--- author: All-In Podcast date: '2026-09-20' guest: '' layout: post.njk source: https://www.youtube.com/watch?v=JtomF4bGxHs speaker: All-In Podcast tags: - mobile-advertising - recommendation-system - machine-learning - capital-allocation - founder-mindset title: 对话AppLovin创始人Adam Foroughi：从股价暴跌92%到千亿广告帝国的AI逆袭之路 summary: AppLovin联合创始人兼CEO Adam Foroughi深度复盘了公司如何依托自研深度学习引擎AXON从移动游戏广告切入电商推荐，在股价暴跌92%后逆势回购并实现85%的EBITDA利润率，展现了顶尖科技创始人的工程与资本运作哲学。 insight: '' draft: true series: '' category: ai-application area: tech-engineering project: [] people: - Adam Foroughi companies_orgs: - AppLovin - Meta products_models: - AXON media_books: [] status: evergreen --- ### 隐形广告巨头 **主持人**: **Adam** 可能是大家从未听说过的最优秀的创业者。在超过 10 万款手机游戏背后，潜藏着一个正在悄然超越 **Facebook** 电商广告效果的广告平台。在过去上千家 IPO 的公司中，市值最高、最具价值的就是 **AppLovin**。创始人的心态必须是：追逐胜利。
Original English **Host**: Adam is probably the best founder no one's ever heard of. There's an ad platform hiding inside 100,000 mobile games and is quietly outperforming Facebook ads for e-commerce brands. Of all those thousand plus IPOs, the number one most valuable is AppL. Appe. The founder mentality has got to be Chase winning.
**主持人**: 他们今年将创造大约 **60 亿美元** 的现金流。
Original English **Host**: They're going to print something like $6 billion in cash this year.
**Adam Foroughi**: 在一个人们无法理解的世界里，大家会觉得你在作弊，而不是意识到你构建了世界上见过的最酷的技术之一。
Original English **Adam Foroughi**: In a world where things don't make sense, people think you're cheating. instead of realizing you built one of the cooler technologies the world's ever seen.
**主持人**: 让我们欢迎 **Adam Foroughi**。
Original English **Host**: Please welcome Adam Farugi.
**主持人**: 好的，欢迎 Adam。嘿，兄弟。
Original English **Host**: All right. Welcome, Adam. Hey, man.
**Adam Foroughi**: 各位大佬好。
Original English **Adam Foroughi**: Big man.
**主持人**: 最近怎么样，兄弟？很高兴见到你。
Original English **Host**: How you doing, bro? Good to see you.
**Adam Foroughi**: 我也是，见到你们很高兴。
Original English **Adam Foroughi**: Likewise.
**主持人**: Adam，感谢你的到来。我们觉得这次交流非常有意义，因为很多人并不经常谈论你——在你的业务发展过程中，你很少出现在新闻头条上。你的经营风格几乎就像是远离媒体一样。你很少接受媒体采访，也不经常公开露面大谈特谈公司，但这是一项极其不可思议的业务。你能否向观众介绍一下 **AppLovin** 究竟是什么，并顺便帮我们梳理一下这个市场的格局？
Original English **Host**: Adam, thanks for being here. Um, we thought it'd be really great to chat because a lot of people don't talk as much about you're not in the headlines all the time with your business. You're operating your business almost like absent media. You don't do a lot of press. You don't get out there a lot talking about the company, but it's such an incredible business. Can you just tell the audience kind of what Apploving is and maybe also frame up the market a little bit for us?
**Adam Foroughi**: 好的，完全没问题。我认为，我们在早期没有依靠风险投资（VC）就建立起了一家规模庞大的公司，这一事实造就了我们只能保持低调、潜心构建业务的环境。当然，起初那个听起来有点滑稽的名字在宣传上也确实没帮上我们太多忙。 但从本质上讲，我们是一家**广告技术公司**，致力于帮助移动游戏开发者在这个领域实现变现。如今大众普遍没有意识到的是，移动游戏整个生态的规模到底有多庞大。每天有超过 **10 亿人** 在玩移动休闲游戏。这些玩家全都是成年人、家庭经济支柱，整个市场的机会极其巨大。 我们在去年 1 月——也就是大约两年前——披露过，仅在我们自己的平台上，每年的广告支出就达到了 **110 亿美元**。自那时起，我们保持了大约每年 **60%** 的同比增长。如果把这个数字粗略折算成如今一个整数，年流水规模大约达到了 **200 亿美元**。 我们在这项业务中的抽成率非常高。我们构建的技术是世界上最复杂的**机器学习模型**之一，正是这个模型在驱动着这些广告的精准匹配与转化。
Original English **Adam Foroughi**: Yeah, totally. I I think the the fact that we were able to build a very big company without having VC funding at the early stage created this world where we just had to build quietly and and then obviously the goofy name didn't help us all that much as well. But what we are ultimately is an advertising company that's helping mobile game developers monetize that space. Now what people don't realize is just how big the mobile gaming universe has become. You've got over a billion people a day playing mobile casual games. These are all adults, heads of households, and the scale of the opportunity is just humongous. We disclosed last January, so nearly two years ago, that on our own platform, there was 11 billion a year of ad spend. Since then, we've grown 60% year-over-year roughly. And so, if you gross that up to a nice round number today, you get 20 billion. Now, we're taking a big take rate on that business. The technology we've built is one of the more complex machine learning models in the world driving those ads.
--- ### 广告与AI演进 **主持人**: 互联网广告的第一波浪潮在很多方面其实是催化许多关键技术诞生的火花，这些技术随后扩散并深刻影响了整个世界。包括我们现在所看到的大模型基础设施，追根溯源，很大程度上都源于广告网络构建的需求。你能否谈谈广告与 AI 之间这种密不可分的关系？
Original English **Host**: The the first wave of internet advertising was in many ways the spark for a lot of critical technologies that then sort of diffused out into the world. And even today, a lot of the large models that we think about and infrastructure were built around ad networks. Can you talk about that bridge?
**Adam Foroughi**: 是的。广告技术就像是 **ML 1.0**，但它确实是当今驱动 AI 发展的所有核心技术的首个实际落地场景。最美妙的地方在于，如果你拥有一个极其庞大的平台，比如 **Google**、**Meta**，或者在移动广告领域的我们，你就拥有了即时商业化的飞轮。 这意味着，只要你的算法能力提升了 **1%**，这些收益就会以几乎 **100% 的毛利率**直接转化为真金白银的收入。世界上几乎没有任何其他业务能拥有如此直接的算力回报机制：你只要投入资本购买算力、聘请顶尖的工程师，不断迭代模型，只要产出哪怕一点点精度的提升，就能立刻获得巨大的财务回报，进而让你有资本购买更多的算力。这正是过去 20 年里支撑 Google 和 Meta 持续巨额资本支出的底层商业引擎。
Original English **Adam Foroughi**: Yeah, I mean advertising is like ML 1.0 but really was the first implementation of all these technologies that now are are driving AI today and the thing that was so beautiful about it was if you have a massive platform like Google does, like Meta does, like we do in mobile advertising, you have an immediate business to monetize. And that meant that if you got a 1% lift in your algorithms, that went straight to your top line with 100% margin. And there's almost no other business in the world where you can invest CapEx into computing power, hire great engineers, and that innovation loop creates an immediate return on investment to buy more computing power. And that's what funded the CapEx for Google and Meta over the last 20 years.
**主持人**: 相比 2006 年，2026 年人类对广告的反应是否存在某种广泛的行为模式演进？在过去 20 年里，消费者的广告心理是否有过某种演变轨迹？
Original English **Host**: Is it is it true that there's just a broad-based behavior around humans reaction to ads in 2026 versus 2006? Like, has there been an evolutionary arc of human reaction to ads over the last 20 years?
**Adam Foroughi**: 是的，这非常有趣。我的职业生涯始于 2005 年，所以我亲眼目睹了那个年代的广告——完全是一堆垃圾，全都是垃圾邮件式的弹窗与骚扰广告，当时的技术根本跟不上。 但随着时间的推移，广告平台不断演进，推荐系统变得极为先进。如今，当你在 Meta、TikTok 或者在玩手机游戏时看到的广告，通常都是极具相关性、引人入胜的内容，甚至让你觉得这本身就是一种内容消费。广告本身变成了一种发现机制。如果你把优质、精准的内容推送给合适的用户，它就不再是传统意义上令人厌烦的骚扰广告，而是一种极佳的**产品发现体验**。
Original English **Adam Foroughi**: Yeah, it's it's interesting. I started my career in 2005, so I saw the ads back then, complete garbage. It was all spam and the technologies just weren't there. But over time, the platforms evolved and the recommendation systems became so good. Today, the ads that you see on Meta, the ads you see on TikTok, the ads you see in mobile gaming, they're relevant, they're engaging. You actually enjoy looking at the content. It becomes a discovery mechanism. And if you're putting great content in front of a consumer that's relevant to them, it's no longer an intrusive ad. It's actually a great experience.
--- ### 搜索与发现广告 **主持人**: 业界对于 AI 对传统广告网络的影响一直有很多担忧和讨论，特别是对 Google 的搜索界面形态。与此同时，**OpenAI** 也开始尝试广告产品。人们总在讨论搜索广告是否会被生成式 AI 颠覆。你是怎么看待这种格局变化的？
Original English **Host**: There's been a lot of hand ringing about the impact AI will have on um the ad networks, specifically Google's interface. And OpenAI has an ad product that they're experimenting with. Um what is how do you think about where all that goes? Because it does feel like it's a zero sum game in a lot of ways.
**Adam Foroughi**: 广告通常分为两个完全不同的阵营。一部分是**漏斗底部（Bottom-of-Funnel）**的意图搜索广告，即消费者已经清楚知道自己想买什么，比如在 Google 上搜索特定型号的跑鞋或航班。这一块是 Google 极为擅长的领域，也是目前大家争论 OpenAI 或聊天机器人界面是否会分流用户搜索流量的核心战场。 但广告的另一半版图是**发现类广告（Discovery）**，也就是**漏斗顶部与中部**的体验。当用户打开社交软件或玩游戏时，他们并没有带着明确的购买意图，而是处在闲暇状态。在这一场景下，广告技术的作用是洞察用户的潜在兴趣，主动向他们推荐他们从未听说过但可能会极度喜爱的商品或游戏。
Original English **Adam Foroughi**: Yeah, I mean there's two sides of advertising. One one part of it is bottom offunnel advertising where a consumer sort of knows what they want to buy, they go search for it. That's Google's domain, and that's the area that people are questioning right now with OpenAI and conversational search. But the other half of the market is discovery. When you're browsing social media, when you're playing a game, you're not there with intent to buy something.
**主持人**: 本质上就是“发现”机制。
Original English **Host**: Discovery basically,
**Adam Foroughi**: 完全就是纯粹的发现。这正是 **Meta** 在广告业务上如此强大的原因，也是我们 **AppLovin** 所追求和深耕的方向。你创造了让用户惊喜的发现时刻。如今消费者大部分的购物行为并不是因为他们早就计划好的，而是在浏览过程中被精准匹配的商品所激发。这种基于深度学习的个性化发现，是单纯的搜索框式 AI 无法轻易替代的。
Original English **Adam Foroughi**: totally complete discovery. And this is what makes Meta so amazing in their ad business and what we aspire to do. You create that discovery moment. Now, most consumer purchasing today doesn't happen because they had premeditated intent. It happens because they discovered something compelling while they were engaged in another digital experience.
--- ### 推荐机制与隐私迷思 **主持人**: 有时候这种精准度甚至让人觉得毛骨悚然。
Original English **Host**: That's creepy.
**Adam Foroughi**: 是的，当这种巧合发生时，确实会让人产生“被窥探”的错觉。
Original English **Adam Foroughi**: Yeah, it feels creepy when it does happen
**主持人**: 深入探讨一下这个边界：当很多人说“我在午餐时随口聊到了某个东西，结果几分钟后手机上就弹出了那个东西的广告”，这背后到底发生了什么？手机真的在窃听我们吗？
Original English **Host**: or just to push the boundaries at him. Like what what is actually happening when people say I say something at lunch and all of a sudden an ad for that thing is in front of me?
**Adam Foroughi**: 事实并不是窃听。我认为真实情况是你在不经意间做了其他可被追踪的操作：比如你曾经搜索过相关话题、浏览过某个网页、或者查看过类似商品，你自己可能转头就忘了，但推荐算法记住了这些微弱的意图信号，并迅速完成了广告出价与展示。
Original English **Adam Foroughi**: I mean I I think you've done other actions that are trackable like do a search, browse a website, do a product search and you don't realize it and then the machine learning picks it up and serves you that ad.
**主持人**: 但还有一种流传很广的理论：比如我们几个人坐在一起吃午餐，应用利用地理位置数据把我们归入同一个临时群组。如果 Chamath 搜索了某个东西，算法就会认为既然我们在一起，我可能也会对它感兴趣，于是把广告推给我。是这样运作的吗？
Original English **Host**: There's a theory though that if we were all at lunch, especially with these apps, you know, our geol location, you've kind of put us into a group. So, if Chimoth searches for something, the ad shows up on my phone because we were together.
**主持人**: 真实情况是这样吗？很多人都是这么对我解释的。
Original English **Host**: is that what's happening? That's what I'm told.
**Adam Foroughi**: 我认为绝大多数独立的广告技术公司根本无法追踪精确的地理位置。至少在 **AppLovin**，我们完全不追踪用户的地理位置。追踪每个人的精确位置是一个极其沉重且敏感的技术概念，无论在隐私合规还是系统开销上都不可行。
Original English **Adam Foroughi**: I don't think advertising companies can track location. So, we don't track location at all. It's a really heavy concept to track people's precise location, and it's not something we do.
**主持人**: 那如果是基于好友关系或社交图谱中的群组关联呢？
Original English **Host**: But what about us being friends and being connected together like groups?
**Adam Foroughi**: 作为第三方广告网络，我们没有那样的社交图谱数据。但如果你使用的是像 Meta 这样的社交网络平台，你的好友关系链和互动网络确实有可能影响广告体验。如果 Chamath 频繁搜索并购买了某样东西，而你们在社交图谱上有高度重合的兴趣标签和互动，社交平台的算法完全有可能将这类商品推荐给类似受众。但在游戏和应用内广告领域，这完全是纯粹的上下文与模型预测。
Original English **Adam Foroughi**: So, we wouldn't have that data. But if you're on a social network, of course, your relationships together might drive an ad experience. If Chimoth searches for something and you guys are connected in their graph, their graph intelligence might surface that. But for us in gaming and across open apps, we don't have that data, and we don't rely on it.
--- ### 团队架构与全球研发 **主持人**: Adam，让我们回到公司的运作模式上。我觉得这家公司最吸引人的地方之一在于你们独特的经营方式。你常驻在洛杉矶对吧？
Original English **Host**: Adam, let's just go back to the um because what I find so fascinating about the business is the the the way you've operated it. You're you're based in LA, right?
**Adam Foroughi**: 是的，我常驻洛杉矶。公司最初在硅谷创立，我们的总部在帕洛阿尔托（Palo Alto）。
Original English **Adam Foroughi**: I'm based in LA. Company started in Silicon Valley. We're in Palo Alto.
**主持人**: 帕洛阿尔托。但你人在洛杉矶，而且你们在中国有大量的研发人员，是这样吗？
Original English **Host**: Palo Alto. But you're here and then you have a lot of developers in China. Is that right?
**Adam Foroughi**: 我们的工程研发办公室主要分布在帕洛阿尔托、北京以及新加坡。
Original English **Adam Foroughi**: We have our engineering offices are in Palo Alto, um, Beijing and Singapore.
--- ### 股价暴跌92%与极限逆袭 **主持人**: 你们早期并没有募集大量的风险投资。你在 2021 年带领公司上市，一挂牌市值就达到了近 300 亿美元。
Original English **Host**: And then the company didn't raise a lot of venture money. You take the company public 2021. It went public like 20 billion market cap out the gate.
**Adam Foroughi**: 是的，我们在 2021 年 4 月上市，当时的 IPO 市值大约是 **280 亿美元**。
Original English **Adam Foroughi**: Yeah, we were co IPO. We went out in April 2021. Uh, it was about $28 billion.
**主持人**: 280 亿美元。然后到了 2023 年，市值最低暴跌到了多少？
Original English **Host**: 28 billion. And then in 2023, what did the market cap collapse to?
**Adam Foroughi**: 这就是公开市场极其戏剧化的地方。我们在 2021 年以上市价每股 80 美元挂牌，市值最高曾冲到接近 400 亿美元。随后整个科技股遭遇宏观退潮，加上苹果隐私政策（ATT）的冲击，在 2022 年底到 2023 年初，我们的股价一路狂跌到每股 9 美元左右，总市值只剩下 **30 亿美元** 出头。 换句话说，市值从高点直接蒸发了 **92%**。但滑稽的是，当时我们账面上每年依然能产生四五亿美元的自由现金流。
Original English **Adam Foroughi**: Well, this is the funny thing about the public market. So, we went out in 2021, $80 a share, market cap got up to around 40 billion. Then everything in tech crashed, and by late 2022/early 2023, our stock went down to $9 a share, and the market cap collapsed to like $3 billion. So we lost 92% of our value, even though we were generating $400-500 million in free cash flow at the time.
**主持人**: 这太不可思议了。
Original English **Host**: That's incredible.
**主持人**: 等等，让我们仔细拆解一下这个过程。当时市场由于种种原因对你们的业务充满了极度的怀疑和不信任。在那个至暗时刻，你作为创始人是怎么应对的？
Original English **Host**: So, so hold on. So, so, so, so let's just go through this. So, the market's in disbelief for some reason about the business. And what do you do?
**Adam Foroughi**: 我本身是金融背景出身，受过非常扎实的金融市场训练。你很快就会明白一个道理：公开市场上的股票价格，在短期内完全是由边际买家和边际卖家的情绪决定的，它往往脱离了底层生意的真实基本面。 如果市场陷入恐慌，所有人都在疯狂抛售，你的股价就会被砸到不可思议的低点。但我清楚地看着我们的财务报表：我们的现金流是真实的，客户依然在使用我们的产品，我们的底层模型正在取得重大技术突破。 既然市场以不可思议的 30 亿美元估值把我们当成破产公司来甩卖，而我们账上又拥有充沛的现金流，那最理性的资本配置行为就是：**疯狂回购我们自己的股票**。我们在底部回购了数亿美元的股票，把被市场错误定价的股份重新收回到公司手中。
Original English **Adam Foroughi**: Yeah. What what you learn pretty quickly and I'm a finance background so I had good education on this is that your price in the markets is determined by the marginal buyer and the marginal seller. It doesn't necessarily reflect the intrinsic value of the business. When everyone is dumping tech, they just sell everything indiscriminately. But I looked at our P&L, we were printing cash, our core business was healthy, and our engineers were making massive progress on our next-gen AI model. So when the market offered us our own company at a $3 billion valuation, we decided to aggressively buy back our own stock. We took our cash flow and bought back massive amounts of shares at the bottom.
**主持人**: 在那段时期，你的个人心理状态是怎样的？你一直都保持着这种超然冷静的信念，还是经历过极度沮丧的低谷期？
Original English **Host**: How did you manage Sorry. Did you feel that way from the like the whole time or was there this period of depression where you're like
**主持人**: 这正是我一直想问的。当公司股价跌去整整 **92%** 时，你究竟是如何稳定内部团队军心的？
Original English **Host**: oh my gosh what that's what I was going to ask. How do you manage the internal culture when the stock is off 92%.
**Adam Foroughi**: 那确实是一段极其艰难的日子。我甚至会接到家里亲戚和朋友打来的电话，小心翼翼地问我：“你还好吗？你不会想不开要自杀吧？” 我对他们说：“看在上帝的份上，我们每年还在赚好几亿美元的净现金，我们活得非常好！” 但在公司内部，员工看着自己的期权和股权激励大幅缩水，恐慌是必然的。普通员工不可能拥有创始人那样的股权比例或大局观。
Original English **Adam Foroughi**: It's tough. I mean, like I'll tell you, like I would get phone calls from family members, friends, are you suicidal? And I'm like, look, we we got stuff going on, but we're printing $500M in cash, we're not going out of business! But for employees, when they see their equity drop by 90%, it's terrifying.
**主持人**: 没错。他们既没有你这样的资历和威望，也没有你那么大的股权缓冲。
Original English **Host**: Exly. And they don't have the the gravitas that you do, nor the ownership.
**Adam Foroughi**: 所以我们给全公司树立了一种**“我们对抗全世界”（Us against the world）**的心态。 我对团队说：“听着，现在全世界都看衰我们，华尔街把我们当成弃子。但这恰恰是我们的机会。我们要用事实打所有人的脸。我们将全力回购股票，我们将专注把新一代 AI 引擎打造出来。只要核心指标恢复增长，市场终究会被迫承认现实。” 我们将这种外部压力转化为了极其强大的团队凝聚力。
Original English **Adam Foroughi**: So, so we built it by just saying, "Look, it's an us against the world mentality." Like, everyone's turned against us. we're going to buy back shares aggressively, we're going to ship this new technology, and we're going to prove every single one of them wrong. And that focus united the entire team.
--- ### AXON 2.0架构跃迁 **主持人**: 后来转折点是如何发生的？投资者是什么时候开始重新注意到你们并大举入场的？
Original English **Host**: and then what investors started showing up and saying hey we're paying attention again it
**Adam Foroughi**: 这是一段非常精彩的技术蜕变。正如我们几分钟前聊到的，广告技术正在经历从 **ML 1.0 到 ML 2.0** 的代际跨越。 在过去，整个行业的广告推荐本质上依赖的是**传统回归模型（Regression Models）**——根据历史特征计算点击率和转化率的加权概率。但在过去几年里，我们的研发团队彻底推倒重来，构建了基于**深度学习神经网络（Deep Learning Neural Networks）**的新一代推荐引擎——**AXON 2.0**。 当我们将 AXON 2.0 部署上线后，模型展现出了不可思议的预测与扩展能力。广告主的投资回报率（ROAS）开始飙升，广告主发现只要在 AppLovin 上加大投放，就能获得极其可预测且高额的用户转化。 我们的营收开始以每季度令人难以置信的速度爆发式增长，华尔街终于如梦初醒。
Original English **Adam Foroughi**: it was interesting because for us what happened was we went from ML 1.0 like we talked about a couple minutes ago to ML2.0 We we went from a regression-based model, which was the old way ad networks worked, to a full deep learning neural network engine with AXON 2.0. And when we rolled that out, the performance exploded. Advertisers started seeing massive ROAS (return on ad spend), and they began pouring billions into our platform because our engine was finding high-value users better than almost anyone else in the mobile ecosystem.
**主持人**: 也就是从你亲自走出去向市场证明：“嘿，我们公司不仅活得好好的，而且我们彻底走出了泥潭。”
Original English **Host**: From me just going out and saying, "Hey, our company still exists. We survived this."
**Adam Foroughi**: 没错。当你坐在投资者会议室里，你能非常轻易地看懂桌子对面那些人的微表情。起初所有人都充满怀疑，但当你把连续几个季度的真实财报、超预期的现金流、以及 AXON 2.0 的算法提升曲线摆在他们面前时，整个房间的气氛瞬间反转了。
Original English **Adam Foroughi**: Yeah. And then people like I'd sit in the meetings and do like it's pretty easy to read the other side of the room if you you do that kind of thing. Once the numbers started printing and the growth re-accelerated, skepticism turned into euphoria.
**主持人**: 那硬币的反面又是什么呢？当华尔街的基金经理们全部转为空头翻多头并重仓你们之后，他们现在的态度是不是变成了追着问你：“好的 Adam，我们接下来怎么进一步扩张？怎么增长得更快？电商广告能做多大？”
Original English **Host**: And then what's what's the opposite side of that? Once they're long, are they now asking you, "Okay, Adam, how do we expand? How do we grow faster? What's next?"
**Adam Foroughi**: 遗憾的是，公开市场上真正具有**逆向思维（Contrarians）**的人少之又少。大多数人永远在羊群效应中随波逐流：在低谷时他们可以把悲观推向极端，而在上涨时他们又会把狂热推向极端。 当股价从 9 美元一路飙升到 300 多美元、市值突破 **1000 亿美元** 时，曾经那些质疑你随时会破产的人，现在每天都打电话催促你赶紧横扫所有新市场。但作为创始人，你必须始终保持清醒，不被极端的悲观击垮，也不被狂热的泡沫冲昏头脑。
Original English **Adam Foroughi**: So, unfortunately, not a lot of people are contrarians. So, you can go to the extreme down and then on the other side of it, you can go extreme up, too. When you're at the bottom, they think you're going to zero; when you're at the top, they think you're going to take over the universe. As a founder, you have to tune out both extremes and just execute on the technology and the product.
--- ### 苹果隐私新规与电商突围 **主持人**: 也许你可以谈谈隐私政策的影响。苹果的 **ATT 政策** 和欧盟的监管机构一直在严厉审视广告追踪生态。很多传统广告公司被苹果的隐私新规打得一蹶不振。你们当初甚至收购了大量移动游戏工作室来作为数据防线，你们后来是怎么应对这一变革的？
Original English **Host**: Maybe you could talk a little bit about privacy. Apple, the EU really are look at companies like yours um and they this is a little too aggressive in terms of tracking. You guys bought a bunch of game studios originally as a data hedge against ATT. How did that evolve?
**Adam Foroughi**: 没错，我们当初之所以收购那些移动游戏资产，最初确实是一次**数据战略布局（Data Play）**。当我们最初研发深度学习模型时，我们需要拥有第一手的第一方数据来训练算法。 但随着我们的深度学习模型（AXON）不断进化成熟，模型本身对跨应用个人隐私追踪的依赖度大幅降低。算法学会了从高维的上下文信号、微观互动特征以及纯粹的数学概率中去进行匹配，而不再需要依赖传统的设备级标识符（如 IDFA）。 因此，当我们确认模型已经不再依赖这些游戏工作室提供第一方数据时，我们就果断将那些非核心的游戏工作室全部剥离出售了，重新变回了一家纯粹、精简的高毛利纯科技平台。
Original English **Adam Foroughi**: Yeah, we we sold all those games. We bought them originally as a data play. We when we built our first deep learning model, we needed to have data to train the initial architectures. But as AXON evolved, we realized that deep learning models could find patterns and optimize matching purely based on contextual and transactional feedback loops, without needing invasive individual tracking. Once the model was strong enough and didn't need the first-party game data, we sold the gaming assets to focus 100% on being a high-margin, pure-play technology platform.
**主持人**: 广告主最终在意的始终是效果与转化。
Original English **Host**: They want to compare probably
**Adam Foroughi**: 毫无疑问。广告主唯一关心的就是投资回报率。如果你能帮一个电商独立站或者品牌商家以 1 美元的成本带来 2 美元甚至 3 美元的确定性销售额，他们就会源源不断地把预算投给你。 这也是为什么我们现在能够顺利将业务从纯游戏广告拓展到更广阔的**电商（E-commerce）**领域。电商商家在我们的平台上看到了媲美甚至超越 Meta 广告的转化效率。
Original English **Adam Foroughi**: totally. They want to track it. And if you told them after the fact, hey, an agent could have done this for you and saved you 20%. I don't think that matters as much as raw return on ad spend. E-commerce merchants care about making profitable sales. When we plugged our recommendation engine into e-commerce, the conversion rates were so high that advertisers naturally allocated more and more budget to us.
--- ### 极致工程与85%惊人利润率 **主持人**: 我想弄清楚你们到底凭什么能赢。世界上两家最聪明、拥有最顶级工程师的公司——**Meta** 和 **Alphabet (Google)**，在广告领域拥有几乎垄断性的资源。作为一家规模相对较小的独立公司，你们究竟是如何在他们眼皮底下击败巨头并建立起如此深厚的护城河的？
Original English **Host**: yeah let me just try and understand how you won because um the two of the smartest companies in the world with the best engineers Meta Alphabet Google dominate this space. How did AppLovin carve out this territory and win against the giants?
**Adam Foroughi**: 我想说的是，我们从不认为自己已经赢了。我们每天醒来时的心态都是：我们今天随时可能被竞争对手打垮，我们必须时刻保持敬畏与饥饿感。 但如果说有什么因素帮助我们走到了今天，那就是**极度的专注与极致的工程文化**。巨头们战线太长，有成千上万的项目在分散注意力。而我们在过去十年里，几乎把所有的精力和算力都押注在移动生态与深度学习推荐引擎这一件事情上。
Original English **Adam Foroughi**: Yeah, I mean, here's here's something that that helped us get to this point. We never think we won. We think every day we wake up and we're probably going to get disrupted if we don't execute. But what helped us is extreme focus. The big tech companies have a million different priorities. We had one single focus: building the absolute best algorithmic auction and recommendation engine for mobile developers and advertisers.
**主持人**: 那么你们业务中的摩擦损耗（Leakage）到底在哪里？如果看你们的损益表（P&L），一般的企业总会有各种各样的渠道成本、履约成本、支付通道摩擦等等。
Original English **Host**: what's the leakage in the business then? So meaning when you look at a P&L, you know, we did this I did this thing with Amazon a decade ago where it's like all this friction. Where is the leakage for you guys?
**Adam Foroughi**: 确实。
Original English **Adam Foroughi**: Yeah.
**主持人**: 你们的损耗形式是什么？是支付基础设施？还是算力与带宽？因为你们的财务数据表现得太不可思议了。
Original English **Host**: Um what's that version for you like there there must be is it payment infrastructure is it uh other kinds of things Jason asked you about apps but I guess what is the structural cost?
**Adam Foroughi**: 我们的理念是赋能广告主与消费者实现无缝对接。在整个公司运营上，我们保持了**极其精简（Extremely Lean）**的组织架构。我们高度依赖算法驱动和全自动化流程，全公司只有几百名核心员工，几乎没有任何冗余的中间环节和人员摩擦，因此我们的成本结构中几乎没有传统软件公司的严重损耗点。
Original English **Adam Foroughi**: We want to power the advertisers to meet the consumer and we've run extremely lean and been so algorithmically focused and automation focused that we haven't had a lot of points of leakage. We keep our team small, our infrastructure highly optimized, and automate everything that can be automated.
--- ### 护城河与跨国研发飞轮 **主持人**: 另一面是，当你们交出高达 **85% 的调整后 EBITDA 利润率**时，市场上的经典反应通常是：“天哪，他们肯定在‘超额赚取超额利润’（Over-earning）。”随后竞争对手会跳出来说：“我可以把 Adam 的高利润率打掉，我愿意只赚 50% 甚至 60% 的利润率来跟你们打价格战。”但这种竞争在现实中并没有发生，你们的超高利润率维持得异乎寻常的持久。
Original English **Host**: The other side then is when you have 85% EBIT to margins people say wow they could be over earning. Right? That's the classic phrase. And then you have competition that says I can compete Adam's margins away. I'm willing to do this at 60% or 50%. But sort of maybe as a correlary to David's question, that hasn't really happened and it's been incredibly sustained.
**Adam Foroughi**: 确实如此。
Original English **Adam Foroughi**: Yeah.
**主持人**: 你认为这是为什么？为什么别人无法轻易通过降价竞争来侵蚀你们的利润？
Original English **Host**: And why do you think that is?
**Adam Foroughi**: 因为这类**深度学习推荐系统的技术复杂度极高**。如果你能在算法架构上持续创新，并且拥有海量、独特的实时交易数据，你就能建立起极其深厚的飞轮优势。 这就像大语言模型领域一样，理论上别人都可以宣称要用更低成本的模型颠覆 Anthropic 或 OpenAI，但当一个强大的模型达到了规模效应并被庞大的生态群体采纳时，数据闭环与网络效应构成的壁垒，是外人极难通过单纯降价就能跨越的。低效率的算法即便收费更低，给广告主带来的总产出依然远逊于高效率模型。
Original English **Adam Foroughi**: Because these technologies are really complex and if you can innovate and you have differentiated data, you can build an advantage. I mean by that token anthropic shouldn't be running away with the large language model space but the power of a model that then reaches a point of scale and gets adopted by a large scale community becomes something that is a a moat that is hard for other people to overcome. If a competing platform has an inferior model, even if they charge zero take rate, the advertiser still loses money because the ads don't convert.
**主持人**: 谈谈你们在**中国**的研发团队吧，这支团队为你们带来了多大的竞争优势？
Original English **Host**: and uh talk to us about this the team in China and how big of an edge these folks are.
**Adam Foroughi**: 中国的工程师群体极其谦逊、极其极其勤奋，而且在数学和工程实现上极其敏锐聪明。无论他们是在中国本土、美国还是世界任何地方工作，只要你能与他们紧密协作，你就是在与全球最顶尖的大脑共事。 当我创立这家公司时，我最核心的目标之一就是：与世界上最优秀的一群人共事，一起攻克最硬核的技术难题。 说实话，每当我走进会议室与我们技术团队坐在一起时，我深知自己大概率是那个房间里最不聪明的人。而这恰恰是每天最让我兴奋、最有动力去上班的原因。
Original English **Adam Foroughi**: I mean, Chinese people are very humble. They're very, very hardworking. They're very sharp. And if you can work with them, whether out of China or United States or any other part of the world, you're working with some of the brightest minds in the world. And so, uh, when I started the business, one of my goals at this company was just work with great people and figure things out. And so, I'm when I sit in a room with some of the people on my team, I know I'm probably the dumbest person in that room. And that gets me excited to show up.
**主持人**: 这句话太有力量了。能意识到这点并以此为动力。
Original English **Host**: Got him. gets me to show up for the
**主持人**: 听到这话感觉真棒。
Original English **Host**: feel when you hear that.
**Adam Foroughi**: 是的。
Original English **Adam Foroughi**: Yeah,
**主持人**: 非常赞同。
Original English **Host**: works for me.
**主持人**: 好的，让我们把掌声送给 Adam！
Original English **Host**: All right, let's give it up for Adam.
**主持人**: Adam，非常感谢你！多谢了，兄弟。
Original English **Host**: Adam, thank you. Thanks, bro.
**Adam Foroughi**: 这次交流太棒了。谢谢各位，非常高兴见到你们！
Original English **Adam Foroughi**: That was great. Thanks, man. Thanks, man. Great to see you.