還有時間可以改進。我們打開它然後出發。

Got it. And a follow-up on that is when you think about this potential to give you more data or maybe increase your capacity to process more, do you think near term, this is -- could be a near -- helpful near-term boost to how you improve the performance of web app business?

知道了。接下來的問題是，當您考慮到這種潛力可以為您提供更多數據或可能提高您的處理能力時，您是否認為在短期內，這是否會對您如何提高 Web 應用程式業務的效能產生近期有益的推動作用？

Look, the more customers we get, the better the model will get. Just if you think about it, just logically, if you've got 10 beauty companies live or you have 1,000 beauty companies live, you're going to be better in the category if you have 1,000 live than 10 because you have more diversity of product to show the end consumer. This type of a product is a recommendation engine. Give it more options to recommend something personalized, and it's going to, by effect, get better. Now I've said this in the past.

你看，我們獲得的客戶越多，模型就會越好。如果你仔細想想，從邏輯上講，如果你有 10 家美容公司或 1,000 家美容公司，那麼如果你有 1,000 家公司，那麼你在這個類別中的表現會比 10 家公司更好，因為你有更多樣化的產品可以展示給最終消費者。這種類型的產品是推薦引擎。給它更多的選項來推薦個人化的東西，它的效果就會變得更好。我以前已經說過這​​個了。

We're at a point where we're not perfect. We're not making it work for every customer, whereas we do that inside gaming. So we don't want to go to 1,000 beauty companies and say, "Look, 800 of you are going to work perfectly well. 200 of you are not going to work." Our shareholders will be ecstatic. If we open up the platform and become thousands to tens of thousands to hundreds of thousands of customers, the business is going to be multiples bigger than it is today.

我們目前還不完美。我們不會讓它適用於每個客戶，但我們在遊戲內部做到了這一點。所以我們不想去 1,000 家美容公司然後說，“看，你們中的 800 家會做得很好。但你們中的 200 家卻不會成功。”我們的股東將會欣喜若狂。如果我們開放平台並擁有數千、數萬甚至數十萬的客戶，那麼業務規模將比現在擴大數倍。

The growth rate is going to be phenomenal. The problem is we want to make it work for everyone. We really set a high bar of perfection in anything that we do. We want to build the best performance product that these customers have ever seen. And so we're not there yet.

成長率將會非常驚人。問題是我們想讓它為每個人服務。我們對所做的每一件事都設定了極高的完美標準。我們希望打造這些客戶所見過的性能最佳的產品。所以我們還沒有到達那裡。

We're going to get there. When we get there, we'll open it up, and that's going to catalyze a lot of growth for the long term.

我們將會到達那裡。當我們到達那裡時，我們就會開放它，這將催化長期的大量成長。

Eric Sheridan, Goldman Sachs.

高盛的艾瑞克·謝裡丹。

Maybe two if I could. Adam, first for you. I got a chance to look at the blog post about TikTok, and I totally get your labeling it as a bit of a long shot, to use your phraseology from the end of the blog. But my understanding was you were more interested in the US ring-fenced process that's going on under the Trump administration now, and this seems like it's a bit of a broader proposal ex China globally as opposed to maybe what the Trump administration is trying to solve for.

如果可以的話也許兩個。亞當，首先請你發言。我有機會看了關於 TikTok 的部落格文章，用部落格末尾的措辭來說，我完全明白你將其標記為有點長遠的想法。但我的理解是，你對川普政府目前正在進行的美國專案保護進程更感興趣，這似乎是一個在全球範圍內除中國之外的更廣泛的提議，而不是川普政府正在試圖解決的問題。

So I just want to make sure I understand a little bit of the messaging off the blog and how to reconcile it with what is going on in D.C. right now with respect to TikTok US and how you're positioning yourself vis-a-vis that.

因此，我只是想確保我理解部落格上的一些信息，以及如何將其與目前華盛頓特區關於 TikTok US 的事件以及你對此的定位相協調。

And then, Matt, you guys continue to produce a lot of efficiencies while growing at very high rates. Any updated views on how you guys are thinking about incremental margins in the business over the medium to long term or steady-state rate on incremental margins?

然後，馬特，你們在以非常高的速度成長的同時，繼續創造許多效率。你們對於中長期業務的增量利潤率或增量利潤率的穩定狀態有何最新看法？

Thanks, Eric. I don't want to spend too much time on TikTok, but just briefly, yes, we're going after world outside of China. We think the biggest priority is to solve national security concerns with regards to biases in the algo and data in the US But that's important outside of China everywhere. And to do that, you really truly do have to have operational control.

謝謝，埃里克。我不想在 TikTok 上花太多時間，但簡單來說，是的，我們正在追求中國以外的世界。我們認為，當務之急是解決美國在演算法和數據方面的偏見所帶來的國家安全問題，但這在中國以外的任何地方都很重要。要做到這一點，你確實必須擁有操作控制權。

Unless the US app is a separate app from the rest of the world, which it cannot be, would just deflate the app's productivity, then some company needs to have control to rewrite parts of the algo and to ensure that this thing complies with the standards around national security. We think our proposal is best suited to do that. In fact, we think we may be the only one who could do that with the knowledge that we have of models and the proposal we put forth. And on top of that, we think we would solve the present and be the best operator of this business long term through this type of partnership structure we propose.

除非美國的應用程式是一個與世界其他地方不同的應用程序，但這不可能，這只會降低應用程式的生產力，然後一些公司需要控制重寫部分演算法並確保這個東西符合國家安全的標準。我們認為我們的提議最適合實現這一點。事實上，我們認為，憑藉我們掌握的模型知識和提出的建議，我們可能是唯一能夠做到這一點的人。最重要的是，我們認為透過我們提出的這種合作結構，我們可以解決當前的問題，並長期成為該業務的最佳營運商。

So I put the thoughts out there. I wanted people to have context. It's absolutely a long shot. We do go after long shots. That shouldn't be surprising to our shareholders at this point.

所以我把這些想法說了出來。我希望人們了解背景。這絕對是一個不可能實現的目標。我們確實在追求遠射。對於我們的股東來說這應該不會感到驚訝。

We think we've got the best performance advertising AI model the world has ever seen today. And that paired with an audience as large scale as TikTok's could unlock immense economic potential and value for shareholders. So it's something that we're excited about, and I put the rest of the details in the blog for others to read.

我們認為我們擁有當今世界上表現最佳的廣告 AI 模型。再加上 TikTok 如此龐大的受眾群體，可以為股東釋放巨大的經濟潛力和價值。所以這是我們感到興奮的事情，我把其餘的細節放在部落格中供其他人閱讀。

And in terms of your second question, Eric, for the margin profile of the business and expectations going forward, I mean, as Adam mentioned, we continue to expect to run the business lean, and it's part of the ethos of the company. So as we continue to grow top line, we're trying to find ways to do that in an efficient manner. Obviously, we've been talking about the self-serve platform. That's one. Obvious way that we're planning to grow by not adding any to the cost base.

關於你的第二個問題，艾瑞克，對於業務的利潤狀況和未來的預期，我的意思是，正如亞當所提到的，我們繼續期望精益經營業務，這是公司精神的一部分。因此，隨著我們繼續增加營業收入，我們正在嘗試尋找有效的方式來實現這一目標。顯然，我們一直在談論自助服務平台。那是一個。顯然，我們計劃透過不增加任何成本來實現成長。

So we do expect that the primary components of the cost base should remain relatively flat with obviously the exception of data center costs, which are more variable. What we've seen thus far as we grow revenue is that we've been adding on data center costs on an annual basis at least around kind of a 10% of the overall revenue growth, and we expect that, that should continue. So margin -- the margin profile of the business should continue to grow from where we're at today on the Advertising business up to that kind of asymptote. And then we'll see where we can take it from there.

因此，我們確實預期成本基礎的主要組成部分應該保持相對平穩，但顯然資料中心成本除外，因為資料中心成本的變化更大。到目前為止，我們看到，隨著收入的成長，我們每年增加的資料中心成本至少佔總收入成長的 10% 左右，我們預計這種情況應該會持續下去。因此，利潤率－廣告業務的利潤率狀況應該會從目前的水平繼續增長，直至達到那種漸近線。然後我們再看看能從那裡得到什麼。

Rob Sanderson, Loop Capital.

桑德森（Rob Sanderson），Loop Capital。

Can you hear me this time?

這次你聽得到我說話嗎？

We can, Rob, yes.

是的，羅布，我們可以。

Okay. Sorry. Apologies for the technical difficulties. And apologies for not being on camera because that's where I failed last time, so we're going to play it safe. I have two questions, please.

好的。對不起。對於技術問題，我們深表歉意。很抱歉沒有出現在鏡頭中，因為上次我失敗了，所以我們會謹慎行事。我有兩個問題請問。

Can you -- first, can you help us understand maybe and delineate, to the extent that you can, sort of trends in the app product versus web? And the lines obviously are really blurring here. And Matt, when you talk about potential for 10% contribution from web this year, like are you talking gross or net? Or how do you even parse out the difference as there's -- you're obviously serving a lot of ads on inventory you'd be buying otherwise. And then I've got a follow-up as well.

首先，您能否盡可能地幫助我們理解並描述應用程式產品與網路之間的趨勢？顯然，這裡的界線已經變得模糊了。馬特，當您談到今年網路貢獻 10% 的潛力時，您指的是總額還是淨額？或者你如何分析出其中的差異——你顯然在你本來會購買的庫存上投放了大量廣告。然後我也會進行後續跟進。

Yes, Rob, I'll go with the first. Matt, can go with the second. Look, over time, this is just content marketed through our platform using our models to drive a result. So whether the advertiser is an app, a website, a web in an app, the model should be able to deliver the result. Now we started with mobile gaming advertisers who had apps.

是的，羅布，我選擇第一個。馬特，可以選擇第二個。瞧，隨著時間的推移，這只是透過我們的平台利用我們的模型來行銷內容以推動結果。因此，無論廣告商是應用程式、網站還是應用程式中的網頁，該模型都應該能夠提供結果。現在我們開始與擁有應用程式的行動遊戲廣告商合作。

And so the first version of our business for 12 years was focused on one specific use case. Now we built into the web use case. Now we're also building into the web and app use case. But over time, the way we envision our product is that advertisers come in. They say, "Here's where I want the user to go.

因此，我們 12 年來業務的第一個版本專注於一個特定的用例。現在我們將其融入網路用例中。現在我們也正在建立網頁和應用程式用例。但隨著時間的推移，我們設想我們的產品會吸引廣告商的注意。他們說，「這就是我希望用戶去的地方。

Here's my app. Here's my website. Here's the return on ad spend target that I want or the cost per purchaser that I want. Here's my budget. Upload a few videos.

這是我的應用程式。這是我的網站。這是我想要的廣告支出回報目標或我想要的每位購買者的成本。這是我的預算。上傳一些影片。

AppLovin go take care of the rest, deliver result, create your own videos, personalize the advertising. And then I'm really happy." And so we think all of this stuff is going to merge together as we get to a place quarters or years from here.

AppLovin 負責處理其餘事務、交付結果、建立您自己的影片、個人化廣告。然後我真的很高興。 」因此我們認為，當我們在幾個季度或幾年後到達某個地方時，所有這些東西都會融合在一起。

And to clarify on the 10% contribution, Rob, so that would be on a net basis. So we believe that the web-based advertising solution should contribute at least a 10% to the overall net revenue of the advertising solution.

羅布，請澄清一下 10% 的貢獻，那將是淨額基礎。因此我們認為基於網路的廣告解決方案應該為廣告解決方案的整體淨收入貢獻至少 10%。

Okay. And then a question on sort of longer-term avenues for self-attribution. It's sort of -- it's a little bit of a hole right now for you just not having the identity and whatnot. But are you always going to be sort of dependent on third parties for this? Or are you limited to just maybe lower consideration, high turnover products like you described earlier, Adam?

好的。然後是關於自我歸因的長期途徑的問題。這對你來說就像是個漏洞，因為你沒有身分等等。但您是否總是會依賴第三方來實現這一點？還是您僅限於像您之前描述的一樣考慮度較低、週轉率較高的產品，亞當？

Or is it sort of you just got lots of demand to serve right now and years of growth in front of you and you'll sort through that challenge over time? What can you share on that at this point?

或者您現在需要滿足大量需求，並且未來幾年將迎來成長，隨著時間的推移，您會逐步克服這些挑戰？現在您能分享什麼呢？

Yes. So look, the third is true. We've got lots of demand and opportunity in front of us. The products that we can service on app are attributing with third-party companies like AppsFlyer and Adjust, so we were built for third-party attribution on that. On web, we built the product to be self-attributing to our own attribution platform.

是的。所以看，第三個是正確的。我們面前有很多需求和機會。我們可以在應用程式上提供服務的產品是與 AppsFlyer 和 Adjust 等第三方公司合作的，因此我們是為第三方歸因而建立的。在網路上，我們建構了可以自我歸因於我們自己的歸因平台的產品。

And it's not high turnover products. I mean, look, most products in the world are not selling something greater than $250. Our product -- our models can go deliver something that's a couple of hundred dollars within a few minutes of the ad being seen, and it's happening quite often, I mean, obviously, scale, the $1 billion run rate that I mentioned. So that's something where we don't get a long window today, but it's not harming us. In fact, it's forcing us to work at a higher level.

而且它不是高週轉產品。我的意思是，世界上大多數產品的售價不會超過 250 美元。我們的產品——我們的模型可以在廣告被看到後的幾分鐘內提供幾百美元的收入，而且這種情況經常發生，我的意思是，顯然是規模，我提到的 10 億美元的運行率。所以，這是我們今天沒有很長的時間來解決的問題，但這並沒有對我們造成傷害。事實上，它迫使我們在更高層次上開展工作。

And that standard makes it so that our product is delivering a specific value in a short window, which allows the advertiser to see more incrementality on the dollars that they spend and question attribution less. Now over time, though, we want our advertisers to be able to measure us based on the way that they want to measure us. And companies use different attribution tools. Some of them have internal tools. So the same way on mobile app, we integrated with third-party solutions, we are working on integrations in the web space as well so that the advertisers can have multiple reference points.

這個標準使我們的產品能夠在短時間內提供特定的價值，從而使廣告商能夠看到他們所花費的資金的更多增量，並減少對歸因的質疑。然而，隨著時間的推移，我們希望廣告商能夠以他們想要衡量我們的方式來衡量我們。而且各公司使用不同的歸因工具。其中一些有內部工具。因此，就像在行動應用程式上我們與第三方解決方案集成一樣，我們也在致力於網路空間的集成，以便廣告商可以擁有多個參考點。

They can feel confident in the dollars that they spend as they scale us. And our aspirations are to become their biggest or their second biggest channel, period. And if you become that big, you want them to have a lot of confidence in the dollars that they're spending. That's it. Thanks, everyone, for joining us today.

隨著我們規模的擴大，他們可以對自己所花的錢充滿信心。我們的願望是成為他們最大或第二大的管道，僅此而已。如果你變得那麼大，你希望他們對自己所花的錢有足夠的信心。就是這樣。感謝大家今天加入我們。