AppLovin Corporation (APP) 2026 Q3 法說會逐字稿

Welcome to AppLovin's Earnings Call for the third quarter ended September 30, 2026. I'm David Hsiao, Head of Investor Relations. Joining me today to discuss our results are Adam Foroughi, our Co-Founder and CEO; and Matt Stumpf, our CFO. Please note, our SEC filings to date as well as our financial data and press release discussing our third quarter performance are available at investors.applovin.com.

歡迎參加 AppLovin 截至 2026 年 9 月 30 日止第三季財報電話會議。我是投資人關係主管 David Hsiao。今天與我一同討論本公司業績的有共同創辦人暨執行長 Adam Foroughi，以及財務長 Matt Stumpf。提醒各位，我們迄今向 SEC 提交的申報文件，以及我們的財務資料與討論第三季表現的新聞稿，皆可於 investors.applovin.com 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 expected future financial performance of the company and other future events.

在今天的電話會議中，我們將作出前瞻性陳述，包括但不限於：我們平台未來的發展與觸及範圍、我們預期的成長機會、公司未來預期的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-Q for the fiscal quarter ended June 30, 2026. Additional information may also be found in our quarterly report on Form 10-Q for the fiscal quarter ended September 30, 2026, which will be filed today. We will also be discussing non-GAAP financial measures.

這些陳述係基於我們目前的假設與信念，除法律要求外，我們不承擔更新之義務。我們的實際結果可能與預測結果存在重大差異。我們建議您查閱我們最近提交的、截至 2026 年 6 月 30 日止會計季度之 Form 10-Q 中的風險因素。其他資訊亦可見於我們截至 2026 年 9 月 30 日止會計季度之 Form 10-Q 季度報告（將於今日提交）。我們也將討論非 GAAP 財務衡量指標。

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

這些非 GAAP 指標並非旨在優於或取代我們的 GAAP 結果。請務必查閱我們在投資人關係網站上提供的財報新聞稿與財務更新中的 GAAP 結果，以及 GAAP 與非 GAAP 財務衡量指標之調節表。本次電話會議將被錄音，並將於一段期間內在我們的 IR 網站提供重播與逐字稿。

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.

現在我把時間交給 Adam 和 Matt 做一些開場致詞，接著由主持人帶領我們進行問答。

Thanks, everyone, for joining us today. We delivered a record quarter revenue of $2.18 billion, up 55% year-over-year, the largest sequential dollar step-up in revenue with well over $200 million more than last quarter. Both revenue and EBITDA came in well above the high end of our guidance. That's the standard we've always run this business by outperforming our expectations. That result belongs to our team.

感謝各位今天加入我們。我們交出創紀錄的單季營收 21.8 億美元，年增 55%，也是營收連續季度增量中金額最大的一次，較上季增加超過 2 億美元。營收與 EBITDA 皆遠高於我們指引區間的上緣。這一直是我們經營這門生意的標準：持續超越我們自己的預期。這樣的成果屬於我們的團隊。

They're carrying out a model transition as significant as AXON 2.0, while running the business at full speed, and they delivered both in the same quarter. I want to thank them. Here's what drove it. During the quarter, we partially rolled out the next generation of our models built on the architectural changes I described last quarter. Performance was strong advertisers saw more spend deploying profitably at their target return on ad spend goals and budgets stepped up.

他們在全速運轉業務的同時，完成了如 AXON 2.0 般重大的模型轉換，並且在同一季度兩者都做到了。我要感謝他們。以下是推動成果的原因。本季度，我們部分推出了下一代模型，該模型建立在我上季所描述的架構變更之上。表現強勁——廣告主在達成其目標廣告投資報酬率（ROAS）的前提下，看到了更高的可獲利投放，因此增加了支出與預算。

And our gaming business grew nearly 50% year-over-year, reaccelerating right alongside the rollout. Here's the part I want you to sit with. This was a partial rollout. The transition is still in front of us, and the new generation gives us meaningfully more room to optimize in the architecture it replaces. It also improves performance most where data is sparser, newer advertisers, smaller advertisers, newer categories, exactly where our consumer vertical lives.

同時，我們的遊戲業務年增近 50%，並在推出的同時再度加速成長。我希望各位特別留意的是：這只是部分推出。轉換仍在前方，而新一代模型相較其所取代的架構，為我們提供了顯著更大的最佳化空間。它也在資料較稀疏的情境下改善效果——例如新廣告主、小型廣告主、新類別——而這正是我們消費者垂直領域所在之處。

As always, model gains won't land on a perfect quarterly schedule and we won't promise the cadence, but the headroom is real, and we're still very early. Just as important, we did it with discipline. While part of the team trained the new models, our engineers optimized how we serve them. So even as training investment increased, data center costs declined as a percentage of revenue. Every dollar of this investment is gated on return.

一如既往，模型的提升不會精準落在每一季的固定節奏上，我們也不會承諾其頻率；但可提升的空間確實存在，而且我們仍處於非常早期的階段。同樣重要的是，我們以紀律執行。當部分團隊在訓練新模型時，我們的工程師也在最佳化模型的服務方式。因此，即便訓練投入增加，資料中心成本占營收比重仍下降。我們對每一美元的投入都以投資報酬為門檻。

In this quarter, you saw both halves of that, the return in revenue and the discipline in margin. Consumer had another strong quarter and with Q4 now underway, the seasonal peak for these advertisers Consumer is set up very well heading into the biggest spending season of the year. On the long term, nothing has changed. We run one auction across multiple advertiser categories. Gaming keeps improving.

在本季度，你們同時看到了這兩個面向：營收端的回報，以及毛利率/利潤率端的紀律。消費者業務再度交出強勁季度表現；隨著第四季已展開、也正值這些廣告主的季節性高峰，消費者業務在迎向全年最大投放旺季時具備非常好的態勢。就長期而言，沒有任何改變。我們在多個廣告主類別之間運行同一個競價拍賣機制。遊戲業務持續改善。

Consumer keeps adding runway. And we believe this business can compound at roughly 30% annually over the long term. And as this year shows, we intend to run well ahead of that whenever we can. Our focus heading into next year is consistent. Complete the model transition and optimize on top of it, expand spend from our current customers while really ramping up the rate of acquiring new customers, keep improving our creative tools and ad formats and start expanding our efforts on CTV.

消費者業務持續延長成長跑道。我們相信，長期而言這門生意可望以約 30% 的年複合成長率持續累積成長。而如今年所示，只要有機會，我們就打算跑在這個水準之上。我們邁向明年的重點一以貫之：完成模型轉換並在其上持續最佳化；擴大現有客戶的投放，同時大幅提升獲取新客戶的速度；持續改進我們的創意工具與廣告格式，並開始擴大我們在 CTV（連網電視）上的投入。

Let me close with this. A quarter ago, we stood here answering questions about a miss. We told you what had happened that it was fixed and that the business was reaccelerating. And everything we told you showed up in this quarter's numbers. That's how we'll keep operating, build for the long term, be transparent about challenges and opportunities and compound through the noise.

最後我想以這點作結。一季之前，我們站在這裡回答關於一次未達預期的問題。我們告訴各位發生了什麼、問題已修復，且業務正在重新加速。而我們所說的一切，都反映在本季度的數字中。這就是我們將持續的運作方式：為長期而建、對挑戰與機會保持透明，並在雜訊中持續複利成長。

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

接下來我把時間交給 Matt，帶各位回顧財務表現。

Thanks, Adam, and thanks to everyone for joining us today. AppLovin posted record third quarter earnings. Net income grew 72% year-over-year to $1.44 billion on record revenue of $2.18 billion up 55% year-over-year with an industry-leading 66% net income margin and 84% adjusted EBITDA margin. Revenue grew 14% sequentially and exceeded the high end of our guidance by approximately $100 million, primarily driven by the core mobile gaming business due to model improvements during the quarter with continued scaling in the consumer vertical. Consumer advertiser spend continues to have healthy growth with spend up double digits quarter-over-quarter and growing triple digits from the same quarter last year.

謝謝 Adam，也感謝各位今天加入我們。AppLovin 第三季繳出創紀錄的獲利表現。在創紀錄營收 21.8 億美元（年增 55%）的帶動下，淨利年增 72% 至 14.4 億美元，淨利率達業界領先的 66%，調整後 EBITDA 利潤率為 84%。營收季增 14%，並較我們指引區間上緣高出約 1 億美元；主要由核心行動遊戲業務所驅動，原因是本季度模型改善，同時消費者垂直領域持續擴張。消費者廣告主支出持續健康成長，支出季增達雙位數，且較去年同期成長達三位數。

Adjusted EBITDA was $1.84 billion, up 59% year-over-year. and also above the high end of our guidance range at an 84% margin, up approximately 190 basis points from the same period last year. Quarter-over-quarter flow-through to adjusted EBITDA was 87%. Operating leverage remains strong. Our data center costs declined as a percentage of revenue due to optimizations rolled out by our infrastructure team, even as we invested in an additional compute for training our existing models and for new model development.

調整後 EBITDA 為 18.4 億美元，年增 59%。同時也高於我們指引區間上緣，利潤率為 84%，較去年同期提升約 190 個基點。本季度營收對調整後 EBITDA 的季對季流轉率（flow-through）為 87%。營運槓桿依然強勁。在我們基礎設施團隊推出的最佳化措施帶動下，資料中心成本占營收比重下降；即便我們增加了運算資源投入，用於訓練既有模型以及開發新模型。

We manage this business to EBITDA dollars and free cash flow rather than to a margin percentage, and we'll continue to deploy dollars when we see an opportunity to produce more revenue. Free cash flow for the quarter was $1.4 billion or 77% of adjusted EBITDA, up from 53% in the second quarter. As I mentioned on last quarter's call, second quarter conversion was depressed by the timing of cash tax payments. We still expect full year free cash flow conversion of roughly 70% to 75% of adjusted EBITDA. To put our results in context, this earnings season, the largest platforms in digital advertising reported third quarter revenue growth between roughly 10% and the mid-20s year-over-year.

我們管理這門業務時，更著重 EBITDA 金額與自由現金流，而非單純的利潤率百分比；只要看到能帶來更多營收的機會，我們就會持續投入資金。本季度自由現金流為 14 億美元，約為調整後 EBITDA 的 77%，高於第二季的 53%。如我在上季電話會議中提到，第二季的轉換率受到現金稅款支付時點影響而偏低。我們仍預期全年自由現金流轉換率約為調整後 EBITDA 的 70% 至 75%。為了讓各位理解我們的表現背景，在本次財報季中，數位廣告領域最大的幾個平台所公布的第三季營收年增率約介於 10% 到 20% 中段。

We grew 55%, several times faster than the market. And we did it at margins well above all of our peers across the advertising space. The contrast is even sharper on cash. This quarter, our asset-light model converted roughly 65% of revenue into free cash flow. We ended the quarter with $3.17 billion of cash against $3.5 billion of long-term debt which puts net leverage below 0.1 times adjusted EBITDA, comfortably below the roughly 1 times net leverage where we expect to operate over the long term.

我們成長了55%，增速是市場的數倍。而且我們是在遠高於廣告領域所有同業的利潤率水準下達成的。在現金表現上對比更為鮮明。本季，我們的輕資產模式將約65%的營收轉化為自由現金流。本季末，我們持有31.7億美元現金，對應35億美元長期負債，使得淨槓桿低於調整後EBITDA的0.1倍，明顯低於我們預期長期運營的大約1倍淨槓桿水準。

During the third quarter, we repurchased and withhold a total of 4.1 million shares for $1.4 billion, a meaningful step-up from the second quarter and ended the quarter with 331 million shares outstanding. Given our continued confidence in the company's future, our Board of Directors increased our share repurchase authorization by an incremental $3.2 billion, bringing our total current authorization to $3.6 billion as of the end of October.

在第三季期間，我們回購並扣留合計410萬股，金額為14億美元，較第二季有顯著提升，季末流通在外股數為3.31億股。基於我們對公司未來持續的信心，董事會額外增加32億美元的股票回購授權，使我們截至10月底的目前總授權額達到36億美元。

Since we began the program, we have repurchased and withheld a total of 104 million shares for $9.4 billion at an average price of $90 per share, returning 85% of our free cash flow over that period and reducing our share count materially. Our weighted average diluted share count has now declined for 8 consecutive quarters and is down approximately 2% from a year ago, even as we invest in the business. That combination, 55% growth with a shrinking share count is one very few companies can match.

自我們啟動該計畫以來，我們已回購並扣留合計1.04億股，總金額94億美元，平均每股價格90美元，在此期間回饋了85%的自由現金流，並實質降低了股數。我們的加權平均稀釋後股數已連續8個季度下降，較一年前約下降2%，即便我們仍在投資業務。這樣的組合——在股數縮減的同時仍能達成55%的成長——是極少數公司能夠匹敵的。

Turning to our outlook for the fourth quarter of 2026. We expect revenue between $2.365 billion and $2.395 billion, representing approximately 43% to 44% year-over-year growth or 8% to 10% sequentially. We expect adjusted EBITDA between $1.995 billion and $2.025 billion, representing approximately 43% to 45% year-over-year growth.

接著談我們對2026年第四季的展望。我們預期營收介於23.65億至23.95億美元，約相當於年增43%至44%，或季增8%至10%。我們預期調整後EBITDA介於19.95億至20.25億美元，約相當於年增43%至45%。

At the midpoint, this guidance implies full year 2026 revenue and adjusted EBITDA growth of more than 50%. That outlook reflects normal seasonality, further scaling in our consumer vertical, significant model improvements that are already live and performing and continued investment in training compute for new model development.

以中位數來看，這份指引意味著2026年全年營收與調整後EBITDA的成長都將超過50%。該展望反映了正常的季節性、我們在消費者垂直領域的進一步規模化、已上線並表現良好的重大模型改進，以及持續投入訓練算力以開發新模型。

To close, we exceeded the high end of our guidance on both revenue and adjusted EBITDA, grew revenue 55% year-over-year, expanded margins, generated $1.4 billion of free cash flow and returned $1.4 billion to shareholders. Our fourth quarter outlook reflects that same trajectory.

最後，我們在營收與調整後EBITDA兩項指標上都超越了指引區間的上緣，營收年增55%，利潤率擴張，產生14億美元自由現金流，並向股東回饋14億美元。我們對第四季的展望也反映相同的走勢。

With that, let's move to Q&A.

接下來，我們進入問答環節。

James Heaney.

James Heaney。

Yeah, great. Just a couple for me. I'll start with Adam. Could you just talk about these new generations of models that you're rolling out? Just curious what the biggest improvement that you see is compared to Axon two and then anything around just timing of the full rollout? And then I'll ask one to Matt.

好的，很棒。我這邊只有幾個問題。我先問Adam。你能談談你們正在推出的這些新一代模型嗎？我想了解相較於Axon 2，你看到最大的改進是什麼，以及完整全面推出的時間點大概如何？然後我再問Matt一題。

Hi James, we talked about it last quarter, but one of our objectives was try to get to a place where we could build bigger models, use more compute and try to train a model that has better output. And so I foreshadowed it for you last quarter. We're now starting to roll those out. We've seen good results. We can't tell you or commit to when we can actually see a full rollout or how the progression of these models will go.

嗨James，我們上季談過，其中一個目標是希望能到達一個狀態：我們可以建更大的模型、使用更多算力，並嘗試訓練出輸出更好的模型。所以我上季已先跟你預告過。我們現在開始逐步推出這些模型。我們看到不錯的結果。但我們無法告訴你或承諾何時能真正完成全面推出，或這些模型的演進節奏會如何。

But what I said in the talk track is these models have a lot more room to optimize them than the old generation that we were on. So we're really excited about the future.

但我在講稿裡提到的是，這些模型相較於我們先前使用的舊一代，有更大的優化空間。所以我們對未來非常興奮。

Great. And then one for Matt. Just on the surface, it feels like the Q4 revenue guide of the 8% to 10% sequential growth seems a get a bit more conservative versus the 14% in Q3 when you layer in the strength of e-commerce in the quarter for Q4? Just how would you frame up that sequential deceleration?

很好。接著問Matt一題。表面上看，第四季營收指引的季增8%到10%，相較第三季的14%似乎更保守一些，尤其考量到第四季電商在本季的強勁表現。你會如何解釋這個季增放緩？

Yes. I mean our guidance philosophy remains the same from what we've done over the past few quarters. We're guiding to what we know at that point in time. So we're guiding to include the points that I said in my script, which are normal seasonality within the quarter, the model updates that we've already launched and are live and performing and then obviously, the increase in the data center costs. So nothing there has really changed from previous quarters.

是的。我們的指引哲學與過去幾季一致。我們會依據當下時間點我們所掌握的資訊來提供指引。因此，我們的指引納入了我在講稿中提到的幾點：本季內正常的季節性、我們已經推出且已上線並表現良好的模型更新，當然還有資料中心成本的增加。所以這些方面其實與前幾季相比並沒有真正改變。

Jason Bazinet, Citi.

Jason Bazinet，花旗。

Just one question. I think last quarter, you guys talked about leaning into the referral program a bit more. I was just wondering if you could give us an update on whether you're pleased or not with the progress you're making with those partners.

只有一個問題。我記得上季你們提到會更積極投入推薦（referral）計畫。我想請你們更新一下：你們對與這些合作夥伴推進的進展是否滿意？

Yes. Jason, still work in progress. Obviously, this is going to be multi-quarter, multiyear trajectory on this business. We're early in doing partnerships. But as you've seen, even in the rate of Pixel accounts that I think you're publishing is that we've been getting more and more customers coming on and the rate is accelerating.

是的，Jason，仍在進行中。顯然，這會是這項業務跨季度、跨年度的發展軌跡。我們在做合作夥伴關係方面還在早期。但如你所見，即便在你發布的Pixel帳戶增長速度上，我們確實吸引到越來越多客戶加入，而且增速正在加快。

That's only going to compound over time because not only you're going to have more partnership deals, we'll start running more ad campaigns, but you're just going to have more positive sentiment on the product in the market as more customers start getting a good experience and doing it over time because the other thing you got to realize is we've got customers that come on. They have to get campaigns that are scaled and optimized on our platform, they have to run an incrementality study. They have to prove that it works and then they really scale for the long term. So -- and we expect this stuff will just keep compounding over time.

這只會隨時間推移而加速累積，因為不僅會有更多合作夥伴交易，我們也會開始投放更多廣告活動；而且隨著更多客戶獲得良好體驗並持續使用，市場對產品的正面口碑也會增加。但你也要理解的是，我們有些客戶加入後，必須先在我們的平台上把投放規模化並完成優化，還要做增量（incrementality）研究。他們必須證明有效，之後才會在長期大幅擴大規模。所以——我們預期這些因素會隨時間持續複利式累積。

Matthew Cost, Morgan Stanley.

Matthew Cost，摩根士丹利。

Great. I guess, does the experience that you've had with this new model that just rolled out does that change your expectations for how much runway there still is ahead to improve gaming ad revenue? Because you've been relatively specific over the course of this year about a 20% to 30% growth expectation and sort of law of large numbers weighing on gaming. Would you reiterate all that at this point? Or is there potentially more headroom than you would have otherwise expected based on how it's gone so far?

很好。我想問，你們剛推出的新模型所帶來的經驗，是否改變了你們對遊戲廣告營收仍有多少提升空間（runway）的預期？因為今年以來你們對20%到30%的成長預期相對具體，也提到遊戲業務受到規模效應（law of large numbers）的影響。你們目前會重申這些看法嗎？或者基於目前的進展，是否可能比你們原先預期有更多上行空間？

So I mean what we've been specific on the last 2 quarters that we're committed to is an expectation that we can grow this business 30% year-over-year for the long term. And we are a platform that's a unified auction. So let's say, in some world, we grew consumer 1,000% in the next quarter. Well, that's going to create more competition, right? So like you can't look at it verticalized as this consumer business is now growing more and more.

所以，我們在過去兩季一直明確承諾的，是我們預期長期可以讓這項業務年增30%。而我們是一個統一拍賣（unified auction）的平台。所以假設在某種情境下，我們的消費者業務在下一季成長了1,000%。那麼這會帶來更多競爭，對吧？因此你不能用垂直領域切分的方式來看，說消費者業務現在越來越成長。

It's very hard to judge a vertical by itself in a silo. Now that said, this model is, for sure, more accurate than the old model, and it has a lot more room to optimize going forward. So if we're able to do that, of course, we would expect that gaming could continue to grow at higher rates for the long term. It's just not something that we need, though, in order to hit that 30% if we have both parts of the business growing pretty quickly. And as we touched on in the talk track, the other side of the business, the consumer side is growing triple digits.

很難把某一個垂直領域單獨放在孤島裡來判斷。不過話說回來，這個模型確實比舊模型更精準，而且未來還有更大的優化空間。所以如果我們能做到這點，當然我們會預期遊戲業務長期仍可維持較高的成長率。但要達到那個 30% 並不一定需要它，因為只要業務的兩個部分都成長得很快即可。而且如同我們在講稿中提到的，業務的另一端——消費者端——正以三位數的速度成長。

So when you have a business like that, that now isn't immaterial in our business, growing at that rate with a lot of opportunity to grow in the future as we go get more customer penetration, it gives us a lot of headroom on future growth opportunity.

所以當你有這樣一個業務，如今在我們整體業務中已不是無足輕重，並且以這樣的速度成長、未來隨著我們提升客戶滲透率還有很大的成長機會時，這會為我們未來的成長機會提供很大的上行空間。

Got it. And then I noticed you mentioned CTV in the prepared remarks. I think it's been a little while since that's something you chose to highlight. I guess, how prominently does that feature into your product road map right now relative to 6 months ago?

了解。另外我注意到你在事先準備的發言中提到了 CTV。我覺得你已經有一段時間沒有特別強調這個了。我想問的是，相較於 6 個月前，這在你們目前的產品路線圖中占比有多重要？

Yes, we're making more progress there. I mean if people watch anything on connected television, you'll start seeing some of our ads and we've got pilot campaigns going now. The product is actually looking very promising. It's early. It's a whole new market.

是的，我們在那方面有更多進展。我的意思是，如果大家有在連網電視上看任何內容，你會開始看到我們的一些廣告，而且我們現在也在跑試點投放活動。產品實際上看起來非常有前景。但還很早。這是一個全新的市場。

So I wouldn't bank on it at all for being accretive to numbers anytime soon. But it's something that the team is actually gaining a lot of confidence in with where they're at in the product road map. And so if we get to the point where it actually starts showing really good result, for customers, we'll talk about it more. Internally, we're now very excited about it.

所以我不會指望它在短期內對數字有任何增益。但團隊在產品路線圖目前的進度上，確實正在建立很大的信心。因此如果我們到了它開始為客戶呈現非常好的成效的階段，我們會更常談到它。在內部，我們現在對此非常興奮。

Benjamin Black, Deutsche Bank.

Benjamin Black，德意志銀行。

Adam, you've spoken in the past the need to collect more data on the consumer side to refine the model and to improve performance and then the flywheel around that. And now with the new architecture, do you have sort of line of sight to potentially getting that flywheel to reach sort of a tipping point to where you were with gaming a few years back? And if not, which incremental investments do you think are required?

Adam，你過去談過需要在消費者端蒐集更多資料，以便精煉模型、提升效能，並帶動那個飛輪效應。那現在有了新的架構，你是否已經能大致看得到，這個飛輪有機會達到某個臨界點，回到幾年前遊戲業務那樣的狀態？如果還沒有，你認為需要哪些額外的增量投資？

No, look, these two things go hand in hand, right? Every day that we stay in business, we get customers that come into the platform, they share data with us so we get more data density. So as you fast forward that quarters, years, you then need to pair it with model architecture that can actually process more data and generate better output. So you write a bigger model, more complex model, you train it on a larger compute cluster, and you end up with better output. And what this new architecture allows us to do on the gaming side and on the consumer side, is just train bigger models.

不，你看，這兩件事是相輔相成的，對吧？我們每天只要還在營運，就會有客戶進入平台，他們與我們分享資料，因此我們的資料密度會提高。所以當你把時間拉到幾季、幾年之後，你就需要搭配能夠處理更多資料並產生更好輸出的模型架構。因此你會寫出更大的模型、更複雜的模型，在更大的運算叢集上訓練它，最後得到更好的輸出。而這個新架構讓我們在遊戲端與消費者端都能做的，就是訓練更大的模型。

So we're going to be starved for data. The more data we get into the system quicker, the better the output of our predictions model is going to become. And so we're never going to stop looking to go get data, whether it's promoting ourselves through referrals and advertising as we just touched on for new customers or potentially some sort of data acquisition we're going to be looking for ways to get data into this model because we've seen every single time we add more data, we're able to write a more complex model, we're able to train a bigger model and just get a better output on the other side.

所以我們會一直覺得資料不夠用。越快把更多資料導入系統，我們的預測模型輸出就會越好。因此我們永遠不會停止尋找資料，不論是像剛才提到的，透過推薦與廣告來推廣自己以獲取新客戶，或是可能進行某種資料收購；我們都會尋找把資料導入這個模型的方法，因為我們看到每一次加入更多資料，我們就能寫出更複雜的模型、訓練更大的模型，並在另一端得到更好的輸出。

Great. And I guess, how are consumer advertiser budget is shaping up for the holiday shopping season relative to last year. And if you think advertisers see strong returns this quarter, how much of that spend could ultimately prove sticky into '27? Or should we expect a seasonal reset as usual in Q1?

很好。那我想問，消費品廣告主在假期購物季的預算，相較於去年目前的走勢如何？如果你認為本季廣告主看到強勁回報，那其中有多少支出最終可能會一路黏著延續到 2027 年？還是我們應該如往常一樣，預期在第一季出現季節性回落？

I think it just follows the normal patterns. I mean, look, we're still early in the curve. It's not even into the holiday season. So that stuff takes form really over the next couple of weeks. But that seasonal reset, I mean, that's consistent in e-com.

我認為就是遵循正常的模式。我的意思是，你看，我們仍在曲線的早期階段。甚至還沒真正進入假期季。所以那些情況會在接下來幾週逐漸成形。但那種季節性回落，在電商裡一直都很一致。

Now we do have the benefit that so long as we're growing this business at this rate and we're pretty early in its evolution, we still have rate of customer acquisition, allowing us to grow the business even in more traditionally seasonally difficult quarters like the Q1 versus Q4 comp. With that said, like the same advertiser like-for-like will have a huge Q4 and followed it up with a weak Q1.

不過我們確實有一個優勢：只要我們以這樣的速度成長，而且這個業務仍處於演進早期，我們仍有客戶獲取的速度，讓我們即使在傳統上季節性較困難的季度（例如第一季相較第四季的比較）也能成長。但話說回來，同一批廣告主的同店（like-for-like）表現會是第四季非常強，接著第一季偏弱。

That's just the reality of the e-commerce business. There, we traditionally, if you looked at the business growth sequentially in our business, our Q1s have been pretty strong because gaming has a phenomenon called Q5. We're going into Q1, they know that there's not as much competition.

這就是電商業務的現實。而我們這邊，傳統上如果你看我們業務的逐季成長，第一季一直都相當強，因為遊戲產業有一個現象叫做 Q5。進入第一季時，他們知道競爭不會那麼激烈。

And so the models whether just selecting gaming more often because return on ad spend is compelling or the customers being willing to take bigger risk because they're coming off of a very high-priced Q4 into this post Q4 season. They end up spending very competitively. So we've been more insulated than other advertising companies. But as we get more deep into e-commerce, we should see bigger Q4 is followed by tougher Q1.

因此，不論是模型因為廣告投資報酬率（ROAS）很有吸引力而更常選擇遊戲，或是客戶在經歷了價格非常高的第四季後、進入第四季後的淡季而願意承擔更大風險，他們最後都會以很有競爭力的方式投放。所以我們比其他廣告公司更能抵禦季節性影響。但隨著我們更深入電商，我們應該會看到更大的第四季之後，接著是更艱難的第一季。

Stephen Ju, UBS.

Stephen Ju，瑞銀。

Right. So I think I was talking to some of the advertisers on your platform and they said something curious. They were anticipating that the pricing on the auction is going to start getting inflated because you're going to have incremental bidders from e-commerce coming in and taking up the prices.

好的。所以我想我最近跟你們平台上的一些廣告主聊過，他們提到一件有點有趣的事。他們預期拍賣的定價會開始被推高，因為會有來自電商的新增競價者進來把價格抬上去。

So they were preemptively taking some of the budgets to some of your competitors in anticipation of what might be compression of ROI. So which kind of seemed like a strange thing to do, but it did make me worry about, I guess, the amount of inventory, which I think you've addressed in prior periods of this being, I guess, a demand-constrained industry as opposed to supply constrained industry.

所以他們先發制人地把部分預算轉到你們的一些競爭對手那邊，以因應可能出現的 ROI 壓縮。這樣做看起來有點奇怪，但也確實讓我擔心——我想——可用庫存的量；而我記得你們在之前的期間談過，這個產業更像是需求受限，而不是供給受限。

So can you talk about how your thinking might have evolved or things have changed at all in terms of AppLovin's ability to absorb all this incremental ad spend without having some of the ROI get compressed for some of your incumbent customers.

所以你能談談你們的看法是否有所演變，或在 AppLovin 吸收這些新增廣告支出、同時不讓既有客戶的 ROI 被壓縮方面，有沒有任何改變嗎？

Yes. I mean if you look at our business over the last three years, we've gotten a lot more competitive. The business is much, much bigger. The ROI or return on ad spend for customers is stronger than it was three years ago because the spend is materially larger, right? So the better these models get at taking an impression and predicting an outcome, the more accurate they can price.

是的。我的意思是，如果你看我們過去三年的業務，我們變得更有競爭力了。業務規模大了很多很多。客戶的 ROI（或廣告投資報酬率）比三年前更強，因為投放規模實質上更大，對吧？所以這些模型越能把一次曝光（impression）與結果做出更好的預測，它們就越能更精準地定價。

Our step-up this quarter in net revenue dollars against the last quarter I don't know how much of our competition that's publicly reporting has in terms of scale to directly compare to that. But I think it's a lot of their annual reported revenue that we just put up in one quarter of growth.

我們本季相較上季的淨營收美元增量，我不確定我們那些有公開揭露的競爭對手在規模上有多少能直接拿來對比。但我認為，我們單一季度的成長，可能就相當於他們全年所揭露的營收規模。

So I think these industry checks tend to end up being flawed because you're looking at a small subsegment of the ecosystem and the customers that are taking time talking to you all probably aren't working that hard. The ones that are working hard and succeeding probably aren't talking to you all. So I would say we're a pretty good indicator on the market and our technology has proven to be great at getting better pricing for customers.

所以我認為這些產業調查最後往往會有偏差，因為你看到的是生態系中一個很小的子區段，而那些願意花時間跟你們聊的客戶，多半也不是最努力在做事的。真正努力且做得成功的那些人，可能根本不會跟你們聊。所以我會說，我們其實是市場上一個相當不錯的指標，而且我們的技術已經證明，能為客戶爭取到更好的定價。

That doesn't mean that just looking at a CPM, the price is not going up. In fact, if we get better targeted so that our conversion rate goes up, the CPM will go up. but the down funnel effect of a better prediction is that the return on ad spend is going up, and that's why we're able to grow so quickly.

這並不代表只看 CPM（千次曝光成本）時，價格就不會上升。事實上，如果我們做到更精準的定向、讓轉換率提高，CPM 反而會上升。但更好的預測在漏斗下游帶來的效果是：廣告投資報酬率（ROAS）在提升，而這也是我們能夠如此快速成長的原因。

Robert Coolbrith, Evercore.

Robert Coolbrith，Evercore。

Just wanted to ask a little bit more about the architecture. Is that something as it becomes larger or allows for more data models get bigger. Is that something that also becomes more generalizable. I know in the past, you talked about keeping the consumer and the gaming model separate, but is that something that something where in the models become sort of more generalizable across media types, across client types over time.

我想再多問一點關於架構的問題。當它變得更大、或允許更多資料模型變得更龐大時，它是否也會變得更具通用性（generalizable）？我知道過去你們談過要把消費者模型和遊戲模型分開，但隨著時間推進，這些模型是否會在不同媒體類型、不同客戶類型之間變得更通用？

Yes, it's possible sorry to interrupt on the second question. But you're right, it is possible that we will eventually be able to take, this architecture in that direction.

是的，有可能——抱歉我先插一句回應第二個問題。但你說得對，我們最終有可能把這個架構往那個方向發展。

Okay. Got it. And then just I wanted to ask a little bit about the brand performance, if you will. We heard from more brands maybe because of the issues in mobile web, both in terms of inventory supply, but also iOS 26 focus more on mobile app inventory. Do you see that as an opportunity that's potentially attractive interesting for AppLovin going forward?

好的，了解。另外我也想問一下品牌端的表現（如果可以這麼說）。我們聽到有更多品牌，可能因為行動網頁（mobile web）上的問題——包括庫存供給，以及 iOS 26 使得大家更聚焦於行動 App 的庫存。你們是否把這視為 AppLovin 未來一個可能很有吸引力、很有意思的機會？

I mean, certainly, if it's a performance marketer, so I would split up performance brand into true performance where they're optimizing the incremental ROAS or like a down funnel purchase transaction. then you've got performance brand that still is looking for audience reach. A lot of the mobile web advertising and a lot of what's called, I guess, traditionally open web is more of the latter. They're just trying to buy audience and trying to price effectively as cheap as possible to reach that audience. we try to price the discovery all the way to the point of transaction.

我的意思是，當然——如果是效果型行銷（performance marketer），我會把「效果型品牌」拆成兩類：一類是真正的效果導向，也就是在優化增量 ROAS，或是像漏斗下游的購買／交易。另一類是所謂的效果型品牌，但其實仍然在追求受眾觸及（audience reach）。很多行動網頁廣告，以及很多傳統上所謂的 open web（開放網路），比較偏後者。他們只是想買受眾，並且希望以盡可能便宜的價格有效觸及該受眾。而我們則是從發現（discovery）一路到交易發生的那一刻來做定價。

So that's what they care about, and they're able to track it all on digital, then yes, we think a lot of those dollars will shift into the app space. It clearly performs really well. it's much, much bigger. I mean if you I said this, I think, on the All-In Pod at the All-In Summit, the scale of this space is getting very, very large. We have I mean, you can back into the gross spend numbers that we have and they're pretty substantial.

所以如果他們在意的是這件事，而且能在數位端把整個流程都追蹤到，那麼是的，我們認為其中很多預算會轉移到 App 端。它的表現顯然非常好。而且規模要大得多、非常非常大。我的意思是——我想我在 All-In Summit 的 All-In Pod 上也提過——這個領域的規模正在變得非常、非常龐大。我們——我的意思是，你們可以從我們的總投放（gross spend）數字回推，規模相當可觀。

And if you then gross that up for the whole space, this is one of the largest advertising opportunities that there are outside of social and search, maybe the largest outside of social and search. So we fully expect more dollars will follow given the performance is there.

如果再把它放大到整個市場來看，這會是社群與搜尋之外最大的廣告機會之一，甚至可能是社群與搜尋之外最大的。所以只要效果存在，我們完全預期會有更多預算跟進。

Clark Lampen, BTIG.

Clark Lampen，BTIG。

I have two, if I may. I guess, Adam, I sort of wanted to pull out this model thread. And based on everything that I understand about where you guys are going right now, I think one of the bigger derivative benefits of success would be opening up a lot more inventory or available inventory, I guess, within the same sort of base of games and supply that you have right now, which probably should get filled because of auction dynamics by non-gaming advertisers. Is that right, I guess, just sort of number one, as we think about like the second order benefit beyond just sort of more ads per 100 that you're buying right now translating to revenue-generating customers for a commerce or gaming developer. And if so, I guess, is there a way to frame for us just how significant, I guess, this sort of available or effective supply expansion could be?

如果可以的話我有兩個問題。Adam，我想把話題拉回到模型這條主線。根據我對你們目前方向的理解，我認為成功帶來的一個更大的衍生效益，是能在你們現有的遊戲基礎與供給之上，開放出更多庫存或可用庫存；而且由於拍賣機制，這些庫存很可能會被非遊戲廣告主填滿。這樣理解對嗎？也就是說，除了你們現在每 100 次曝光買到更多廣告、進而轉化為電商或遊戲開發者的營收型客戶之外，這算是第二層的效益。如果是的話，你們能否幫我們框架一下：這種可用供給或有效供給擴張，可能會有多大幅度？

Yes. I mean, look, I think the challenge with supply expansion is that the MAX market or the gaming advertising market is a lot bigger than people realize. So when you've got this market that's in the tens of billions, we can go get new supply. And in fact, this model that we're now on and that consumer vertical growing gives us a lot of opportunity to go get more supply. But there aren't that many supply sources that are sort of open to third parties that could be a huge needle mover in one shot.

是的。我的意思是，你看，我認為供給擴張的挑戰在於：MAX 市場或遊戲廣告市場其實比大家想像的大得多。所以當你面對的是一個數百億美元規模的市場時，我們是可以去取得新的供給的。而且事實上，我們現在採用的這個模型，以及消費者垂直領域（consumer vertical）的成長，確實給了我們很多機會去取得更多供給。但能對第三方開放、並且一次就能帶來巨大增量的供給來源，其實沒有那麼多。

So then you need to compound it over time. The reason CTV is compelling to us is there's not a true top of funnel to bottom of funnel performance offering on connected TV. If we can figure that out and be the first to market with something that's scalable there, there's a lot of available inventory that's under monetized there. It's sort of what mobile looked like to us 10, 15 years ago when we started the business. So something like that could be impactful.

因此你需要隨時間去複利式地累積。CTV（連網電視）之所以對我們有吸引力，是因為在連網電視上，還沒有真正從漏斗頂端到漏斗底端的效果型產品供給。如果我們能把這件事做出來，並且成為第一個推出可規模化方案的業者，那裡有大量尚未被充分變現的可用庫存。對我們來說，它有點像我們 10 到 15 年前創業時看到的行動端。所以像這樣的事情可能會很有影響力。

Anything else not so much. But that's why we've always talked about this conversion rate. We're still very low on what we think is the possible conversion rate inside this gaming ecosystem. What consumer gives us is a chance to diversify the ads that someone is seeing. So we are showing the same user 20 gaming ads in a week.

其他的就沒那麼多。但這也是為什麼我們一直在談轉換率。在我們看來，在這個遊戲生態系裡，轉換率距離可能達到的水準仍然非常低。消費者業務帶給我們的，是讓某個人看到的廣告類型更多元的機會。所以我們不再讓同一個使用者一週看到 20 個遊戲廣告。

We might show them 15 product ads and 5 gaming ads. That doesn't mean that the gaming customers lost 75%. It just means that the model can get much more targeted. So those same 5 impressions that would have driven the conversion are still going to sit there. And now all of a sudden, you have all these product ads.

我們可能讓他看到 15 個商品廣告和 5 個遊戲廣告。這不代表遊戲客戶流失了 75%。它只是表示模型可以做到更精準的定向。所以原本那 5 次能帶來轉換的曝光仍然會存在。而現在你突然多了這些商品廣告。

So that's really the focus area. That's the largest needle mover in the short, medium term, but long term, for sure, supply expansion is something that we're going to be focused on.

所以這才是真正的重點方向。在短期到中期，這是最大的槓桿；但長期來看，供給擴張當然也是我們會聚焦的事情。

Okay. That makes sense. And for Matt, I guess, just sort of net of what I think we were seeing just a moment ago in terms of after-hour share price reaction, your stock is still down ballpark 40% year-to-date, and it's only retraced most of the sort of post 2Q earnings step down, does the buyback give you flexibility or the added authorization that you guys just layered on give you flexibility to be more aggressive, I guess, and opportunistic. If you think that the share is or the share price is meaningfully disconnected from intrinsic value? Or is this mostly going to be programmatic going forward?

好的，這很合理。另外問 Matt：撇開我們剛剛看到的盤後股價反應不談，你們的股票今年迄今大約仍下跌 40%，而且也只是收復了 2Q 財報後下跌的大部分幅度。回購是否讓你們更有彈性？或者你們剛新增的授權額度，是否讓你們能更積極、更具機會主義地進行回購？如果你們認為股價與內在價值（intrinsic value）之間存在顯著脫鉤？還是說，接下來主要會以程式化（programmatic）的方式進行？

No. I mean you saw within the quarter as well, Clark, though, we did a meaningful step up. And so that was directly as a result of what you're referring to, right? That was us opportunistically ramping up. the buyback because we saw this dislocation in the stock price versus the actual intrinsic value of the company.

不。我的意思是，Clark，你也看到在本季度內，我們確實做了顯著的提升。所以那正是你所提到的情況所直接導致的，對吧？那是我們在機會出現時加大回購力道。因為我們看到股價相對於公司實際內在價值出現了錯位（dislocation）。

So we have the ability to do that going forward as well.

因此，我們未來也有能力做到這一點。

Martin Yang, OpCo.

Martin Yang，OpCo。

A question for Adam. At the industry conference, you referenced a dynamic product ad for e-commerce, sometimes it can work better than video. So I want to get your thoughts on the longer term, do you think a more diversified ad format is the ultimate goal to make e-commerce more effective? Are you still just waiting on Gen AI video creation to make volume and the quality of video ads higher, and this dynamic product ad is just an intermittent step for your advertisers to get volume to in front of customers?

有一個問題想問 Adam。在產業研討會上，你提到用於電商的動態產品廣告，有時候效果可能比影片更好。所以我想聽聽你對長期的看法：你是否認為，更加多元化的廣告格式是讓電商更有效的終極目標？你們是否仍在等待生成式 AI 的影片創作，讓影片廣告的量能與品質提升，而動態產品廣告只是讓廣告主先把量做起來、觸及客戶的過渡性步驟？

Look, I'd say it's both. I mean, as the large language models that are applied to creative continuously improve and there's a lot of companies in this ecosystem building ads now out of the box, you're going to see better advertiser performance. we're also going to create more ad formats that can service the advertisers in hopefully better ways. What we want is diversity of what the consumer sees. When you serve as many ads a day as we do, that diversity of content can create an uplift.

我想兩者都有。我的意思是，隨著應用在創意上的大型語言模型持續進步，而且這個生態系裡也有很多公司現在可以開箱即用地製作廣告，你會看到廣告主的成效變得更好。我們也會打造更多能以更好的方式服務廣告主的廣告格式。我們想要的是消費者看到的內容要多樣化。當我們每天投放的廣告量像現在這麼大時，內容的多樣性可以帶來成效提升。

And so that's why in my talk track, that was one of the bullets. It's an area of focus for us. This product is still really new. So I mean I know we're a couple of years in, but it's a really new product. the ad format hasn't evolved that much yet.

所以在我的講稿脈絡裡，那是其中一個重點。這也是我們的聚焦領域。這個產品其實還很新。所以我的意思是，我知道我們已經做了幾年，但它仍然是個很新的產品。這個廣告格式到目前為止還沒有演進太多。

I would fully expect it to evolve a lot more over the next couple of years.

我完全預期在接下來幾年它會有更大的演進。

Got it. A follow-up on e-comm, if I may. When you compare this year going into the holiday season versus last year. How would you say the learning budget or learning time is for e-comm customers? How has that improved, if any?

了解。如果可以的話，我想追問一個電商相關問題。把今年進入假期旺季與去年相比，你會怎麼形容電商客戶的學習預算或學習時間？如果有的話，這方面有什麼改善？

Yes. I mean, look, learning is two-sided. So our model has gotten better. And this new model architecture also, as I said in the talk track, is faster when there's less data. On the other hand, customers are risk averse going into a new relationship.

有。我是說，學習是雙向的。所以我們的模型變得更好了。而且如我在講稿中提到的，這個新的模型架構在資料較少時也更快。另一方面，客戶在建立新合作關係時會比較風險趨避。

So when they buy on us, even if the result looks better, they still need to validate it. So in this category, if you talk to the customers, not only do they want to see spend and return on ad spend, they need to run incrementality studies, too. So there is a slow ramp-up. It's just a fact of the industry, but our product is getting better. Every quarter that we go on and we have a new release part of the space is getting it better is just improving it for current customers.

所以當他們在我們平台上投放時，即使結果看起來更好，他們仍需要驗證。因此在這個類別裡，如果你跟客戶聊，不只是想看到花費與廣告投資報酬（ROAS），他們也需要做增量（incrementality）研究。所以會有一個較慢的爬坡期。這就是產業現實，但我們的產品正在變得更好。每一季我們持續推進並發布新版本，這個領域的一部分就是把它做得更好，也就是持續為現有客戶改善。

The other space of getting it better is that so a customer now clicks in, they don't have much data on our system. How quickly can you show them a good result because that's how you lower your churn rates.

另一個把它做得更好的面向是：當客戶現在進來點擊時，他們在我們系統裡沒有太多資料。你能多快讓他們看到好的結果，因為這就是降低流失率的方法。

Rob Sanderson, Loop Capital.

Rob Sanderson，Loop Capital。

A couple of questions on consumer as well, growing triple digits, obviously, good trajectory here. Any commentary on sort of conversion of new leads coming through self-service or through your referral partners or any specifics on customer outcomes. I'm just kind of wondering, there is still a bit of hit and miss kind of depending on sophistication of the advertisers. Are we still seeing sort of similar trends or anything you can comment on sort of how the go-to-market is progressing?

也有幾個關於消費者（consumer）的問題，成長達到三位數，顯然走勢很好。對於透過自助式（self-service）或透過你們的推薦合作夥伴帶來的新線索轉換，或任何客戶成果的具體情況，有沒有什麼評論？我只是有點想了解，依廣告主成熟度不同，似乎仍有一些時好時壞的情況。我們是否仍看到類似趨勢？或你能否談談你們的市場推進（go-to-market）進展如何？

Yes. I mean look, we're still early in the product. So like if a customer goes through our team, much higher likelihood to succeed than they just go through self-service channel and they're on their own. Now if they go through self-service and they're on their own and they're sophisticated, they absolutely can have success. If they're a very small business, something that is just what you would classify as complete SMB, that's still further into the future for us.

有。我是說，你看，我們的產品仍在早期。所以如果客戶是透過我們的團隊導入，成功的機率會高很多；相較之下，如果他們只是走自助式通路、自己摸索。當然，如果他們走自助式、自己操作，而且本身很成熟，他們絕對也能成功。但如果他們是非常小的企業，也就是你會歸類為純 SMB（中小企業）的那種，這對我們來說還是更久以後的事。

So there is a road map on this thing where the more data you get into the system, the more predictive you can be in the absence of data, and we're not at the point yet where somebody can just plug in tiny business, selling jewelry. We've never seen them before, very little data, very little GMV, and it just plugs in and then they go. But five years from now, if we've got a lot of scale of data and the model has time to get evolved to the point of five years from now, you're going to have much better outcomes across the board. So this is a road map. I mean, we got to build this product out over time to tackle every single circumstance, it is working really well for the customers that we either handhold or they come in and are sophisticated.

所以這件事有一個產品路線圖：你把越多資料放進系統，在資料不足的情況下就越能做出預測；而我們還沒到那個階段——有人可以只是接入一個很小的生意、賣珠寶。我們以前從沒見過他們，資料很少、GMV（商品交易總額）也很小，然後一接上就能跑起來。但五年後，如果我們有很大規模的資料，而且模型也有時間演進到五年後的水準，你會看到整體各方面的成果都會好很多。所以這是一個路線圖。我的意思是，我們必須隨時間把這個產品建起來，去應對每一種情境；目前它對於我們會手把手協助的客戶，或是本身就很成熟、自己進來的客戶，效果非常好。

Is ad creative still a primary barrier? Like that was a lot of the points of failure, a lot of things that weren't working it's kind of stalled out at the ad creative. Is that still kind of the primary sticking point? And if we kind of asked this question a year ago, I mean, did you expect the industry to be a little further ahead on Gen AI tools by this point, one year ago? It seems like to me at least, maybe things are a little bit slower than we would have thought?

廣告創意仍然是主要障礙嗎？像是很多失敗點、很多跑不動的地方，似乎都卡在廣告創意上。這仍然是主要的卡點嗎？另外，如果我們一年前問這個問題，你當時是否預期到現在為止，產業在生成式 AI 工具上會走得更前面一些？至少在我看來，進展似乎比我們想像的慢一點？

Yes. I mean, look, on the latter, the large language model companies in the West are focused on something else. And so you have less R&D outside of us on this problem, right? On the former I wouldn't call it a barrier, but it is an opportunity cost or opportunity loss for customers that don't invest in creative. If you talk to customers that buy on Meta, they buy on TikTok, they buy on us.

是。我是說，先講後者，西方的大型語言模型公司把重心放在別的事情上。因此在這個問題上，除了我們之外的研發投入就比較少，對吧？至於前者，我不會稱它為障礙，但對於不投資創意的客戶而言，這是一種機會成本或機會損失。如果你跟在 Meta 投放、在 TikTok 投放、也在我們這裡投放的客戶聊，

There's a constant like point that they make that you need to put more creatives into AppLovin to have success on our platform. So the ones that know that have more success than the ones that don't know that. The ones that don't might be smaller, they may it's probably not a barrier to working at all but they're missing a point of scale. And the challenge is that our ads are just much longer and they are seen. And when you go into another platform and you load 10 ads, they're not seen that long and a lot of them aren't seen.

他們會一直強調一點：你需要在 AppLovin 放入更多創意素材，才能在我們平台上成功。所以知道這點的人，比不知道的人更容易成功。不知道的人可能規模較小；這大概不是完全無法運作的障礙，但他們少了一個做大規模的關鍵。而挑戰在於，我們的廣告就是長得多，而且會被看到。當你到另一個平台上傳 10 支廣告，它們不會被看那麼久，而且很多根本沒被看到。

Load 10 ads into our platform, on average, you're getting over 35 seconds of watch time, you just can't do anything else on the screen. It's a full screen ad. And so they are being seen with 100% certainty. So ad count and quality of ad combination is what creates scale in our platform. Some customers know it, they're able to become our large spenders.

在我們平台上傳 10 支廣告，平均你會拿到超過 35 秒的觀看時間，而且螢幕上你無法做其他事。這是全螢幕廣告。因此它們是以 100% 的確定性被看到。所以，廣告數量與廣告品質的組合，才是在我們平台上創造規模的關鍵。有些客戶知道這點，因此能成為我們的大額投放客戶。

The others have to learn it.

其他人必須學會。

I'm going to try to sneak one more in. So you're talking about significantly larger models, right, step function in scale. Are you generating enough data organically to feed the models? You did talk a little bit, Adam, about maybe opportunities to acquire data. I'm assuming you'd be looking for customer data, transaction data.

我想再偷偷塞進一個問題。所以你談的是顯著更大的模型，對吧，在規模上有一個階躍式的提升。你們是否能有足夠的有機生成資料來餵養這些模型？你確實稍微提到過，Adam，可能有機會去收購資料。我猜你們會尋找客戶資料、交易資料。

What types of data to be most helpful? Does it need to be evergreen? Do you need it even are you generating enough organically?

哪些類型的資料最有幫助？它需要是可長期持續更新（evergreen）的嗎？你們需要它嗎，還是你們有機生成的資料已經足夠？

I mean, look, like obviously, this quarter, we had a fantastic quarter, best we've ever had with the data that we had. So there was no incremental data into the system, right? So there's, on the one hand, just processing the data you have available to you in smarter ways, can create uplifts. On the other hand, like these models just need more data. I mean there's -- that's consistent across any company in a modeling space.

我的意思是，你看，很明顯，本季我們表現非常出色，是我們在既有資料條件下做過最好的季度。所以系統裡並沒有新增的增量資料，對吧？因此一方面，只要用更聰明的方式處理你手上可用的資料，就能帶來提升。另一方面，這些模型就是需要更多資料。我的意思是——這在任何做建模的公司都是一致的。

And so getting more data into the system is important for us. We're not rush. The business is doing really well. Every customer we get live is a data relationship on top of a business revenue relationship. But long term, if we can add ways to get data into our system quicker, the business is going to have much more runway in the short term.

所以把更多資料導入系統對我們很重要。我們不急。業務做得非常好。我們每讓一位客戶上線，除了帶來營收關係之外，也同時建立了一段資料關係。但長期來看，如果我們能增加一些方式，更快把資料導入我們的系統，短期內這個業務就會有更長的跑道。

Shweta Khajuria, Wolfe.

Shweta Khajuria，Wolfe。

Okay. I have a high-level question. Some of my questions were asked. As an ad tech company for investors who are questioning the visibility of your business model and how much confidence you have in the durability of your top line growth. Is there something else incremental that you would consider sharing that give more confidence in terms of the way your product works, the visibility you can offer and the moat that you're building?

好的。我有一個高層次的問題。我有些問題已經被問到了。作為一家廣告科技公司，對於那些質疑你們商業模式可見性、以及你們對營收（top line）成長耐久性有多大信心的投資人而言，你是否願意再額外分享一些增量資訊，讓大家對你們產品的運作方式、你們能提供的可見性，以及你們正在建立的護城河更有信心？

Yes. I mean, look, I think over the last -- I don't know, it's been 16 quarters or something. We've had 15 phenomenal quarters. Our growth rate has been higher than almost any tech company in the world that's public at the scale that we operate at and our margins are probably second to none. And so you've got a business that, while is unpredictable in the future, and we can't really give KPIs to predict something that isn't there, a lot of it is tied to actually us being able to improve our technology.

是的。我的意思是，你看，我想在過去——我不知道，大概是16個季度左右。我們有15個非常出色的季度。以我們目前的營運規模來看，我們的成長率高於全球幾乎任何一家上市科技公司，而我們的利潤率可能也是數一數二。所以你看到的是一個業務：雖然未來仍然不可預測，而且我們也無法提供KPI去預測一個尚未發生的東西，但其中很大一部分其實取決於我們能否持續改善技術。

And then a lot of it is tied to new customer growth, but we're so early in that consumer segment, that there's nothing that's material there yet that can be predictive in the future. You sort of have to make your own guesses. I said this, I think I don't know, maybe 12 quarters ago, early in AXON 2. I said, when human beings tend to think about predicting something, everyone gets negative on the future. In our business, they can be negative and positive.

另外很大一部分取決於新客戶成長，但我們在那個消費者細分市場仍然非常早期，因此目前還沒有任何具實質性的東西可以用來預測未來。你某種程度上必須自己做判斷。我以前說過，我想我不知道，可能在12個季度前，在AXON 2早期。我說過，當人們試圖去預測某件事時，大家往往會對未來變得悲觀。在我們的業務裡，他們可能會悲觀，也可能會樂觀。

But almost all those quarters, I mean, I think it's been 14 out of 16 that we've beaten the high end of our guidance and not by an immaterial amount, we've been consistently outperforming what anyone would have expected. So I think a lot of this is over time. People have to go, this is a bigger ad tech company that's sort of independent ad tech without a first-party site, bigger than anything we've ever seen before. We're a very large business at this point. We create a lot of cash flow and we're delivering to our customers in a very large amount of scale, return on ad spend, tens of billions of dollars of revenue on the other side of the ad spend that they put into our platform in any given year.

但幾乎在那些季度裡，我的意思是，我想在16個季度中有14個季度，我們都超越了指引區間的上緣，而且不是小幅超越；我們一直在持續跑贏任何人原本會預期的表現。所以我認為很多事情需要時間。人們必須理解：這是一家更大型的廣告科技公司，是某種獨立的ad tech，沒有第一方網站，規模大到超過我們過去見過的任何案例。我們現在是一個非常大的業務。我們創造大量現金流，並且以非常大的規模為客戶帶來廣告投資報酬（return on ad spend）；在任何一年，客戶在我們平台投入的廣告支出，能在另一端帶來數百億美元的營收。

So we're a very large economic force at this point. That ad spend is creating GDP. It's creating new transactions that otherwise wouldn't have been there. So as a very large business, we're growing quickly. We can't give those KPIs that you're looking for because we're just not at that point of stability.

所以我們此刻是一股非常龐大的經濟力量。那筆廣告支出正在創造GDP，正在創造原本不會存在的新交易。因此，作為一家非常大的企業，我們仍在快速成長。我們無法提供你想要的那些KPI，因為我們還沒有到那個足夠穩定的階段。

But we have something that is a catalyst that is also very valuable. Our engineers improving our models and more data into the system is not predictable, but it can create very high upside.

但我們有一個同時也非常有價值的催化劑。我們的工程師改善模型、以及把更多資料導入系統，這些都不可預測，但可能帶來非常高的上行空間。

Ralph Schackart, William Blair.

Ralph Schackart，William Blair。

Adam, on the talk track, you talked about a reference this model step-change is as big as AXON 2.0, I think you said. I'm just curious, is it fair to think of this as a AXON 3.0. Just really trying to frame this model update because when the model transition to AXON 2.0, you had a fairly significant inflection point. Obviously, the business is a lot larger now. But just if you could help conceptualize how impactful this new model will be once it's fully rolled out and impact on the business would be great.

Adam，關於你剛才的說法，你提到這次模型的階躍式變化，規模跟AXON 2.0一樣大，我想你是這麼說的。我只是好奇，把它視為AXON 3.0是否合理？我主要是想替這次模型更新建立一個框架，因為當時模型轉換到AXON 2.0時，你們出現了一個相當顯著的拐點。顯然，現在業務規模大得多了。但如果你能幫我們概念化：一旦全面推出，這個新模型會有多大影響、對業務的影響會是什麼，那就太好了。

Yes. I'd put it from a technology standpoint in terms of technical complexity on that level in terms of business. Look, this is the next generation of the model. We're not a direct-to-consumer AI company. And so we don't increment the model.

是的。從技術角度來看，就技術複雜度而言，我會把它放在那個等級；至於業務層面——你看，這就是下一代模型。這是模型的下一代。我們不是一家直接面向消費者的AI公司。所以我們不會只是做小幅度的模型增量更新。

But the reality is this gives us a lot bigger path in the future in terms of processing more data and spinning up bigger clusters and hopefully getting better output. Team is really excited about what it means. We're still in partial rollout. So it has to roll through all of our models. There's still a lot of complex R&D work there.

但現實是，這讓我們在未來有更大的路徑：能處理更多資料、啟動更大的叢集，並希望得到更好的輸出結果。團隊對它所代表的意義非常興奮。我們仍在部分推出階段。所以它必須逐步覆蓋我們所有的模型。那裡仍然有很多複雜的研發工作。

But it gives us a much longer road map than the old models would have given us. So that gets the team really excited.

但它讓我們擁有比舊模型更長的產品/技術路線圖。所以這讓團隊非常興奮。

Great. One follow-up. Obviously, the results and I might speak for themselves. But after last quarter, there was a ton of competitive concern. Maybe just give us an update on the market and the competition as you see it today.

很好。我再追問一個。顯然，結果——我想它們可能已經不言自明。但在上個季度之後，市場上有很多對競爭的擔憂。也許你可以就你今天看到的市場與競爭態勢，給我們一個最新看法。

Yes. I mean, look, like the industry checks out there show you all. Other companies are performing really well in the market. We've said since we went public, this isn't zero-sum. Our prediction, the impression, what we price for a customer is going to be different than any other model.

是的。我的意思是，你看，外部的產業調查都已經告訴大家了。其他公司在市場上的表現也非常好。我們自從上市以來就一直說，這不是零和遊戲。我們的預測、曝光（impression）、以及我們為客戶定價的方式，會和任何其他模型都不同。

What makes this category great is that the sum total of all the models improving, give these customers better businesses. So it's never been the case that other companies growing should squeeze us. I said on the last earnings call, that was not something that actually happened and it was our own footfall. That now shows true to you all. In the face of other companies doing well and it's not just one, it's multiple all at the same time, we put up a massive quarter.

這個類別之所以很棒，是因為所有模型整體的持續改進，會讓這些客戶的生意變得更好。所以從來都不是其他公司成長就會擠壓我們。我在上一通財報電話會議上說過，那其實並沒有發生，而是我們自己踩到自己的腳。而現在你們也都看到了這一點。在其他公司表現良好、而且不只一家、是多家同時表現良好的情況下，我們仍然交出了一個非常亮眼的季度。

Bryan Smilek, JPMorgan.

Bryan Smilek，摩根大通。

I guess, just really curious, Adam, you had mentioned that the All-In Summit, I believe that gross spend flowing through AppLovin's ecosystem was approaching like $20 billion annualized. Can you just talk about going forward in order to achieve that 30% long-term growth algorithm overall? How much of that growth will be from greater realized achievements through this new model versus incrementally adding new customers, both on the gaming and consumer side. And then, I guess, Matt, more near term, obviously, good to see that performance in 3Q implied 8% to 10% sequential totally appreciate greater consumer seasonality into year-end. But just curious like what's embedded there from a gaming perspective as you continue to roll out the new model?

我想我只是很好奇，Adam，你先前提到在 All-In Summit 上，我相信流經 AppLovin 生態系的總支出（gross spend）年化接近約 200 億美元。你能否談談，展望未來，為了達成整體 30% 的長期成長演算法（growth algorithm），你們打算怎麼做？這些成長有多少會來自於透過這個新模型帶來更高的已實現成效（realized achievements），相較於逐步新增新客戶（不論是遊戲端或消費端）？另外我想問 Matt，較近期來看，很高興看到第三季（3Q）的表現，指引隱含約 8% 到 10% 的季增；我完全理解年底前消費端季節性會更強。但我也好奇，從遊戲的角度來看，隨著你們持續推出新模型，這裡面包含了哪些因素？

Yes, look, long-term growth for us like the components of are hard to predict. Of course, new customers coming on is now going to be a component, not in gaming. We have every mobile gaming company live probably. But in consumer, we're really early. The data in the system will improve the model architecture now gives us a long runway to improve it.

是的，你看，我們的長期成長，組成要素其實很難預測。當然，新客戶的加入現在會成為一個組成部分，但不是在遊戲端。我們大概已經讓所有行動遊戲公司都上線了。但在消費端，我們還非常早期。系統中的資料會持續改善，而模型架構現在也讓我們有很長的跑道可以去優化它。

So we don't know. It's also interesting that every quarter, it can be different. If you verticalize the business, we could have a lift on the gaming side that doesn't apply to the consumer side, and that takes off one quarter and the next quarter flips. What gets us really excited is that consumer is growing very quickly and it's starting to become more and more a part of our business. That on its own has much longer runway because we're so lightly penetrated into the market of customers.

所以我們也不確定。而且有趣的是，每一季可能都不一樣。如果我們把業務做垂直化（verticalize），我們可能在遊戲端看到提升，但這不一定適用於消費端；某一季可能遊戲端起飛，下一季又反過來。真正讓我們很興奮的是，消費端成長非常快，而且它開始在我們的業務中占比越來越高。單就這一點而言，因為我們在客戶市場的滲透率還很低，所以它本身就有更長的成長跑道。

as each quarter goes on, more customers find out about us, that business is going to grow really quickly.

隨著每一季推進，會有更多客戶認識我們，那塊業務就會成長得非常快。

Yes. On the guide, Bryan, we don't break out, obviously, between gaming and consumer. I won't give you the specific splits. But what we've assumed for gaming and for consumer is consistent. I mean, there's seasonality, right, as we look into Q4, obviously, both are strong for different reasons.

是的。關於指引（guide），Bryan，我們當然不會把遊戲和消費端拆開揭露。我也不會給你具體的拆分比例。但我們對遊戲與消費端的假設是一致的。我的意思是，當我們看向第四季（Q4）時，確實存在季節性，顯然兩邊都會因不同原因而走強。

You have the consumer period, right, of Black Friday, Cyber Monday and then the holidays, you also have on the gaming side of the business, there's seasonality as well, and Q4 is always strong. So we've assumed that within the guide and then obviously, the model improvements that we've had thus far, those are reflected in the 8% to 10% as well.

消費端有黑色星期五、網購星期一，接著是節慶假期；同時在遊戲端業務也有季節性，而第四季一向都很強。所以我們已在指引中納入這些假設；此外，我們迄今為止所做的模型改善，也同樣反映在那 8% 到 10% 的數字之中。

Alex Brondolo, Wells Fargo.

Alex Brondolo，富國銀行。

Appreciate the question. Maybe two for me. I think, first, you mentioned in response to a question about e-commerce or web advertising, the idea of a data acquisition. And I think that there's this kind of idea percolating that there could be tuck-in M&A leverage to kind of accelerate the capabilities in e-commerce. And so perhaps if you could talk about what's the ideal acquisition candidate or what the ideal data you're looking for is that would help people kind of frame that opportunity?

謝謝讓我提問。我這邊可能有兩個問題。首先，我想你在回應關於電商或網路廣告的問題時，提到了「資料收購」（data acquisition）的概念。我覺得市場上有一種想法正在發酵：可能透過補強型併購（tuck-in M&A）來加速電商能力的建置。所以也許你可以談談，理想的收購標的會是什麼，或你們在尋找的理想資料類型是什麼，好讓大家能更好地理解這個機會？

And then secondly, you mentioned the Pixel data, so I guess it's fair game. Clearly, APAC, most of the inflection in the Pixel data, it seems like most of the inflection in e-commerce customer growth is coming from APAC. Why do you think that region is finding kind of stronger product market fit first relative to the U.S. and Europe?

第二，你提到了 Pixel 資料，所以我想這應該可以談。很明顯，Pixel 資料的大部分拐點（inflection）似乎都在亞太（APAC）；而且看起來電商客戶成長的拐點也多來自亞太。你認為為什麼相較於美國與歐洲，那個區域會更早找到更強的產品市場契合度（product-market fit）？

Okay. I want to take the second first. We've got sales teams on both North America, I guess, EMEA as well. But APAC and the North America team are the 2 biggest. We've always traditionally been focused on both those two markets, which are big ones on the Western advertising side.

好。我先回答第二個。我們在北美有銷售團隊，我想在 EMEA 也有。但亞太與北美團隊是最大的兩個。我們傳統上一直聚焦在這兩個市場，這兩個在西方廣告市場都是很大的市場。

The APAC customers because they're tapping into the Asian supply chain. A lot of times, they spin up sites that are specific to certain type of products that aren't buying elsewhere. So for them, if they go spend $1,000 on us. They either make money or they lose money, but it's very easy because there's no attribution complexity. So sometimes these Pixel sites are brand-new sites that just didn't exist that are spun up specifically for our channel.

亞太客戶因為能連結到亞洲供應鏈。很多時候，他們會建立一些針對特定產品類型的網站，而這些產品在其他地方不一定買得到。所以對他們來說，如果在我們這裡花 1,000 美元，他們要嘛賺錢、要嘛賠錢，但判斷非常容易，因為沒有歸因（attribution）複雜度。因此有時候這些 Pixel 網站是全新的網站，原本並不存在，而是專門為了我們這個渠道而建立起來的。

North American customers are bigger enterprises. They're not going to go spin up the next version of Wayfair, right? It's one domain. And so you -- it doesn't -- I wouldn't read into the counts as I'm seeing more Pixel counts out of APAC, so they're getting more penetration into the customer base over there. It might be just one is creating a lot because they don't have the same attribution complexity that some of the Western companies do because they're just longer-standing brands.

北美客戶則是更大型的企業。他們不會去打造下一個 Wayfair 對吧？就是一個網域（domain）。所以你——我不會因為看到亞太的 Pixel 數量更多，就解讀成我們在那邊的客戶滲透更高。可能只是因為他們不會像一些西方公司那樣有同樣的歸因複雜度；西方公司多半是經營更久的品牌。

On the acquiring data, point. The M&A that we did in the past that maybe is our best is when we bought MoPub. We were already succeeding with MAX, so it's a great product. And at the time, no one understood it. This crazy company bought a company and put it out of business.

關於收購資料這點。我們過去做過的併購中，可能最成功的是我們收購 MoPub 的那次。我們當時已經靠 MAX 做得很成功，所以那是一個很棒的產品。而在當時，沒有人理解它。大家覺得這家公司很瘋，買了一家公司然後把它關掉。

I mean we put it out of business in 90 days. But we got all the customers over to our platform because we just build great products. And by doing that, we ended up with what MAX is today, which everyone knows is a really valuable business. That transaction has paid for itself many times over now. What we would look for is can we buy data relationships, customer relationships that are valuable for us.

我的意思是，我們在 90 天內就把它關掉了。但我們把所有客戶都轉到我們的平台上，因為我們就是能打造很棒的產品。而透過這麼做，我們最終成就了今天的 MAX——大家都知道那是一個非常有價值的業務。那筆交易到現在已經回本很多倍了。我們會尋找的是：我們能否買到對我們有價值的資料關係、客戶關係。

We're going to get all these customers over the next decade anyways. But if we could buy it and get the shot in the arm to bring all these customers onto our platform quickly and get all that data into our platform quickly, it will accelerate this business to get through what could be a decade's worth of work in a much shorter amount of time. Same sort of structure is that kind of deal is what we would look for.

反正未來十年我們終究會把這些客戶都拿下。但如果我們能透過收購，打一劑強心針，把這些客戶更快帶到我們的平台上，並把所有資料更快導入我們的平台，就能加速這門業務——把原本可能需要十年完成的工作，在更短時間內完成。我們會尋找的就是這種結構的交易。

Omar Dessouky, BofA.

Omar Dessouky，美國銀行（BofA）。

In your old models, you used to continuously retrain on live impression data and that resulted in a steady 3% to 5% self-learning growth. and that's distinct from engineer directed model uplifts. You didn't address it specifically on the last conference call. So can you tell us if you see that same 3% to 5% tailwind from self-learning and intra-quarter model retraining off of live impression data. Did you see it in Q3?

在你們舊的模型中，你們會持續用即時曝光（live impression）資料進行再訓練，並帶來穩定的 3% 到 5% 自我學習成長。而這與工程師主導的模型提升（engineer-directed model uplifts）是不同的。你們在上一通法說電話中沒有特別回應這點。所以你能否告訴我們，你們是否仍看到同樣來自自我學習、以及在季度內用即時曝光資料再訓練模型所帶來的 3% 到 5% 順風（tailwind）？第三季（Q3）有看到嗎？

Is it assumed in your Q4 guidance? And kind of looking forward, how much visibility do you have as to that component of growth? Or are you just wholly reliant on engineer-directed model enhancements going forward?

第四季（Q4）指引中有把它納入假設嗎？以及往前看，對於這個成長組成部分，你們有多少能見度？還是說，未來你們將完全仰賴工程師主導的模型強化？

Look, in all these models, you have this sort of reinforcement that happens. You get a prediction, you get data that comes into the system and they get smarter. This is true in any sort of category of companies that are writing these deep learning models at this point. So you don't actually know is it 3%? Is it 3.2%?

你看，在所有這些模型中，都會發生某種強化效應。你得到一個預測，資料進入系統，模型就會變得更聰明。在目前任何一類撰寫這些深度學習模型的公司中，這都是真的。所以你其實不知道是 3%？還是 3.2%？

Is it 4%? You can't really track what the model is improving on its own. But there is always going to be this self-learning notion that exists in a company that's running a model, creating a prediction, reassessing its own prediction and hopefully getting smarter as it goes forward. that's what these models are built to do, right? Like they're structured similar to the human brain.

是 4%？你其實無法追蹤模型自行在改進什麼。但在一家運行模型、產生預測、重新評估自身預測並希望隨著時間推進變得更聰明的公司裡，總是會存在這種自我學習的概念。這就是這些模型被設計來做的事，對吧？它們的結構類似人類大腦。

We, as humans, learn from our memories and get smarter as we go forward. These models do the same exact thing. So we just wouldn't split it up into this is a growth driver. Let's plug this number in here. And then on top of it, let's put a model enhancement because we talked about back in the day also we'd have one big one a year.

我們人類會從記憶中學習，並在往前走的過程中變得更聰明。這些模型做的完全一樣。所以我們不會把它拆成「這是一個成長驅動因素」。「把這個數字填在這裡」。然後再加上一個「模型增強」，因為我們以前也談過，過去一年可能只有一次大的提升。

Well, probably over the last 16 quarters, we've had multiple AV test lifts pretty much every single one of those quarters, except the last one that contributed to very outsized growth for the business. The stuff is not all that predictable nor is it able to be dissected down to a specific percentage, but it all compounds into the business that we have today.

嗯，大概在過去 16 個季度裡，我們幾乎每一個季度都有多次 A/B 測試帶來的提升，除了上一個季度之外；而這些提升為業務帶來了非常顯著的成長。這些事情既不那麼可預測，也無法拆解到某個特定百分比，但它們都會累積，形成我們今天的業務。

Jim Callahan, Piper Sandler.

Jim Callahan，Piper Sandler。

Adam, I think last year, you mentioned supply growth across MAX growing double digits. Any comment on what that looks like this year or to 2027.

Adam，我記得去年你提到 MAX 的供給成長是兩位數。能否評論一下今年的情況，或是到 2027 年會是什麼樣子？

Yes. I mean, look, it's still strong double digits. I think last quarter, we told you in a single quarter, it grew double digits. So it's a very fast-growing space in the advertising ecosystem, where I think like most ad spaces in terms of the supply growth and value of that supply aren't growing all that fast, probably like very low double digits to possibly single digits, this MAX market is growing at very strong double-digit growth rate every year.

是的。我的意思是，你看，仍然是強勁的兩位數。我想上一季我們告訴過你，在單一季度內就成長了兩位數。所以在廣告生態系中，這是一個成長非常快的領域；我認為多數廣告版位在供給成長以及供給價值方面其實成長沒那麼快，可能是很低的兩位數甚至個位數，但這個 MAX 市場每年都以非常強勁的兩位數成長率在成長。

Great. And then anything on the cost of supply acquisition. We've kind of heard this come up in checks a little bit that the cost of impression broadly has gone up. Any way to frame how that looks today and kind of how that could look into next year as well?

很好。那供給取得成本方面有什麼看法嗎？我們在一些調查/訪查中有稍微聽到這點：整體而言，每次曝光的成本上升了。能否說明一下目前的狀況，以及明年可能會如何發展？

Yes. I mean, look, there's two drivers to MAX growth, more supply and the cost going up. So cost effectively, company is getting better at monetizing the impressions. The value of the auction goes up, the publisher makes more. Now on the other side of that, we're either going to have to pay out more for publisher inventory or improve our own technology at that same rate or improve our own technology more than that rate to grow our business even faster than that market.

是的。你看，MAX 成長有兩個驅動因素：更多供給，以及成本上升。所以從成本角度來看，公司在把曝光變現方面做得更好。拍賣的價值上升，出版商（publisher）賺得更多。而在另一面，我們要嘛得為出版商的庫存支付更多，要嘛就必須以同樣的速度提升我們自己的技術，或是以更快的速度提升我們的技術，才能讓我們的業務成長得比市場更快。

And so a lot of the innovation that we've had over the last 4 years has been the latter. We've been improving our technology better than the market, and therefore, our business is growing much faster than the market is growing.

因此，我們過去 4 年的許多創新都屬於後者。我們提升技術的速度快於市場，所以我們的業務成長也遠快於市場成長。

Vasily Karasyov, Cannonball.

Vasily Karasyov，Cannonball。

Adam, I wanted to ask you to comment on the take rate directionally maybe compared to last quarter and maybe year-on-year, whatever is easier? And then the question I had with this model, separate models for consumer and gaming, is there would there be a reason why the take rate fluctuation on one would be greater than the other. Can you help us like frame it for us, how to think about it?

Adam，我想請你就 take rate（抽成率/分成率）的方向性做個評論，可能是相較上一季，或是年對年，哪個比較容易都可以？另外我對這個模型有個問題：消費者與遊戲是分開的模型，那是否有理由使得其中一個的 take rate 波動會比另一個更大？你能幫我們建立一個思考框架嗎？

Yes. I mean, look, we don't talk about take rate in the business. And frankly, it happens after the fact when you have a conversion rate that's a little bit over 1%. If we were buying on 1,000 impressions, somewhere around like 987 of them we're losing money on. So there's no take rate, we lost everything.

是的。你看，我們不談業務中的 take rate。坦白說，這是在事後才會發生的事情，當你的轉換率只有略高於 1% 的時候。如果我們買了 1,000 次曝光，其中大概有 987 次我們是在虧錢。所以沒有什麼 take rate，我們什麼都虧掉了。

On the other ones, you have almost an infinite take rate because you're making a ton of revenue with a very small amount of outlay. So it's just not a function that we optimize to. And it really does tie to the point that I made earlier. If our engineers created a massive uplift in the next two weeks, we'd be able to take more because we out-innovated to everyone else in the market. If we innovate at the rate of the market, the market pricing stays consistent, we're sort of where we always are, which is pretty consistent.

而在其他情況下，你幾乎會有無限大的 take rate，因為你用非常小的支出就賺到大量收入。所以這不是我們用來最佳化的指標。而這也確實呼應我先前提到的重點。如果我們的工程師在接下來兩週做出巨大的提升，我們就能拿更多，因為我們在創新上超越了市場上其他所有人。如果我們的創新速度跟市場一樣，市場定價就會維持一致，我們大致就會維持在一貫的位置，也就是相當穩定。

And then if we fall way behind, we just disappear for five years, our take rate will contract because the market pricing will get better than the rate of improvement on our technology. Now what you've seen with our team and the opportunity that we put in front of us with these different business lines and where we are in them, is that we're really good at executing.

然後如果我們遠遠落後——比如我們消失五年——我們的 take rate 就會收縮，因為市場定價的改善速度會超過我們技術改進的速度。而你從我們團隊、以及我們在這些不同業務線上所面對的機會與所處的位置所看到的是：我們真的很擅長執行。

We've been able to execute better than others in the ecosystem for a very long amount of time now, our team is very seasoned on improving these technologies. And when they have R&D breakthroughs like we just had, usually those end up being before everyone else. So before anyone else even understands how to have a technological breakthrough like that, we're many quarters or years ahead.

我們長期以來在生態系中一直能比其他人執行得更好；我們的團隊在改進這些技術方面非常老練。而當他們像我們剛剛那樣取得研發突破時，通常都會發生在其他人之前。所以在其他人甚至還沒弄懂如何達成那樣的技術突破之前，我們就已經領先很多個季度甚至好幾年。

And is there something inherent to consumer versus gaming that would drive different...

那消費者相較於遊戲，有沒有什麼內在因素會驅動不同的……

It will be consistently, again, like similar. So I mean the way this is why I always say it's like one auction. Let's say, gaming has a predictive value of $20 CPM, so $0.02 for the impression. And a consumer comes in and wins that impression at $21. It's so close to each other.

它會一貫地、再說一次，類似。所以我的意思是，這也是為什麼我總說它就像是一個拍賣。假設遊戲的預測價值是 20 美元 CPM，也就是每次曝光 0.02 美元。然後某個消費者（類別）進來，以 21 美元贏得那次曝光。兩者非常接近。

You're paying out to the publisher. And if the model is very predictively accurate, you're not going to have like one bid at $50 and the other one bid at $20, they're pretty competitive to each other.

你要支付給出版商。而如果模型的預測非常準確，你不會出現一個出價 50 美元、另一個出價 20 美元那種情況；它們彼此之間會相當有競爭力。

Tim Nollen, SSR.

Tim Nollen，SSR。

I have a question about data and it's about data provenance versus data processing. Question is really who owns the data in your transactions? Is it your mediation platform that processes the data? Or is it like the original game developer or the monetization platform that handled the transaction?

我有一個關於資料的問題，涉及資料來源/溯源（data provenance）與資料處理（data processing）。問題其實是：在你們的交易中，資料的所有權屬於誰？是你們處理資料的中介（mediation）平台嗎？還是像原始的遊戲開發商，或是處理該筆交易的變現平台？

What data are you talking about? Just mediation data? I don't we have a lot of data in our system. So I didn't follow exactly...

你指的是什麼資料？只是 mediation 的資料嗎？我不知道——我們系統裡有很多資料。所以我沒有完全跟上你的意思……

Yes. Well, the data -- the data originates through transactions. However, whoever processes them, but your mediation platform kind of creates these packages and you have the sort of new package data. So I guess my question is, who owns that data? Is it...

是的。嗯，資料——資料是透過交易產生的。不過，不管是誰在處理它們，你們的中介平台某種程度上會把它們打包，於是你們就有了某種新的封包化資料。所以我想問的是，誰擁有那些資料？它是……

What is packaged data? If you just give me an example of like an ad served, it will help me answer it for you.

什麼是封包化資料？如果你能給我一個例子，例如投放了一則廣告，會有助於我回答你的問題。

So if an ad is served and you process the transaction versus somebody else processes the transaction, you can see the data that is going through these transactions. You incorporate these data into your media -- into your mediation platform. So I guess my question is, is the data -- does the data belong to the original game developer or ad tech company that processes the transaction or through the mediation platform, do you gain providence over that data? Does that make sense?

所以如果投放了一則廣告，而你們處理該筆交易、相對於由其他人處理該筆交易，你們就能看到在這些交易中流動的資料。你們把這些資料納入你們的媒體——納入你們的中介平台。所以我想問的是，這些資料——這些資料是屬於原始的遊戲開發商或處理該交易的廣告技術公司，還是透過中介平台，你們就取得對那些資料的支配權？這樣說有道理嗎？

I think there's not really because I think there's a disconnect on what you think we have access to mediation. When we have a mediation auction run, there's every company bids into the auction. The only data that's available. So to the MAX system, of course, we see all the bids. We don't know what people bid for.

我覺得其實不是，因為我認為你對我們在中介中能接觸到什麼資料有些誤解。當我們進行一次中介競價時，每家公司都會對該次拍賣出價。唯一可用的資料是——所以對 MAX 系統而言，當然我們能看到所有出價。我們不知道大家是為什麼而出價。

But the only data that's available to our AXON models or our system are did we win the ad? Or did we not win the ad? What was the price for the winning auction? And then everyone who loses gets the price for the losing auction. That's it.

但對我們的 AXON 模型或我們的系統而言，唯一可用的資料是：我們是否贏得該次廣告？或是我們沒有贏得該次廣告？贏得拍賣的成交價格是多少？然後所有輸掉的人會拿到輸標的價格。就這些。

No one should have access to what other people serve. We don't have visibility into an auction cleared, someone else want it. They serve exactly X. That exact X is the ownership of the advertising company. The company that wrote a model that created that process a ton of data has a relationship with customers and consumers to show an advertisement.

任何人都不應該能存取別人投放了什麼。我們也無法看見某次拍賣結算後，別人贏了。他們投放了某個特定的 X。那個特定的 X 的所有權屬於該廣告公司。那家撰寫模型、建立流程並處理大量資料的公司，與客戶與消費者之間有關係，用以展示廣告。

That's their data. It's an output of their models that remains with them. And so there's nothing that like gives us magical data access from the mediation. In fact, when we built MAX, what we did was we commoditized it all. We said everyone is going to have the same access to data as we will as a bidder into MAX because what we wanted was a fair marketplace.

那是他們的資料。那是他們模型的輸出，並且仍然由他們保有。所以並不存在什麼能讓我們透過中介就神奇地取得資料存取權。事實上，當我們打造 MAX 時，我們做的是把它商品化。我們說，每個人都會擁有與我們作為 MAX 競價方相同的資料存取權，因為我們想要的是一個公平的市場。

The reason MAX out monetizes the rest of the world of mobile gaming is because we made it fair, and that fairness is what drove the price of every single impression to be the highest possible. Does that answer your question?

MAX 之所以能比行動遊戲世界其他地方帶來更高的變現，是因為我們讓它變得公平，而這種公平性推動每一次曝光的價格達到可能的最高水準。這有回答你的問題嗎？

That answers it. Thanks very much, Adam.

有回答到。非常感謝你，Adam。

Matthew Swanson, RBC.

RBC 的 Matthew Swanson。

Great. Congratulations on the quarter. I guess, on the consumer side, Adam, you mentioned there's some handholding in terms of like the initial on-ramp, if you're not a more sophisticated customer. I guess, a 2-parter there. One, how much of it is like on you versus them in terms of improving that initial performance?

很好。恭喜本季表現。我想在消費者端，Adam，你提到如果客戶不是比較成熟的客戶，在最初導入上會需要一些手把手的協助。我想這裡有兩個問題。第一，在改善初期表現方面，有多少是你們的責任、又有多少是他們的責任？

Is it still about the models? Is it about them learning how to kind of better optimize their usage of the platform. And then secondarily, when you're kind of targeting customers, it seems like these real sophisticated users are more likely to generate revenue immediately. Is that more the priority? Or are you looking for kind of the best long-term partners for AppLovin?

這仍然主要取決於模型嗎？還是取決於他們學會如何更好地優化他們對平台的使用？第二，當你們在鎖定客戶時，看起來那些真正成熟的使用者更可能立刻帶來營收。這是不是更優先？或者你們是在尋找對 AppLovin 而言最佳的長期合作夥伴？

Yes. And so I mean, I think a lot of this is a learning curve. We're new platform. So a lot of these customers have not bought in games. They're not used to an ad that actually gets watched, and certainly not used to an ad that's watched for that much time.

是的。所以我的意思是，我認為其中很多是一個學習曲線。我們是一個新平台。因此很多這些客戶以前沒有在遊戲裡買過廣告。他們不習慣有一種廣告真的會被看完，當然也不習慣一則廣告會被觀看那麼久。

And if you have a shop, I mean, like think about like when I sometimes I think about like we're asking the customer to create 60 seconds of content to showcase a shop. I mean, what do you even show in that ad? It's like shops aren't that exciting. On the gaming side, they're creating mini games, so someone's playing a preview of their game. It's straightforward.

如果你有一家店，我的意思是——想想看，有時我會想，我們是在要求客戶製作 60 秒的內容來展示一家店。那你到底要在那支廣告裡展示什麼？店其實沒那麼令人興奮。但在遊戲端，他們是在做迷你遊戲，所以有人是在玩他們遊戲的預覽。這就很直接。

So this is just there's a learning process here that has to happen. So that the media buyers really understand the framework. And then once they do and more and more media buyers understand it and more agencies are trained, then it will become standardized. There will be playbooks for our platform. but it's not dissimilar to buying on social 10 years ago.

所以這裡就是需要一個學習過程。讓媒體採買方真正理解這個框架。然後一旦他們理解了，越來越多媒體採買方理解它、更多代理商受過訓練，它就會變得標準化。我們的平台會有操作手冊（playbooks）。但這和 10 年前在社群平台上投放廣告並沒有太大不同。

It was brand new. The scale was way smaller. People didn't know what to do. Now everyone knows what to do because if you don't, you're probably not scaling your business. And on the priority of the type of customers we go for, we are looking for the more sophisticated ones.

那時是全新的。規模小得多。大家不知道該怎麼做。現在每個人都知道該怎麼做，因為如果你不知道，你大概就無法擴大你的業務規模。至於我們優先鎖定哪種類型的客戶，我們在尋找更成熟的客戶。

That's why these partnership deals make a lot of sense rather than just run AppLovin ads everywhere through strategic partnership deals, we can get real customers coming on.

這也是為什麼這些合作夥伴協議非常合理；與其只是透過策略合作把 AppLovin 的廣告到處投放，我們可以透過策略合作讓真正的客戶上線。

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

以上為本季問答環節的結束。感謝各位今天加入我們。祝各位下午愉快。