XR收入涨28%，下一代AR+AI眼镜在研 | Meta Q2 电话会议纪要全文（中/英文）| 福利
科技
科技
2024-08-02 10:30
美国
![]()
Meta公司于7月31日发布2024年第二季度财报,披露其Reality Labs部门(负责AR/VR业务)收入同比增长28%。CEO扎克伯格首次披露正与依视路陆逊梯卡集团合作开发新一代AR+AI眼镜。AR圈将在周日（8月4日）面向XR研究院（知识星球）会员直播，直播主题为《AI终端茶话会》，欢迎扫描下方二维预约观看。主题：在这个轻松愉快的线上聚会中，我们将共同探讨和分享人工智能在手机、PC、XR、智能眼镜、智能汽车等领域的最新发展、未来趋势和投资机会。欢迎大家在文末“发消息”提问、说明感兴趣的方向，或者申请分享看法。- 黑毛警长008：自媒体“AR圈”创始人（5W+粉丝）、“XR研究院”主理人，前H公司战规专家，XR产业链与电子行业资深专家。
- 粲然一笑2021：雪球、公众号同名自媒体（5W+粉丝）和“笑研科技”知识星球主理人和二级私募基金公司创始人。
- 刘延：知乎大V（ID：刘延，5W+粉丝），数码榜前三，某汽车相关智库团队成员，对端侧AI、电子创新和智能汽车有深刻见解！
- 杨臻：深耕XR、AI、内容一级投资，体验派深入洞察科技前沿，现任某动画大厂MR&AI战略负责人。
方式：线上直播（录音会后在知识星球发布，仅会员可见）对象：为了更好互动闭门交流，仅限“XR研究院”与“笑研科技”知识星球成员参加。限8月3日16:00前入会的成员。福利：扫下方二维码，享受98元“XR研究院”入会福利![]()
我们的主要长期重点仍然是元宇宙工作。上个季度我讨论了AI的进步如何加快了一些产品的时间表。几年前，我会预测AR眼镜可能会在AI之前实现，但现在看来这些技术实际上会以相反的顺序准备好。我们因为已经进行的Reality Labs投资而处于有利位置。Ray-Ban Meta眼镜继续比我们预期的更受欢迎——部分原因是AI。需求仍然超过我们的生产能力，但我希望我们能很快满足这种需求。EssilorLuxottica一直是一个很好的合作伙伴，我们很高兴与他们合作，继续构建未来几代的AI眼镜，继续建立我们的长期合作关系。Quest 3的销售也超过了我们的预期——我认为这是因为它不仅是价格最优的MR头显，而且是市场上最好的头显。除了游戏，人们越来越多地利用Quest的能力作为一个通用计算平台，花时间看视频，浏览网站，通过虚拟桌面扩展他们的PC等等。Horizon在VR、移动和桌面平台上的增长也在继续——我预计它将成为这一生态系统中越来越重要的一部分。我们将在9月25日举办我们的年度Connect会议，我们将有很多关于我们所有AI和元宇宙工作的令人兴奋的更新——所以我鼓励大家关注这个会议。此外，我们即将推出下一代Ray-Ban Meta眼镜，与EssilorLuxottica的合作关系将进一步深化。我们将继续在未来几代产品中集成更多的AI功能，以提升用户体验。这些新的功能将包括改进的智能助手、增强的现实显示技术以及更强的计算能力。这将使我们的眼镜不仅仅是一款智能设备，更成为日常生活中不可或缺的工具。在Reality Labs板块，第二季度收入为3.53亿美元，增长了28%，主要由Quest头显销售推动。Reality Labs的支出为48亿美元，同比增长21%，主要是由于与人员相关的费用和Reality Labs库存成本增加。Reality Labs的营业亏损为45亿美元。我们预计2024年Reality Labs的运营亏损将同比大幅增加，原因是我们持续的产品开发工作和进一步扩大我们生态系统的投资。虽然我们无意在第四季度电话会议之前提供2025年的任何定量指导，但我们预计基础设施成本将是明年费用增长的重要驱动因素，因为我们将在扩展的基础设施足迹中确认折旧和运营成本。Reality Labs仍然是我们继续进行大量投资的另一个长期项目。Quest 3销售良好，Ray-Ban Meta智能眼镜显示出非常有希望的吸引力，需求、使用和保留率的早期信号增加了我们对AR眼镜长期潜力的信心。Meta Platforms, Inc.（META）2024年第二季度业绩电话会议 2024年7月31日谢谢大家。下午好，欢迎参加Meta Platforms 2024年第二季度财报电话会议。今天与我一同讨论我们业绩的有CEO马克·扎克伯格和CFO苏珊·李。在我们开始之前，我想提醒大家，我们今天的发言将包括前瞻性声明。实际结果可能与这些前瞻性声明中提到的结果有实质性不同。可能导致这些结果有实质性差异的因素在今天的财报新闻稿和我们向SEC提交的10-Q季度报告中有详细说明。我们今天在电话会议上作出的任何前瞻性声明都是基于今天的假设，我们不承担因新信息或未来事件而更新这些声明的义务。在本次电话会议中，我们将介绍GAAP和某些非GAAP财务指标。GAAP与非GAAP指标的对账已包含在今天的财报新闻稿中。财报新闻稿和相关投资者介绍材料可在我们的网站investor.fb.com上查看。现在，我将电话交给马克。谢谢你，肯。大家好，谢谢今天的参与。对于我们的社区和业务来说，这是一个强劲的季度。我们估计现在每天有超过32亿人在使用我们的至少一个应用程序。我们在美国看到的增长尤其令人振奋。WhatsApp现在在美国有超过1亿月活跃用户，我们看到Facebook、Instagram和Threads在美国和全球范围内的年同比增长也很不错。我对我们在Facebook上吸引年轻成年人的进展尤其满意。我们看到的数字，尤其是在美国，确实与公众关于谁在使用这个应用的说法相矛盾。几年前我们开始更关注18-29岁的用户，现在很高兴看到这些努力带来了良好的结果。另一个亮点是Threads，它即将达到2亿月活跃用户。我们正在稳步推进，将其建设成另一个主要的社交应用。我们看到更深入的参与度，我对这里的轨迹感到非常满意。目前的主要主题当然是AI，我今天的评论将集中在三个方面：AI对我们的应用家族和核心业务的意义，我们看到的新AI体验和机会，以及AI如何影响我们的元宇宙工作。首先，在Facebook和Instagram上，AI的进步继续提高推荐质量并推动参与度。我们发现，随着我们开发更多的一般推荐模型，内容推荐变得更好。这个季度我们在Facebook上推出了全屏视频播放器和统一的视频推荐服务，将Reels、长视频和Live整合到一个体验中。这使我们能够扩展我们的统一AI系统，这已经比我们最初从CPU转向GPU时增加了更多的参与度。随着时间的推移，我希望我们能够朝着一个单一的、统一的推荐系统迈进，该系统能够驱动所有内容，包括“你可能认识的人”等推荐。我们还没有达到这一点，所以仍有上升空间，但我们在这里取得了良好的进展。AI也将在一些令人兴奋的方式中显著进化我们的广告服务。过去，广告客户会带着他们想要接触的特定受众来到我们这里，比如某个年龄段、地理位置或兴趣。最终，我们的广告系统能够比广告客户自己更好地预测谁会感兴趣。但今天，广告客户仍然需要自己制作创意。在未来几年，AI将能够为广告客户生成创意——并且还能够在用户看到它时对其进行个性化。长期来看，广告客户基本上只需要告诉我们一个商业目标和预算，我们将为他们完成其余的工作。我们会逐步实现这一目标，但我认为这将是一个非常重要的事情。接下来，谈谈AI带来的全新体验，上个季度我们开始广泛推出我们的助手Meta AI，并且它有望在今年年底前成为最常用的AI助手。我们有一个令人兴奋的路线图，我们想要添加的功能，但这里的核心是Meta AI正朝着成为一个重要服务的目标快速发展，其智能和功能都在迅速提高。一些用例是实用性的，比如搜索信息或在与他人进行困难对话前进行角色扮演，其他用途则更具创意，比如新的“想象你自己”功能，可以让你以任何你想要的风格创建自己的图像。AI的美妙之处在于它的通用性，所以我们还在发掘它的广泛用例。我们愿景的重要部分是，我们不仅在创建一个AI，还在让许多人创建他们自己的AI。本周我们推出了AI Studio，任何人都可以在我们的应用中创建与他们互动的AI。我认为创作者们尤其会觉得这非常有价值。在我们的应用中有数百万的创作者——这些人想要更多地与他们的社区互动，而他们的社区也想要更多地与他们互动——但一天中只有那么多小时。现在他们可以使用AI Studio创建AI代理，代表他们与社区聊天，回答问题，创建内容等等。所以我对此感到非常兴奋。但这不仅仅是针对创作者的。任何人都将能够基于他们的兴趣或不同的话题创建自己的AI，并与之互动或与朋友分享。企业AI是另一个重要部分。我们仍在与越来越多的企业进行alpha测试。目前收到的反馈是积极的。随着时间的推移，我认为就像每个企业都有一个网站、社交媒体账号和一个电子邮件地址一样，未来每个企业也会有一个客户可以与之互动的AI代理。我们的目标是让每个小企业，最终每个企业，都能够轻松地将其所有内容和目录整合到一个AI代理中，从而推动销售并为他们节省成本。当这一切规模化运作时，我认为这将大大加速我们的业务消息收入。还有很多其他新的机会让我感到兴奋，但我会在我们准备推出时再谈。驱动所有这些新体验的引擎是Llama系列基础模型。这个季度我们发布了Llama 3.1，它包括第一个前沿级开源模型，以及新的行业领先的小型和中型模型。405B模型在成本性能方面优于领先的封闭模型，并且由于它是开源的，立即成为调整和提炼你所需大小的自定义模型的最佳选择。我认为我们会回顾Llama 3.1，把它看作是开源AI成为行业标准的转折点，就像Linux一样。我经常被问到为什么我对开源如此乐观。我与Llama 3.1发布时写了一封信，解释了为什么我认为开源对开发者、对Meta以及对整个世界都更好。我的观点是，开源将更安全，将更快地推动改善我们所有人生活的创新，并且还将创造更多的共享繁荣。对于Meta的自身利益来说，我们在构建最佳消费者和广告体验的业务中。为了做到这一点，我们需要访问领先的技术基础设施，而不受竞争对手允许我们做什么的限制。但这些模型是生态系统，它们不仅仅是我们可以自己开发的独立软件。所以如果我们想要在我们的模型周围开发最强大的工具、效率改进、硅优化和其他集成生态系统，我们需要它们被行业内的开发者广泛使用。一旦我们知道我们将能够访问领先的模型，那么我相信我们将能够构建最好的社交和广告体验。我对这一点如此乐观的部分原因是我们在开源方面有长期的成功记录。通过开放计算项目，我们通过让供应链标准化我们的基础设施设计，节省了数十亿美元。像PyTorch和React这样的开源工具带来了行业贡献的实际好处。这种方法一直对我们有效，我也期待它在这里同样有效。另一个主要关注领域是确定支持更多和更先进模型训练的基础设施容量的正确水平。Llama 3已经与最先进的模型竞争，我们已经开始着手Llama 4的工作，我们的目标是明年成为行业最先进的模型。我们正在规划未来几年的计算集群和数据。训练Llama 4所需的计算量可能是训练Llama 3所需的近10倍——未来的模型将继续增长。很难预测这一趋势在未来几代中的发展方向，但在这一点上，我宁愿冒着提前建设容量的风险，而不是太晚，因为启动新基础设施项目的准备时间很长。随着我们扩大这些投资，我们当然会继续致力于公司各个方面的运营效率。我想讨论的最后一个领域是AI如何影响我们的元宇宙工作[MH1] ，这仍然是我们的另一个长期重点。上个季度我讨论了AI的进步如何加快了一些产品的时间表。几年前，我会预测全息AR可能会在智能AI之前实现，但现在看来这些技术实际上会以相反的顺序准备好。我们因为已经进行的Reality Labs投资而处于有利位置。Ray-Ban Meta眼镜继续比我们预期的更受欢迎——部分原因是AI。需求仍然超过我们的生产能力，但我希望我们能很快满足这种需求。EssilorLuxottica一直是一个很好的合作伙伴，我们很高兴与他们合作，继续构建未来几代的AI眼镜，继续建立我们的长期合作关系。Quest 3的销售也超过了我们的预期——我认为这是因为它不仅是价格最优的MR头显，而且是市场上最好的头显。除了游戏，人们越来越多地利用Quest的能力作为一个通用计算平台，花时间看视频，浏览网站，通过虚拟桌面扩展他们的PC等等。Horizon在VR、移动和桌面平台上的增长也在继续——我预计它将成为这一生态系统中越来越重要的一部分。我们将在9月25日举办我们的年度Connect会议，我们将有很多关于我们所有AI和元宇宙工作的令人兴奋的更新——所以我鼓励大家关注这个会议。最后，我们在核心产品和业务方面的强劲表现使我们有机会对未来进行深度投资——我计划充分抓住这一机会，构建一些令人惊叹的事物，这些事物将在未来几十年为我们的社区和投资者带来回报。我们在基础技术和产品体验方面的进展表明我们走在正确的道路上。我为我们的团队迄今取得的成就感到自豪，我对我们在未来机会上的执行能力感到乐观。像往常一样，感谢我们的团队推动所有这些重要工作，感谢你们与我们一起踏上这段旅程。现在请苏珊发言。谢谢马克，下午好，大家好。让我们从综合结果开始。除非另有说明，所有比较均为同比。第二季度总收入为391亿美元，同比增长22%，按不变汇率计算增长23%。第二季度总支出为242亿美元，同比增长7%。收入成本增加了23%，主要由于基础设施和Reality Labs库存成本的增加。研发费用增加了13%，主要是由于与人员相关的费用和基础设施成本增加，部分被重组成本的减少所抵消。市场营销与销售费用减少了14%，主要是由于重组和与人员相关的成本减少。一般管理费用减少了12%，主要是由于法律相关费用减少。我们在第一季度结束时拥有近70,800名员工，比第一季度增加了2%。第二季度营业收入为148亿美元，营业利润率为38%。本季度税率为11%。净收入为135亿美元，每股收益为5.16美元。资本支出，包括融资租赁的本金支付，为85亿美元，主要用于服务器、数据中心和网络基础设施的投资。自由现金流为109亿美元。我们回购了63亿美元的A类普通股，并向股东支付了13亿美元的股息，本季度末拥有581亿美元的现金和有价证券，以及184亿美元的债务。现在进入我们的业务板块结果。我将从我们的应用家族板块开始。我们的应用家族社区继续增长，6月份有大约32.7亿人每天使用我们的至少一个应用。第二季度应用家族总收入为387亿美元，同比增长22%。第二季度应用家族广告收入为383亿美元，同比增长22%，按不变汇率计算增长23%。在广告收入中，在线商务领域是同比增长的最大贡献者，其次是游戏和娱乐及媒体。在用户地理位置方面，广告收入增长最强劲的是其他地区和欧洲，分别增长33%和26%。亚太地区增长了20%，北美增长了17%。在广告客户地理位置方面，总收入增长最强劲的是亚太地区，增长了28%，尽管增长低于第一季度41%的速度，因为我们经历了来自中国广告客户的强劲需求期。第二季度，我们服务中展示的广告总量和每个广告的平均价格均增加了10%。展示量增长主要由亚太地区和其他地区推动。价格增长则由广告客户需求增加推动，部分原因是广告效果的提高。这被展示量的增长部分抵消，特别是来自低变现区域和展示面的展示量增长。应用家族的其他收入为3.89亿美元，增长了73%，主要由WhatsApp Business Platform的业务消息收入增长推动。我们继续将大部分投资用于我们的应用家族的开发和运营。第二季度应用家族的支出为194亿美元，占我们总体支出的约80%。应用家族的支出增长了4%，主要是由于基础设施和与人员相关的费用增加，部分被重组成本的减少所抵消。应用家族的营业收入为193亿美元，营业利润率为50%。在Reality Labs板块，第二季度收入为3.53亿美元[MH2] ，增长了28%，主要由Quest头显销售推动。Reality Labs的支出为48亿美元，同比增长21%，主要是由于与人员相关的费用和Reality Labs库存成本增加。Reality Labs的营业亏损为45亿美元。现在谈谈业务前景。驱动我们收入表现的主要因素有两个：我们提供令人愉悦的体验给我们的社区的能力，以及我们随着时间的推移将这种参与变现的能力。为了提供令人愉悦的体验，我们继续专注于执行我们的优先事项，包括视频和信息流推荐。在Instagram上，Reels的参与度随着我们对推荐系统的持续改进而继续增长。我们的一部分工作集中在增加推荐中的原创帖子比例，以便人们可以发现Instagram的最佳内容，包括新兴创作者的内容。现在，在美国的推荐中，超过一半来自原创帖子。在Facebook上，我们看到我们在六月全球推出的统一视频播放器和排名系统的早期结果令人鼓舞。这个计划使我们能够将Facebook上的所有视频类型整合到一个观看体验中，我们预计这将为短视频的进一步增长机会打开大门，因为我们将越来越多地将短视频混合到Facebook视频参与的总体基础中。随着我们从去年开始使用新系统支持Facebook Reels推荐，这些新系统已经显示出令人有希望的收益。我们预计将在今年和明年扩大这些新系统以支持更多的展示面。我们在更长期的参与优先事项方面也看到良好的势头，包括生成式AI和Threads。自我们首次推出Meta AI以来，人们已经使用它进行了数十亿次查询。我们在WhatsApp上的保留率和参与度方面看到特别有希望的迹象，这与印度成为我们最大的Meta AI使用市场同时发生。你现在可以在超过20个国家和8种语言中使用Meta AI，在美国，我们推出了Imagine edit新功能，允许人们编辑他们通过Meta AI生成的图像。在生成式AI之外，Threads社区也在继续增长并加深他们的参与度，因为我们发布了新功能并改进了我们的内容推荐系统。现在谈谈我们收入表现的第二个驱动因素：提高变现效率。这个工作有两个部分。首先是优化有机参与中的广告水平。我们继续看到在低变现展示面如视频上增加广告供应的机会，包括在Facebook内随着总体视频参与向短视频的转移，随着时间的推移创造更多的广告插入机会。更广泛地说，我们在确定最佳广告展示和展示时间方面变得越来越好，这使我们能够在不增加广告数量甚至在某些情况下减少广告负载的情况下推动收入增长和转化。提高变现效率的第二部分是增强营销表现。我们对这里的进展继续感到满意，AI在其中扮演着越来越核心的角色。我们通过采用AI进步所带来的更复杂的建模技术来改进广告投放，包括我们的Meta Lattice广告排名架构，它在第二季度继续提供广告表现和效率的提升。我们还在使广告客户更容易最大化广告表现并自动化更多的广告活动设置，利用我们的Advantage+套件解决方案。我们看到这些工具继续释放性能增益，今年进行的一项研究表明，在美国广告客户采用Advantage+购物广告活动后，其广告支出回报率提高了22%。广告客户对这些工具的采用在继续扩大，我们正在添加新功能使它们更加有用。例如，这个季度我们向Advantage+购物引入了灵活格式，允许广告客户在单个广告中上传多个图像和视频，我们可以从中选择并自动确定哪种格式最能提供最佳表现。我们还扩大了使用Advantage+购物优化的转化类型列表，增加了10种新的转化类型，包括“添加到购物车”等目标。展望未来，我们相信生成式AI将在企业如何大规模进行营销和与客户互动方面发挥越来越大的作用。我们预计这项技术将继续使企业更容易开发定制和多样化的广告创意。自我们推出首批生成式AI广告功能——图像扩展、背景生成和文本生成以来，我们已经看到了令人鼓舞的早期结果，过去一个月内有超过100万广告客户使用了至少一种这些解决方案。今年5月，我们开始在Advantage+创意中推出完整的图像生成功能，我们已经看到使用该工具的广告客户表现有所提升。最后，我们预计AI将帮助企业通过消息传递更有效地与客户沟通。我们从测试企业使用AI与客户聊天来帮助销售商品和服务以及生成潜在客户开始。虽然我们处于早期阶段，但我们继续扩大测试的广告客户数量，并自开始使用Llama 3以来在回复质量方面取得了良好进展。接下来，我想讨论我们对资本分配的方式，这一方式没有改变。我们继续投资于提升核心体验的短期发展以及我们认为将改变人们未来几年与我们服务互动方式的技术。我们预计拥有足够的计算能力将是这些机会的核心，因此我们正在大量投资于支持我们的核心AI工作、内容排名和广告以及生成式AI和高级研究工作的基础设施。我们在核心AI能力上的持续投资是基于我们看到的强劲回报，并期望在未来继续带来收益，因为我们提升了在我们平台上推荐内容和广告的相关性。虽然我们预计生成式AI的回报将在更长时间内实现，但我们正在将这些投资与我们期望在定制广告创意、业务消息传递、领先的AI助手和有机内容生成方面解锁的重大变现机会相对照。随着我们扩大生成式AI训练能力以推进我们的基础模型，我们将继续以灵活的方式建设我们的基础设施，使我们能够随着时间的推移灵活使用这些能力。这将使我们能够在将训练能力转向生成式AI推理或我们的核心排名和推荐工作时，根据需要进行调整。Reality Labs仍然是我们继续进行大量投资的另一个长期项目[MH3] 。Quest 3销售良好，Ray-Ban Meta智能眼镜显示出非常有希望的吸引力，需求、使用和保留率的早期信号增加了我们对AR眼镜长期潜力的信心。最后，当我们在短期和长期优先事项上进行这些投资时，我们将继续专注于有效地运营业务。现在转向收入前景。我们预计2024年第三季度的总收入将在385亿至410亿美元之间。我们的指导假设外汇是同比总收入增长的2%逆风因素，基于当前的汇率。现在转向费用前景。我们预计2024年全年的总费用将在960亿至990亿美元之间，与之前的预测没有变化。对于Reality Labs，我们继续预计2024年的运营亏损将同比大幅增加，原因是我们持续的产品开发工作和进一步扩大我们生态系统的投资。虽然我们无意在第四季度电话会议之前提供2025年的任何定量指导，但我们预计基础设施成本将是明年费用增长的重要驱动因素，因为我们将在扩展的基础设施足迹中确认折旧和运营成本。现在转向资本支出前景。我们预计2024年全年的资本支出将在370亿至400亿美元之间，比之前的预测范围（350亿至400亿美元）有所更新。虽然我们继续完善明年的计划，但我们目前预计2025年资本支出将显著增长，因为我们投资支持我们的AI研究和产品开发工作。关于税收。除非我们的税收环境发生变化，否则我们预计2024年全年的税率将在中期水平。此外，我们继续关注活跃的监管环境，包括在欧盟和美国日益增加的法律和监管阻力，这可能会显著影响我们的业务和财务结果。总结一下，第二季度是另一个良好的季度。我们继续在我们的业务优先事项上执行得很好，并且在我们面前有令人兴奋的机会，可以为全球使用我们产品的人们和企业带来更多的价值。操作员：谢谢。我们现在将开放提问环节。要提问，请按下你的触键电话上的星号1。要取消提问，请再次按下星号1。在提问前，请拿起你的听筒以确保清晰。如果你今天正在流式播放电话会议，请静音你的计算机扬声器。第一位提问者是摩根士丹利的布莱恩·诺瓦克。请继续。布莱恩·诺瓦克：很高兴你们接受我的提问。我有两个问题，一个给马克，一个给苏珊。马克，我想回到一些为用户和广告客户启用生成式AI的新用例。你谈到了Meta AI、工作室、聊天机器人、扩散模型。如果你能专注于你最兴奋的一两个，我们看到的信号表明它们可能是2025年、2026年业务的真正驱动因素，这将有助于我们了解你最关注的机会。第二个问题，苏珊，你有很多资本支出优先事项，从为下一代模型建设新基础设施到计算能力。请再次向我们解释资本支出哲学以及你为投资资本确保健康回报设置的任何限制。谢谢。马克·扎克伯格：我可以回答第一个问题。我认为在2025年和2026年推动最多结果的事情实际上是我在评论中提到的第一个类别，即AI如何塑造现有产品。AI改进推荐并帮助人们找到更好的内容以及使广告体验更有效的方法。我认为那里有很大的上升空间。这些已经是规模化的产品。我们正在进行的AI工作将改进这些。这将改善体验和业务结果。我们正在进行的其他领域，我认为你们都知道这一点，因为你们关注我们的业务已经有一段时间了，但我们有一个相对较长的业务周期，从启动一个新产品到将其规模化到触及10亿人或更多，并且只有在真正关注大规模变现之后。所以实际上，对于像Meta AI或AI Studio这样的事情，我认为这些是会增加我们产品的参与度并在短期内带来其他好处的事情。但在我们真正谈论这些东西的单独变现之前，我不认为任何人会惊讶我会认为这将是几年的事情。我们已经看到这种情况在Reels和所有这些东西上发生。但我认为对于那些长期关注我们业务的人来说，你也可以很好地预见这些东西在多年之前会成功。我认为那些早期下注的人往往会做得很好，这就是为什么我在评论中分享我们对Meta AI的早期指标。它是早期的。上个季度，我们推出的时间仅仅是一周或两周前的收益电话。这一次，我们已经几个月了。我们可以说的是，我认为我们有望在今年年底前实现成为最常用AI助手的目标。我认为这是一件大事。这是我们想做的唯一事情吗？不，我们显然希望扩大它的参与度，然后随着时间的推移关注变现。但这些早期信号是好的，我认为这是我们目前可以合理预见的所有情况。但我确实认为AI的根本性之一是它最终将影响我们所有产品的某种方式。它将改进现有的，并将使许多新的可能成为现实。所以有很多关于所有技术CEO在这些收益电话中一直谈论AI的笑话。因为这实际上是超级令人兴奋的，并且将在多个时间跨度上改变所有这些不同的东西。苏珊·李：布莱恩，我可以回答第二个问题。在你的投资回报部分问题上，我会大致将我们的AI投资分为两类，核心AI和生成AI。两者在驱动我们业务收入和我们衡量回报的能力方面处于不同阶段。在我们的核心AI工作中，我们继续采用非常基于投资回报的方式进行投资。我们仍然看到强劲的回报，因为参与度和广告表现的改进已经转化为收入增长，并且我们继续在这里投资是有意义的。生成AI是我们刚刚提到的早期阶段。我们不期望我们的生成AI产品在2024年成为收入的重要驱动因素。但我们确实期望它们随着时间的推移开辟新的收入机会，使我们能够在开源Llama的后续版本时获得良好的投资回报。我们已经谈到了我们在生成AI机会上的四个主要领域——增强核心广告业务，帮助我们在业务消息传递中增长，Meta AI的机会，以及随着时间的推移增加核心参与度的机会。另一个我想说的是，我们继续建立我们的AI基础设施，具有灵活性，以便我们可以在我们认为最有用的地方灵活部署能力。我们为生成AI训练建立的基础设施也可以用于生成AI推理。我们也可以通过进行某些修改如添加通用计算和存储，将其用于排名和推荐。我们还采用了分阶段开发数据中心的策略，这使我们能够在需要时更快地满足更多需求，同时限制我们在外年承诺的支出。因此，虽然我们预计2025年资本支出将显著增长，但我们觉得我们有一个良好的框架来思考机会，并确保我们有灵活性，以最有意义的方式部署它。操作员：你的下一个问题来自伯恩斯坦的马克·施穆利克。请继续。马克·施穆利克：是的，谢谢你接受我的问题。正如我们看到的收入指导和前景，苏珊，你能分享一下当前整体数字广告市场的状况吗？你提到了你看到的强劲领域以及你为提高广告产品效果所做的一些特定努力。谢谢。苏珊·李：嗨，马克。我们继续看到健康的全球广告需求，并且我们也在提供持续的广告表现改进，仅与我们一直以来在改善广告、定位、排名、投放等基础设施方面的投资有关。我们预计所有这些将在第三季度继续受益于广告支出。我们确实预计第三季度的同比增长将放缓，因为我们在经历来自中国广告客户的强劲增长以及一年前的Reels展示量强劲增长。同时，我们还预计第三季度的外汇逆风略大于第二季度的水平，基于当前汇率。操作员：你的下一个问题来自高盛的埃里克·谢里丹。请继续。埃里克·谢里丹：非常感谢你们接受我的问题。我只问一个。你提到了在美国建立社区规模以及Threads的发展。你如何看待这些新的、增长较快的Messaging或Threads平台的元素，以及在参与增长和整体变现之间的潜力？你对这些平台的哪些方面最感兴趣，以便在实现规模的同时将其转化为变现机会？谢谢。马克·扎克伯格：我可以先回答，苏珊如果有其他想补充的也可以加入。WhatsApp的统计数据，我认为作为一个业务趋势非常重要，因为美国在我们的收入中占比很大。因此，WhatsApp之前在世界许多国家是领先的消息应用，但在美国不是。我认为现在越来越多的人使用这个产品并意识到，这是一个非常好的体验，它是跨平台通信和群组交流的最佳体验之一，这将意味着我们在未来增长业务机会的所有工作将有一个大的推动力，如果美国成为一个大的市场。我认为这是它的一个重要原因。显然，看到我们周围的人开始使用WhatsApp，这也是个人上的一种满足感，所以我觉得这很有趣，但从业务角度来看可能不太相关。Threads，我认为这是另一个例子，它的启动几乎是我能想到的任何应用程序中最好的。我认为它是最快达到1亿用户的应用程序。这是一个很好的提醒，即使你有这样的开始，从那里到10亿人使用一个应用程序的路径仍然是多年的时间。这是我们的产品周期。我认为这是Meta在构建消费者产品和围绕它们建立业务方面与许多其他公司有所不同的一点，它们推出一些东西并立即开始销售和变现。所以我认为分析我们业务的投资者和相关人士需要理解的是，所有这些新产品，我们发布它们，然后在规模化和将它们扩展为不仅是消费者体验而且是非常大的业务之间有一个多年的时间跨度。但我认为Threads令人兴奋的一点是，我们已经建立这家公司20年了，能有机会再增加一个10亿用户的应用程序并不多。全世界可能有一打左右。我们公司内部可能比外部更多，但我们做得很好，如果我们在这一点上执行得非常好，将其加入我们的产品组合是非常令人兴奋的。现在显然在这之间还有很多工作。我们几乎达到了2亿用户。这是一个很好的里程碑，我对此感到兴奋。还有很多工作要做，使其成为业务的重要组成部分。但我确实认为这些机会非常罕见，这也是我们非常兴奋的事情。我认为团队在这方面做得很好。苏珊·李：埃里克，我想补充一下，就近期的展示增长来源而言，我们确实预计视频将在下半年继续成为展示增长的来源。在Instagram上，我们预计Reels将继续推动增长，而在Facebook上，我们预计将整体视频时间增长，同时增加短视频的比例，这将创造更多的展示增长机会。总体上，我们预计我们的应用社区将在继续增长。操作员：你的下一个问题来自摩根大通的道格·安默斯。请继续。道格拉斯·安默斯：嗨，谢谢接受我的问题。一个给马克，一个给苏珊。马克，在基础设施和资本支出方面，你提到了不仅在建设Llama 3和Llama 4，还可能在向Llama 7前进，并且Llama 4所需的计算量是Llama 3的10倍。既然你在大力建设未来的容量，这如何影响多年的资本支出曲线？苏珊，你能多谈一下第三季度的前景吗？我知道你提到了更艰难的比较，但这实际上表明在高端只有一点外汇中立的下降。所以有没有什么其他具体的因素驱动了预期的强劲？谢谢。苏珊·李：谢谢，道格。我可以谈谈这两个问题。关于长期资本支出的前景。我们还没有分享长期资本支出的轨迹。部分原因是基础设施是我们现在非常动态的计划领域。我们继续研究生成AI路线图的范围如何。显然，我们的预期是明年将在AI基础设施上大幅增加投资，并将在适当的时候提供进一步的指导。但我们在建设所有这些资本支出时，考虑到我之前谈到的因素，即如何灵活构建它，以便我们可以在需要时用于核心AI和生成AI的用例，并确保我们对核心AI投资的回报感到满意，这些我们可以立即测量。你的第二个问题是关于第三季度的收入前景。我之前提到过，我们在平台上看到了健康的全球广告需求。我们在持续提供广告表现改进，我们认为这是许多季度努力的结果，这些努力积累并将继续为我们的平台带来价值。我们在第二季度看到了22%的收入增长，各地区和垂直领域的广泛强劲，特别是在较小广告客户中表现强劲，我们预计这将在第三季度继续。操作员：你的下一个问题来自美国银行的贾斯汀·波斯特。请继续。贾斯汀·波斯特：太好了，谢谢。我想回到你关于美国年轻成年用户增长的评论，特别是在Facebook和Instagram上。我知道你们几年前对Reels进行了重大改变。这些用户在Facebook和Instagram上做什么？你能否给我们一些使用增长的量化数据？苏珊·李：谢谢，贾斯汀。近年在Facebook团队的核心优先事项之一是为年轻成年人构建产品，我们很高兴看到这些努力转化为这一群体的参与度增长。我们在美国和加拿大看到年轻成年用户的应用使用量健康增长。我们看到像Groups和Marketplace这样的产品在年轻成年人中特别受欢迎。在美国和加拿大，发布到Groups的活动在增长，主要由年轻成年人推动。我们还看到他们是Marketplace的活跃用户，Marketplace受益于产品改进和美国对二手产品的强劲需求。操作员：你的下一个问题来自Evercore ISI的马克·马哈尼。请继续。马克·马哈尼：是的，我本来要问Marketplace的问题，这个话题转得很自然。它是一个很大的、可能被低估甚至被严重低估的资产。我知道你们间接变现了它，它是一个非常大的市场。它甚至可能比eBay还大。你对未来可能在变现方面做些什么的看法，部分是为了改善Marketplace的质量。其次，我想问一下人员编制。它同比下降了约1%。你们的员工人数几乎回到了重大减少之前的水平。我们应该如何看待未来的人员编制增长？你提到了资本支出的显著增长，我们应该预期人员编制增长温和吗？显著吗？任何相关的想法都会有所帮助。苏珊·李：谢谢你，马克。关于你第一个问题Marketplace，我们显然很高兴它成为年轻成年人增长的驱动因素之一。我可能会说，整体而言，Marketplace是我们更广泛商业策略的一部分，该策略继续专注于在我们平台上创建最佳购物体验。Marketplace显然是面向消费者的。更广泛的商业策略是使企业更容易在我们平台上广告他们的产品，使买家更容易找到并购买相关项目。为此，我想说，我们对Shops广告的投资也非常满意。Shops广告收入以强劲的同比速度增长，我们看到Shops广告为广告客户带来了额外的表现，并且它与我们其他产品如Advantage+ Shopping的结合也非常有效。你的第二个问题是关于人员编制。我们继续严格管理新员工的分配，以确保它们真正专注于我们的核心公司优先事项，但我们也在减少先前的招聘不足。随着我们在今年余下时间进一步弥补这一招聘不足，我确实预计我们将在2024年结束时报告的员工人数明显高于2023年结束时的水平。我们尚未提供2025年的人员编制增长预期，因为我们还没有开始预算制定过程。但我预计我们主要会将招聘目标集中在优先领域，并将运行一个非常严格的人员编制过程。操作员：你的下一个问题来自Truist Securities的优素福·斯奎利。请继续。优素福·斯奎利：非常感谢。使用Llama 3.1的AI助手在不同版本中被整合，看起来非常令人印象深刻，似乎越来越接近成为一个全面的搜索引擎，除了目前的商业查询。所以你们有没有计划将其开放给更广泛的网络？就像OpenAI在alpha测试中所做的那样，也许将其链接到第三方市场进行商业搜索等。关于Ray-Ban，你能多谈一下与EssilorLuxottica深化合作的机会吗？这会是什么样子？在这个关系中你最感兴趣的是什么领域？谢谢。马克·扎克伯格：是的，我对Llama 3.1的表现非常满意。我认为团队在从第一个版本的Llama到去年的Llama 2取得了很大的进步，Llama 2比前沿领先的模型落后了一代，现在Llama 3.1基本上与其他顶级封闭模型竞争，甚至在某些方面领先。Meta AI使用了一个版本的Llama 3.1以及我们构建的许多其他服务来构建一个连贯的产品。当我之前谈到我们有关于Meta AI的初始使用趋势时，我提到有很多我们想要添加的东西，比如商业功能，你可以逐个垂直领域构建特定功能，使其在所有这些不同领域中有用，最终我认为这就是我们需要做的，使其成为理想的AI助手。所以这是一条长长的路线图。我不认为这些东西会在接下来的几个季度内完成。但这是未来几年我们要做的一部分，我认为这将成为一个非常广泛使用的服务。所以我对此感到非常兴奋。我们还将继续研究Llama。你提到了Llama，我认为问题更多是关于Meta AI，但它们是相关的。Llama是驱动产品的引擎，它是开源的，我对我们在这两个方面的进展感到兴奋。关于智能眼镜，EssilorLuxottica是一个很好的合作伙伴。我们现在已经在第二代Ray-Ban Meta眼镜上合作了[MH4] 。它们表现良好，比我们预期的要好得多，而我们预计它们会比第一代有显著增长，所以这是一个非常积极的惊喜。我认为部分原因是它们与我们看到的AI革命很好地契合，提供了各种新的功能。所以这非常好。EssilorLuxottica是一家伟大的公司，有很多不同的产品，我们希望能与他们合作，继续构建新一代的眼镜，深化AI产品并使其越来越好。我认为这里还有很多进展空间。与我们目前的预期相比，它表现良好。与我们认为它需要成为一个真正领先的消费电子产品相比，我认为我们还处于早期阶段，但所有的迹象都是好的。操作员：你的下一个问题来自花旗的罗恩·乔西。请继续。罗纳德·乔西：太好了，谢谢你们接受我的问题。马克，我想回到关于开源和Llama 3的评论。我完全理解Meta没有提供公共云，所以我想听听你对Llama 3即将推出的产品的产品愿景。也就是说，可能将一些这些产品提供给其他公司，例如用于客户服务或呼叫中心的产品或其他垂直领域。你对如何让开源和Llama 3.1为其他公司提供更大的企业服务有何见解？谢谢。马克·扎克伯格：Llama是基础模型，人们可以将其塑造成各种不同的产品。无论是我们自己的Meta AI，还是AI Studio，还是企业代理，还是Ray-Ban眼镜中的助手，所有这些不同的东西基本上都是用Llama构建的。类似地，任何开发者都将能够利用它构建出更多不同的东西。正如我所说，我认为开源对我们来说如此有价值的原因是我们希望确保我们有领先的基础设施来支持我们构建的消费者和业务体验。但基础设施不仅仅是我们可以独立构建的软件。它确实是一个生态系统，包含许多不同的能力，其他开发者和人们也在加入这个生态系统中，无论是新的工具，用于将模型提炼成你想要的自定义模型的大小，还是更好地微调东西，或者使推理更高效，或者其他我们尚未想到的方法，硅优化，硅公司正在做的所有事情。这是一个生态系统。所以我们不能自己完成所有这些。如果我们构建了Llama并将其保留在我们公司内部，那么它实际上在我们构建所有产品时不会像最终会的那样有价值。所以这是商业策略所在。这就是为什么我们不觉得需要建立一个云并直接销售它，以使其成为一个非常积极的商业策略的一部分。部分工作是与AWS密切合作，我认为他们在这个发布中做得很好。其他公司如Databricks、NVIDIA，当然还有像微软Azure和谷歌云这样的其他大公司，他们都在支持这个。我们希望开发者能在任何地方获取它。我认为这是像Llama这样的开源模型的一个优势——你不会被锁定在提供该模型的某个云中，无论是微软的OpenAI还是谷歌的Gemini或其他什么。你可以将其带到任何地方使用，我们想鼓励这一点。所以我对此感到非常兴奋。我们将最关注的企业和商业应用，除了优化广告客户体验如我之前在评论中所说的，还有企业代理部分。我认为这有巨大的潜力。正如我之前说的，我认为几乎每个企业今天都有一个电子邮件地址。他们有一个网站。他们有社交媒体账号。我认为在未来，他们将至少有一个，如果不是多个企业代理，可以做从互动帮助人们购买东西到帮助支持销售，如果他们对产品有问题，如果他们需要联系你的一切事情。我们已经看到人们在低劳动力成本的国家通过消息与企业互动效果很好。但问题是，要让人回答每个人的问题在许多国家是相当昂贵的。我认为AI非常适合做这件事。当我们能够使数以亿计的使用我们平台的企业轻松将其所有信息、目录、历史和所有分享的内容整合到一个代理中，并真正快速建立一个代理时，我认为这将是非常棒的。所以我们可以将Llama与我们在业务团队中做的很多定制工作结合起来，再加上生态系统其他部分对Llama进行的所有投资，我认为这将是巨大的，但这只是一个领域。这确实跨越了我们构建的所有不同产品，无论是消费者还是商业产品。所以有很多令人兴奋的东西。肯尼斯·多雷尔：克里斯塔，我们有时间最后一个问题。操作员：谢谢。这个问题来自巴克莱的罗斯·桑德勒。请继续。罗斯·桑德勒：很棒，马克，周一你在接受詹森采访时提到，如果有一天扩展停止，你们会有五年的产品工作要做。除了代理或AI助手之外，你在五年路线图中还在考虑哪些AI领域？其次，也许再问一下关于容量的问题。你们说Llama 3.1是在16000个H100上训练的。你们还说到年底将有600000个GPU可用。所以即使我们将Llama 4的计算量增加到160000个GPU，我们也有足够的额外容量用于推理和未来的产品。你们是如何建模整个资本支出路线图的，包括训练、推理和未来的事情？谢谢。马克·扎克伯格：我可以先回答第一个问题，然后让苏珊回答第二个问题。这是一个有趣的问题，有点假设性，因为我确实认为，如果没有未来的基础模型，我认为行业会带来大量的产品创新，这需要时间。但同时，未来会有基础模型，它们会很棒，解锁新的能力，我们正在围绕这些规划我们的产品。所以我并没有假设没有未来创新来规划我们的产品路线图。相反，我们正在基于我们认为最重要的能力规划Llama 4、Llama 5及其后的产品路线图，以实现我刚刚描述的效用广度，使其成为像Meta AI这样的东西，使企业、创作者和个人能够建立任何他们想要的AI代理，使你能够随时随地使用这些实时的、多模式的眼镜，这些眼镜将越来越有用。我猜这与我期望苏珊谈到的下一点有关——我们确实有一套关于人们想要发现内容和与朋友互动以及企业接触人们的现有用例，所有这些东西都在变得更好。所以——我想我在那里说的意思是——从技术变得可用到产品在空间中被充分探索之间存在一些滞后。我只是认为这是我说的这将是一个非常令人兴奋的领域，未来将有很多创新的方式。我会让苏珊来谈谈GPU和所有这些方面的数字。苏珊·李：谢谢。我们显然正在建设大量的容量，马克在他的评论中提到了我们之前的训练需求和Llama的下一代需求。我们谈到2025年的资本支出显著增长，这是驱动因素之一。我们现在还不能分享更长期的展望。当我们考虑任何新的数据中心项目时，我们会考虑它在数据中心生命周期中的使用情况。我们会考虑它在训练Llama的后续几代中的容量需求，并围绕这一点进行架构设计。然后我们也会考虑它在生命周期中的其他用途，包括生成AI推理、核心广告和内容排名及推荐的支持需求，以及不同类型的服务器支持这些不同用例的影响。所以我们真的在进行跨越广泛潜在用例的规划。我们会考虑到长时间跨度，因为数据中心启动的准备时间很长，也认识到每个数据中心生命周期中有多个决策点，思考什么时候订购服务器以及订购哪些服务器，这给我们灵活性以便基于未来的信息做出最佳决策。肯尼斯·多雷尔：感谢大家今天的参与。我们感谢你们的时间，并期待与你们再次交谈。操作员：本次电话会议到此结束。感谢你们的参与，现在可以断开连接。
3 会议纪要全文（英文）
Meta Platforms, Inc. (NASDAQ:META) Q2 2024 Earnings Conference Call July 31, 2024 5:00 PM ETKenneth Dorell - Director of IRMark Zuckerberg - Founder, Chairman and CEOSusan Li - CFOConference Call ParticipantsBrian Nowak - Morgan StanleyMark Shmulik - Bernstein ResearchEric Sheridan - Goldman SachsDouglas Anmuth - JPMorganJustin Post - Bank of AmericaMark Mahaney - Evercore ISIYoussef Squali - Truist SecuritiesRonald Josey - CitiRoss Sandler - BarclaysGood afternoon. My name is Krista and I will be your conference operator today. At this time, I would like to welcome everyone to the Meta Second Quarter Earnings Conference Call. All lines have been placed on mute to prevent any background noise. After the speakers’ remarks there will be a question-and-answer session. [Operator Instructions] This call will be recorded.Thank you very much. Kenneth Dorell, Meta's Director of Investor Relations, you may begin.Thank you. Good afternoon and welcome to Meta Platform's Second Quarter 2024 Earnings Conference Call. Joining me today to discuss our results are Mark Zuckerberg, CEO, and Susan Li, CFO.Before we get started, I would like to take this opportunity to remind you that our remarks today will include forward-looking statements. Actual results may differ materially from those contemplated by these forward-looking statements. Factors that could cause these results to differ materially are set forth in today's earnings press release and in our Quarterly Report on Form 10-Q, filed with the SEC. Any forward-looking statements that we make on this call are based on assumptions as of today and we undertake no obligation to update these statements as a result of new information or future events.During this call, we will present both GAAP and certain non-GAAP financial measures. A reconciliation of GAAP to non-GAAP measures is included in today's earnings press release. The earnings press release and an accompanying Investor Presentation are available on our website at investor.fb.com.And now I'd like to turn the call over to Mark.All right, thanks Ken. And hey everyone, thanks for joining today. This was a strong quarter for our community and business. We estimate that there are now more than 3.2 billion people using at least one of our apps each day. The growth we're seeing here in the US has especially been a bright spot. WhatsApp now serves more than 100 million monthly actives in the US, and we're seeing good year-over-year growth across Facebook, Instagram, and Threads as well, both in the US, and globally.I'm particularly pleased with the progress that we're making with young adults on Facebook. The numbers we are seeing, especially in the US, really go against the public narrative around who's using the app. A couple of years ago, we started focusing our apps more on 18 to 29 year olds and it's good to see that those efforts are driving good results. Another bright spot is Threads which is about to hit 200 million monthly actives. We're making steady progress towards building what looks like it's going to be another major social app. And we are seeing deeper engagement, and I'm quite pleased with the trajectory here.The big theme right now is, of course, AI. And I'll focus my comments today on three areas. What AI means for our family of apps and core business. What new AI experiences and opportunities we see, and how AI is shaping our metaverse work. So let's start. Across Facebook and Instagram, advances in AI continue to improve the quality of recommendations and drive engagement.And we keep finding that as we develop more general recommendation models, content recommendations get better. In this quarter, we rolled out our full screen video player and unified video recommendation service across Facebook, bringing Reels, longer videos, and live into a single experience. And this has allowed us to extend our unified AI systems, which had already increased engagement on Facebook Reels more than our initial move from CPUs to GPUs did. Over time I'd like to see us move towards a single unified recommendation system that powers all of the content including things like people you may know, across all of our surfaces. We're not there yet. They're still upside, and we're making good progress here.AI is also going to significantly evolve our services for advertisers in some exciting ways. It used to be that advertisers came to us with a specific audience they wanted to reach, like a certain age group, geography, or interests. Eventually, we got to the point where our ad systems could better predict who would be interested than the advertisers could themselves. But today, advertisers still need to develop creative themselves. And in the coming years, AI will be able to generate creative for advertisers as well. And we'll also be able to personalize it as people see it.Over the long term, advertisers will basically just be able to tell us a business objective and a budget, and we're going to go do the rest for them. We're going to get there incrementally over time, but I think this is going to be a very big deal. Moving on to some of the brand new experiences that AI enables, last quarter we started broadly rolling out our assistant Meta AI, and it is on track to achieve our goal of becoming the most used AI assistant by the end of the year.We have an exciting roadmap ahead of things that we want to add, but the bottom-line here is that Meta AI feels like it is on track to be an important service and it's improving quickly both in intelligence and features. Some of the use cases are utilitarian like searching for information or role-playing difficult conversations before you have them with another person and other uses are more creative like the new imagine yourself feature that lets you create images of yourself doing whatever you want in whatever style you want and part of the beauty of AI is that it's general. So we're still uncovering the wide range of use cases that it's valuable for.An important part of our vision is that we're not just creating a single AI, but enabling lots of people to create their own AIs. And this week we launched AI Studio, which lets anyone create AIs to interact with across our apps. I think the creators are especially going to find this quite valuable. There are millions of creators across our apps, and these are people who want to engage more with their communities, and their communities want to engage more with them, but there are only so many hours in the day.So now they are going to be able to use AI Studio to create AI agents that can channel them to chat with their community, answer people's questions, create content and more. So I'm quite excited about this. But this goes beyond creators too. Anyone is going to be able to build their own AIs based on their interests or different topics that they are going to be able to engage with or share with their friends. Business AIs are the other big piece here. We're still in Alpha testing with more and more businesses. The feedback we're getting is positive so far. Over time, I think that just like every business has a website, a social media presence, and an email address, in the future I think that every business is also going to have an AI agent that their customers can interact with. And our goal is to make it easy for every small business, eventually every business, to pull all of their content and catalog into an AI agent that drives sales and saves them money.When this is working at scale, I think that this is going to dramatically accelerate our business messaging revenue. There are a lot of other new opportunities here that I'm excited about too, but I'll save those for another day when we're ready to roll them out. The engine that powers all these new experiences is the Llama family of foundation models. In this quarter we released Llama 3.1 which includes the first frontier level open source model as well as new and industry leading small and medium sized models.The $405 billion model has better cost performance relative to the leading closed models, and because it's open, it is immediately the best choice for fine-tuning and distilling your own custom models of whatever size you need. I think we are going to look back at Llama 3.1 as an inflection point in the industry where open source AI started to become the industry standard, just like Linux is. I often get asked why I'm so bullish on Open Source. I wrote a letter along with the Llama 3.1 release, explaining why I believe that Open Source is better for developers, for Meta app, and for the world more broadly.My view is that Open Source will be safer, will enable innovation that improves all of our lives faster, and we'll also create more shared prosperity. For Meta's own interests, we're in the business of building the best consumer and advertiser experiences. And to do that, we need to have access to the leading technology infrastructure and not get constrained by what competitors will let us do. But these models are ecosystems. They're not just isolated pieces of software that we can develop by ourselves.So if we want the most robust ecosystem of tools, efficiency improvements, silicon optimizations, and other integrations to develop around our models, then we need them to be widely used by developers across the industry. And once we know that we're going to have access to the leading models, then I'm confident that we are going to be able to build the best social and advertising experiences. Part of why I'm so optimistic about this is that we have a long track record of success with open source.We've saved billions of dollars with open compute project by having supply chains standardized on our infrared designs. Open sourcing tools like PyTorch and React has led to real benefits for us from all the industry's contributions. This approach has consistently worked for us and I expect it will work here too. Another major area of focus is figuring out the right level of infra capacity to support training more and more advanced models. Llama 3 is already competitive with the most advanced models, and we're already starting to work on Llama 4, which we're aiming to be the most advanced in the industry next year.We are planning for the compute clusters and data we'll need for the next several years. The amount of compute needed to train Llama 4 will likely be almost 10 times more than what we used to train Llama 3, and future models will continue to grow beyond that. It's hard to predict how this trend -- how this will trend multiple generations out into the future. But at this point, I'd rather risk building capacity before it is needed rather than too late, given the long lead times for spinning up new inference projects.And as we scale these investments, we're of course, going to remain committed to operational efficiency across the company. The last area that I want to discuss is how AI is shaping our metaverse work, which continues to be our other long-term focus. Last quarter, I discussed how advances in AI have pulled in the timelines for some of our products.A few years ago, I would have predicted that holographic AR would be possible before Smart AI, but now it looks like those technologies will actually be ready in the opposite order. We're well positioned for that because of the Reality Labs investments that we've already made. Ray-Ban Meta Glasses continue to be a bigger hit sooner than we expected, thanks in part to AI. Demand is still outpacing our ability to build them, but I'm hopeful that we'll be able to meet that demand soon.EssilorLuxottica has been a great partner to work with on this, and we are excited to team up with them to build future generations of AI glasses, as we continue to build our long-term partnership. Quest 3 sales are also outpacing our expectations. And I think that's because it is not just the best MR headset for the price, but it's the best headset on the market, period. In addition to gaming, people are increasingly taking advantage of Quest's capabilities as a general computing platform, spending time watching videos, browsing websites, extending their PC via virtual desktop, and more. Horizon also continues to grow across VR, mobile, and desktop, and I expect that it will become an increasingly important part of that ecosystem as well. We're hosting our Annual Connect Conference on September 25th, and we will have lots of exciting updates around all of our AI and Metaverse work, so I encourage you to tune into that.At the end of the day, we are in the fortunate position where the strong results that we're seeing in our core products and business give us the opportunity to make deep investments for the future. And I plan to fully seize that opportunity to build some amazing things that will pay off for our community and our investors for decades to come. The progress we're making on both the foundational technology and product experiences suggests that we're on the right track. I'm proud of what our team has accomplished so far, and I'm optimistic about our ability to execute on the opportunities ahead. As always, thank you to our teams who are pushing all this important work forward, and thanks to all of you for being on this journey with us.Thanks Mark and good afternoon everyone. Let's begin with our consolidated results. All comparisons are on a year-over-year basis unless otherwise noted. Q2 total revenue was $39.1 billion up 22% or 23% on a constant currency basis. Q2 total expenses were $24.2 billion, up 7% compared to last year. In terms of the specific line items, cost of revenue increased 23% driven primarily by higher infrastructure and reality labs inventory costs.R&D increased 13%, primarily driven by higher headcount-related expenses and infrastructure costs which were partially offset by lower restructuring costs. Marketing and sales decreased 14%, due mainly to lower restructuring and headcount-related costs. G&A decreased 12%, mostly due to lower legal-related expenses. We ended the first quarter with almost 70,800 employees, up 2% from Q1. Second quarter operating income was $14.8 billion, representing a 38% operating margin.Our tax rate for the quarter was 11%. Net income was $13.5 billion, or $5.16 per share. Capital expenditures, including principal payments on finance leases, were $8.5 billion, driven by investments in servers, data centers, and network infrastructure. Free cash flow was $10.9 billion. We repurchased $6.3 billion of our Class A common stock and paid $1.3 billion in dividends to shareholders, ending the quarter with $58.1 billion in cash and marketable securities and $18.4 billion in debt.Moving now to our segment results. I'll begin with our Family of Apps segment. Our community across the Family of Apps continues to grow, with approximately 3.27 billion people using at least one of our Family of Apps on a daily basis in June. Q2 total Family of Apps revenue was $38.7 billion, up 22% year-over-year. Q2 Family of Apps ads revenue was $38.3 billion, up 22% or 23% on a constant currency basis. Within ad revenue, the online commerce vertical was the largest contributor to year-over-year growth, followed by gaming and entertainment and media.On a user geography basis, ad revenue growth was strongest in rest of world and Europe at 33% and 26% respectively. Asia Pacific grew 20% and North America grew 17%. On an advertiser geography basis, total revenue growth continued to be strongest in Asia Pacific at 28%. The growth was below the first quarter rate of 41%, as we lapped a period of stronger demand from China-based advertisers.In Q2, the total number of ad impressions served across our services and the average price per ad both increased 10%. Impression growth was mainly driven by Asia Pacific and rest of world. Pricing growth was driven by increased advertiser demand in part due to improved ad performance. This was partially offset by impression growth particularly from lower monetizing regions and surfaces.Family of Apps other revenue was $389 million, up 73%, driven primarily by business messaging revenue growth from our WhatsApp business platform. We continue to direct the majority of our investments toward the development and operation of our Family of Apps. In Q2, Family of Apps expenses were $19.4 billion, representing approximately 80% of our overall expenses.Family of Apps expenses were up 4%, mostly due to higher infrastructure and headcount related expenses, which were partially offset by lower restructuring costs. Family of Apps operating income was $19.3 billion, representing a 50% operating margin. Within our Reality Labs segment, Q2 was $353 million, up 28% driven primarily by Quest headset sales. Reality Labs expenses were $4.8 billion, up 21% year-over-year, driven mainly by higher headcount-related expenses and Reality Labs inventory costs. Reality Labs operating loss was $4.5 billion.Turning now to the business outlook. There are two primary factors that drive our revenue performance, our ability to deliver engaging experiences for our community and our effectiveness at monetizing that engagement over time. To deliver engaging experiences, we remain focused on executing our priorities, including video and in-feed recommendations. On Instagram, Reels engagement continues to grow as we make ongoing enhancements to our recommendation systems. Part of this work has been focused on increasing the share of original posts within recommendations so people can discover the best of Instagram, including content from emerging creators.Now, more than half of recommendations in the US come from original posts. On Facebook, we're seeing encouraging early results from the global rollout of our unified video player and ranking systems in June. This initiative allows us to bring all video types on Facebook into one viewing experience, which we expect will unlock additional growth opportunities for short-form video, as we increasingly mix shorter videos into the overall base of Facebook video engagement.We expect the relevance of video recommendations will continue to increase as we benefit from unifying video ranking across Facebook and integrating our next-generation recommendation systems. These have already shown promising gains since we began using the new systems to support Facebook Reels recommendations last year. We expect to expand these new systems to support more services beyond Facebook video over the course of this year and next year. We are also seeing good momentum with our longer-term engagement priorities, including Generative AI and Threads. People have used Meta AI for billions of queries since we first introduced it.We're seeing particularly promising signs on WhatsApp in terms of retention and engagement, which has coincided with India becoming our largest market for Meta AI usage. You can now use Meta AI in over 20 countries and eight languages, and in the US we are rolling out new features like Imagine Edit, which allows people to edit images they generate with Meta AI. Beyond Generative AI, the Threads community also continues to grow and deepen their engagement, as we ship new features and enhance our content recommendation systems.Now to the second driver of our revenue performance, increasing monetization efficiency. There are two parts to this work. The first is optimizing the level of ads within organic engagement. We continue to see opportunities to grow ad supply on lower monetizing surfaces like video, including within Facebook as the mix of overall video engagement shifts more to shorter videos over time, which creates more ad insertion opportunities. More broadly, we are continuing to get better at determining the best ads to show and when to show them during a person session across both Facebook and Instagram. This is enabling us to drive revenue growth and conversions without increasing the number of ads or in some cases even reducing ad load.The second part of improving monetization efficiency is enhancing marketing performance. We continue to be pleased with our progress here, with AI playing an increasingly central role. We're improving ad delivery by adopting more sophisticated modeling techniques made possible by AI advancements, including our Meta Lattice ad ranking architecture, which continued to provide ad performance and efficiency gains in the second quarter.We're also making it easier for advertisers to maximize ad performance and automate more of their campaign setup with our Advantage+ suite of solutions. We're seeing these tools continue to unlock performance gains, with a study conducted this year demonstrating 22% higher return on ad spend for US advertisers after they adopted Advantage+ Shopping campaigns.Advertiser adoption of these tools continues to expand and we are adding new capabilities to make them even more useful. For example, this quarter we introduced flexible format to Advantage+ Shopping, which allows advertisers to upload multiple images and videos in a single ad that we can select from and automatically determine which format to serve in order to yield the best performance.We have also now expanded the list of conversions that businesses can optimize for using Advantage+ shopping to include an additional 10 conversion types, including objectives like add to cart. Looking forward, we believe Generative AI will play a growing role in how businesses market and engage with customers at scale. We expect this technology will continue to make it easier for businesses to develop customized and diverse ad creatives. We've seen promising early results since introducing our first Generative AI ad features, image expansion, background generation, and text generation with more than 1 million advertisers using at least one of these solutions in the past month.In May, we began rolling out full image generation capabilities into Advantage+ Creative, and we're already seeing improved performance from advertisers using the tool. Finally, we expect AI will help businesses communicate with customers more efficiently through messaging. We're starting by testing the ability for businesses to use AI in their chats with customers to help sell their goods and services and to generate leads. While we are in the early stages, we continue to expand the number of advertisers we are testing with and have seen good advances in the quality of responses since we began using Llama 3.Next, I’d like to discuss our approach to capital allocation which remains unchanged. We continue to invest both in enhancing our core experiences in the near-term and developing technologies that we believe will transform how people engage with our services in the years ahead. We expect that having sufficient compute capacity will be central to many of these opportunities. So we’re investing meaningfully in infrastructure to support our core AI work in content ranking and ads, as well as our generative AI and advanced research efforts.Our ongoing investment in core AI capacity is informed by the strong returns we've seen and expect to deliver in the future, as we advance the relevance of recommended content and ads on our platform. While we expect the returns from Generative AI to come in over a longer period of time, we’re mapping these investments against the significant monetization opportunities that we expect to be unlocked across customized ad creative, business messaging, a leading AI assistant and organic content generation.As we scale generative AI training capacity to advance our foundation models, we’ll continue to build our infrastructure in a way that provides us with flexibility in how we use it over time. This will allow us to direct training capacity to gen AI inference or to our core ranking and recommendation work, when we expect that doing so would be more valuable. We will also continue our focus on improving the cost efficiency of our workloads over time. Reality Labs remains our other long-term initiative that we continue to invest meaningfully in.Quest 3 is selling well and Ray-Ban Meta smart glasses are showing very promising traction with the early signals that we’re seeing across demand, usage and retention increasing our confidence in the long-run potential of AR glasses. Finally, as we pursue these investments across near and long-term priorities, we will remain focused on operating the business efficiently.Turning now to the revenue outlook. We expect third quarter 2024 total revenue to be in the range of $38.5 billion to $41 billion. Our guidance assumes foreign currency is a 2% headwind to year-over-year total revenue growth based on current exchange rates. Turning now to the expense outlook. We expect full year 2024 total expenses to be in the range of $96 billion to $99 billion, unchanged from our prior outlook.For Reality Labs, we continue to expect 2024 operating losses to increase meaningfully year-over-year due to our ongoing product development efforts and investments to scale our -- to further scale our ecosystem. While we do not intend to provide any quantitative guidance for 2025 until the fourth quarter call, we expect infrastructure costs will be a significant driver of expense growth next year. As we recognize depreciation and operating costs associated with our expanded infrastructure footprint.Turning now to the CapEx outlook. We anticipate our full year 2024 capital expenditures will be in the range of $37 billion to $40 billion, updated from our prior range of $35 billion to $40 billion. While we continue to refine our plans for next year, we currently expect significant CapEx growth in 2025 as we invest to support our AI research and our product development efforts.On to tax. Absent any changes to our tax landscape, we expect our full year 2024 tax rate to be in the mid-teens. In addition we continue to monitor an active regulatory landscape, including the increasing legal and regulatory headwinds in the EU and the US that could significantly impact our business and our financial results.In closing, Q2 was another good quarter. We continue to execute well across our business priorities and have exciting opportunities in front of us to deliver more value to the people and businesses using our products around the world.With that, Krista let's open up the call for questions.Question-and-Answer SessionThank you. We will now be open the line for question-and-answer session [Operator Instructions] And your first question comes from the line of Brian Nowak from Morgan Stanley. Please go ahead.Great. Thanks for taking my questions. I have two, one for Mark, one for Susan. Mark I wanted to sort of go back to some of the new generative AI-enabled use cases for users and advertisers. You talked about Meta AI, Studio, chatbots, [foundation] (ph) models. If you could just sort of hone in on one or two of those that you are most excited about, we are seeing good signal that could be a real driver for the business in '25, '26 just so we sort of know where are you most focused on all those opportunities it would be helpful.And the second one, Susan, you have a lot of CapEx priorities from building new infrastructure for next-generation models, compute capacity. Just walk us through again on the CapEx philosophy and any guardrails you have around ensuring you generate a healthy return on invested capital for investors from all the CapEx. Thanks.I can take the first one. So I think the things that will drive the most results in 2025 and 2026 are actually the first category of things that I talked about in my comments which are the ways that AI is shaping the existing products. So the ways that it is improving recommendations and helping people find better content, as well as making the advertising experiences more effective. I think there is a lot of upside there. Those are already products that are at scale. The AI work that we are doing is going to improve that. It will improve the experience and the business results.The other areas that we are working on, I mean I think you all know this from following our business for a while, but we have a relatively long business cycle of starting a new product, scaling it to something that reaches 1 billion people or more and only then really focusing on monetizing at scale. So realistically, for things like Meta AI or AI Studio, I mean these are things that I think will increase engagement in our products and have other benefits that will improve the business and engagement in the near term.But before we are really talking about monetization of any of those things by themselves, I mean I don't think that anyone should be surprised that I would expect that -- that will be years, right? It's just -- I think that that's like what we've seen with Reels, it's what we saw with all these things. But I think for those who have followed our business for a long time, you can also get a pretty good sense of when things are going to work years in advance. And I think that the people who bet on those early indicators tend to do pretty well, which is why I wanted to share in my comments the early indicator that we had on Meta AI, which is I mean look, it is early.Last quarter, we -- I think it just started rolling it out a week or two before our earnings call. This time we are a few months later. And what we can say is I think we are on track to achieve our goal of being the most used AI assistant by the end of this year. And I think that is a pretty big deal. Is that the only thing we want to do? No. I mean we obviously want to kind of grow that and grow the engagement on that to be a lot deeper, and then we will focus on monetizing it over time.But the early signals on this are good and I think that -- that's kind of all that we could reasonably have insight into at this point. But I do think that part of what's so fundamental about AI is, it is going to end up affecting almost every product that we have in some way. It will improve the existing ones and will make a whole lot of new ones possible. So it is why there are all the jokes about how all the tech CEOs get on these earnings calls and just talk about AI the whole time. It is because it is actually super exciting, and it is going to change all these different things over multiple time horizons.And Brian, I can take the second question. On the ROI part of your question, I’d broadly characterize our AI investments into two buckets; core AI and Gen AI. And the two are really at different stages, as it relates to driving revenue for our businesses and our ability to measure returns. On our core AI work, we continue to take a very ROI based approach to our investment here. We are still seeing strong returns as improvements to both engagement and ad performance have translated into revenue gains and it makes sense for us to continue investing here.Gen AI is where we are much earlier, as Mark just mentioned in his comments. We don't expect our Gen AI products to be a meaningful driver of revenue in 2024. But we do expect that they are going to open up new revenue opportunities over time that will enable us to generate a solid return off -- of our investment while we are also open sourcing subsequent generations of Llama. And we've talked about the four primary areas that we are focused here on the Gen AI opportunities to enhance the core ads business, to help us grow in business messaging, the opportunities around Meta AI, and the opportunities to grow core engagement over time.The other thing I’d say is, we are continuing to build our AI infrastructure with fungibility in mind, so that we can flex capacity where we think it will be put to best use. The infrastructure that we build for gen AI training can also be used for Gen AI inference. We can also use it for ranking and recommendations by making certain modifications like adding general compute and storage. And we are also employing a strategy of staging our data center sites, at various phases of development, which allows us to flex up to meet more demand and less lead time if needed while limiting how much spend we are committing to in the outer years.So while we do expect that we are going to grow CapEx significantly in 2025, we feel like we have a good framework in place in terms of thinking about where the opportunities are and making sure that we have the flexibility to deploy it, as makes the most sense.Your next question comes from the line of Mark Shmulik with Bernstein. Please go ahead.Yes. Hi, thanks for taking my question. Just as we look at the revenue guidance and the outlook, Susan, any color you can share on just kind of the state of the overall digital ad market? And you've highlighted some areas where you are seeing strength versus kind of some of the idiosyncratic efforts you've made to kind of improve the efficacy of the ad product. Thank you.Hi, Mark. We are continuing to see healthy global advertising demand, and we are also delivering ongoing ad performance improvements just related to all of the investments that we've continued to make over time. And improving the sort of ads, targeting ranking, delivery, all of the fundamental infrastructure there. And we expect that all of that will continue to benefit ad spend in Q3. We do expect year-over-year growth to slow in Q3, as we are lapping strong growth from China-based advertisers, as well as strong Reels impression growth from a year ago. And we also expect modestly larger FX headwinds in Q3 based on current rates.Your next question comes from the line of Eric Sheridan with Goldman Sachs. Please go ahead.Thank you so much for taking the question. I'll just ask one. You called out building community size and what's happened in the United States, as well as Threads. How are you thinking about those newer faster-growing elements of either Messaging or Threads as a platform and the mix between the potential for engagement growth and overall monetization longer term of either the messaging layer or Threads and what you are most excited about there to build to sort of capitalize on scale but bring it back towards monetization? Thank you.I can start and Susan can jump in, if she has anything else that she wants to add. So the WhatsApp stat I think, is really important as a business trend just because the United States punches above its weight in terms of, it is such a large percent of our revenue. So before, WhatsApp was sort of the leading messaging app in many countries around the world but not in the US. And I think now that we're starting to make inroads into leading in the US as more and more people use the product and realize that, hey, it was a really good experience, the best experience for cross-platform communication and groups and on all these different things. I think that -- that's going to just mean that all of the work that we are doing to grow the business opportunity there over time is just going to have a big tailwind if the US ends up being a big market. So that's one reason why it's really relevant.It is obviously also personally somewhat gratifying to see all the people around us starting to use WhatsApp, so I think that is pretty fun but maybe somewhat less relevant from a business perspective. Threads, I think it is another example of something that it got off to about as good of a start of any app that I can think of. I think, it was the fastest growing app to 100 million people. And it is a good reminder that even when you have that start, the path from there to 1 billion people using an app is still multiple years. And that's our product cycle.And I think that -- that's something that is a little bit different about Meta in the way we build consumer products and the business around them than a lot of other companies that ship something and start selling it and making revenue from it immediately. So I think that's something that our investors and folks thinking about analyzing the business, if needed to always grapple with, is all these new products, we ship them and then there is a multiyear time horizon between scaling them and then scaling them into not just consumer experiences but very large businesses.But the thing that I think is just super exciting about Threads is that we've been building this company for 20 years, and there are just not that many opportunities that come around to grow 1 billion person app. I mean, there are -- I don't know, maybe a dozen of them in the world or something, right? I mean, there are certainly more of them outside the company than inside the company, but we do pretty well and being able to add another one to the portfolio if we execute really well on this is just really exciting to have that potential.Now obviously, there is a ton of work between now and there. I mean, we are almost at 200 million. So it is a really good milestone, I'm excited about that. A lot of work between this and it being a large part of the business. But I do think that these kind of opportunities are pretty rare and that's something that we are just really excited about. I think the team is doing great work on it.Eric, I would just add to that in terms of [nearer-term] (ph) sources of impression growth, we really expect that video is going to remain a source of impression growth for us in the second half. On Instagram, we expect Reels to continue to drive growth, while on Facebook, we expect to grow overall video time, while increasing the mix of short-form video, which creates more impression growth opportunities. And generally we expect continued community growth foracross our apps.Your next question comes from the line of Doug Anmuth with JPMorgan. Please go ahead.Hi, thanks for taking my questions. One for Mark, one for Susan. Mark just in terms of infrastructure and CapEx, you've talked about currently building out not just for Llama 3 and 4 but really out to 7 perhaps and then Llama 4, 10x the compute required versus Llama 3. Just given how much you are building ahead, how does that influence the shape of the CapEx curve over a multiyear period? And then Susan, if you could talk a little bit more about the 3Q outlook. I know you are talking about tougher comps, but at the same time, it really suggests only 1 point of FX neutral [decel] (ph) at the high end. So just curious if there's anything else you can point to more specifically that's driving the expected strength here. Thanks.Thanks, Doug. I can go ahead and talk about both of those. So your first question was sort of about the longer-term CapEx outlook. We haven't really shared an outlook sort of on the longer-term CapEx trajectory. In part infrastructure is an extraordinarily dynamic planning area for us right now. We are continuing to work through what the scope of the Gen AI road maps will look like over that time. Our expectation obviously again, is that we are going to significantly increase our investments in AI infrastructure next year, and we'll give further guidance as appropriate.But we are building all of that CapEx, again with the factors in mind that I talked about previously thinking about both how to build it flexibly, so we can deploy to core AI and Gen AI use cases as needed. And making sure that we both feel good about the returns that we're seeing on the core AI investments, which we are able to measure more immediately. And then we feel good about the opportunities in the gen AI efforts.Your second question was about the Q3 revenue outlook. Again I mentioned this earlier. We have seen healthy global advertising demand on our platform. We are delivering ongoing ad performance improvements, which again we feel like is a result of many, many quarters of effort that have accrued and will continue to accrue value to our platform. And we saw basically in Q2 where revenue grew 22% that there was broad-based strength across regions and verticals including particular strength among smaller advertisers, and we expect that generally to continue into Q3.Your next question comes from the line of Justin Post with Bank of America. Please go ahead.Great. Thank you. I just want to get back to the comment on US young adult user growth, especially maybe on Facebook and Instagram. I know you made a big change with Reels a couple of years ago. But what are those users doing on Facebook and Instagram? And can you give us any quantification of the usage growth? Thank you.Thanks, Justin. So building products with young adults in mind has been a core priority area for the Facebook team in recent years, and we've been very encouraged to see these efforts translate into engagement growth with this cohort. We have seen healthy growth in young adult app usage in the US and Canada for the past several quarters. And we have seen that products like Groups and Marketplace have seen particular traction with young adults.Posting to groups in the US and Canada has been growing. That's been boosted mainly by young adults. And we also see that they are active users of Marketplace, which has benefited from product improvements and strong demand for second-hand products in the US.Your next question comes from the line of Mark Mahaney with Evercore ISI. Please go ahead.I was going to ask about Marketplace so that's a nice segue. It is a great, somewhat under-monetized or arguably very under-monetized asset. I know you indirectly monetize it and it's a very large marketplace. It may even be bigger than eBay. Your thoughts on what you may want to do in the future in terms of monetizing it, in part maybe even improve the quality of the Marketplace.And then secondly, I just want to ask you about headcount. It is down about 1% year-over-year. You are pretty much back at par with where the employee headcount was prior to significant reduction. How should we think about headcount growth going forward? Did you talk about a significant growth in CapEx? Should we expect a moderate growth in headcount significant? Any thoughts on that would be helpful. Thank you.Thanks Mark. On your first question about Marketplace, again we are obviously excited that it's been one of the drivers of strength in young adults. I would probably just say that more generally, Marketplace is one prong in a broader commerce strategy that we have which continues to be focused on basically creating the best shopping experience on our platform. Marketplace is obviously consumer oriented. The broader part of the commerce strategy is about making it easier for businesses to advertise their products, for buyers to find and purchase relevant items on our platform.And to that end, I’d say that we feel quite happy with also the investments we've been making in Shops ads. Shops ads revenue is growing at a strong year-over-year pace. We are seeing Shops ads drive incremental performance for advertisers, and it's also working well in combination with some of our other products like Advantage+ shopping.Your second question was about headcount. We continue to be disciplined about where we are allocating new headcount to ensure that it's really focused on our core company priorities, but we are also working down a prior hiring underrun. And as we further close that hiring underrun over the course of this year, I do expect that we will end 2024 with in-seat reported headcount that is meaningfully higher than where we ended 2023. We aren't providing sort of 2025 headcount growth expectations yet as we haven't started our budgeting process yet. But again, I expect that we’ll primarily target our hiring to focus on priority areas, and we will be running a very disciplined headcount process.Your next question comes from the line of Youssef Squali with Truist Securities. Please go ahead.Great. Thank you very much. So the AI system using Llama 3.1 has been incorporated in different variations and looks really impressive and seems to be getting closer to becoming a full search engine for virtually everything except for commercial queries so far. So are there any plans to open it up to the broader web? Kind of like what may be OpenAI is off to testing, maybe link it to third-party marketplaces for commercial search, et cetera.And then on Ray-Ban, can you maybe talk a little bit more about the opportunity to deepen your relationship with EssilorLuxottica? What would that look like? What kind of areas are the most exciting to you, Mark in that relationship? Thank you.Yes. I'm very excited with how Llama 3.1 landed. I think the team there is doing really great work going from the first version of Llama, the Llama 2 last year that was a generation behind the frontier and now Llama 3.1, which is basically competitive and in some ways, leading the other top-closed models. Meta AI uses a version of Llama 3.1 as well as a bunch of other services that we've built to kind of build a cohesive product. And when I was talking before about we have the initial usage trends around Meta AI but there is a lot more that we want to add. Things like commerce and you can just go vertical by vertical and build out specific functionality to make it useful in all these different areas are eventually, I think what we're going to need to do to make this just as -- to fulfill the potential around just being the ideal AI assistant for people.So it is a long road map. I don't think, that this stuff is going to get finished in the next couple of quarters or anything like that. But this is part of what's going to happen over the next few years as we build something that will I think, just be a very widely used service. So I'm quite excited about that. And we are going to continue working on Llama 2. So I mean, you mentioned Llama, and I think the question was a little more about Meta AI but they are both -- I mean, they're related Llama is sort of like the engine that powers the product and it's open source, and I'm just excited about the progress that we are making on both of those.On the smart glasses, EssilorLuxottica is a great partner. We are now in the second generation of the Ray-Ban Meta glasses. They are doing well, better than I think we had expected, and we expected them to grow meaningfully from the first generation so that's been a very positive surprise. And I think part of that is that it is just well-positioned to dovetail well with the AI revolution that we are seeing and offering all kinds of new functionality there. So that was great.But EssilorLuxottica is a great company that has a lot of different products that we hope to be able to partner with to just continue building new generations of the glasses and deepen the AI product and make it better and better. I think there is a lot more to go from here. And compared to what we thought at this point, it's doing quite well compared to, I think what it needs to be, to be like a really leading piece of consumer electronics, I think we are still early but all the signs are good.Your next question comes from the line of Ron Josey with Citi. Please go ahead.Great. Thanks for taking the question. I want to get back, Mark, to the commentary on open source and Llama 3. Totally understand that Meta is not offering a public cloud, and so what I wanted to hear from you is maybe a little bit more on the product vision of products that come out of Llama 3. And meaning potentially offering some of these products to other companies, call it for customer service or call center offerings or other verticals. And so any insights on just how you envision maybe the open source and Llama 3.1, can sort of offer greater enterprise services for others to benefit from? Thank you.So Llama is the foundation model that people can shape into all kind of different products. So whether it's Meta AI for ourselves or the AI Studio or the business agents or like the assistant that's in the Ray-Ban glasses, like all these different things are basically products that have been built with Llama. And similarly, any developer out there is going to be able to take it and build a whole greater diversity of different things as well.Like I talked about, I think -- the reason why open sourcing this is so valuable for us is that we want to make sure that we have the leading infrastructure to power the consumer and business experiences that we are building. But the infrastructure, it's not just a piece of software that we can build in isolation. It really is an ecosystem with a lot of different capabilities that other developers and people are adding to the mix, whether that's new tools to distill the models into the size that you want for a custom model, or ways to fine-tune things better or make inference more efficient or all different other kinds of methods that we haven't even thought of yet, the silicon optimizations that the silicon companies are doing, all the stuff. It is an ecosystem.So we can't do all that ourselves. And if we built Llama and just kind of kept it within our walls, then it wouldn't actually be as valuable for us to build all the products that we are building as it is going to end up being. So that's the business strategy around that. And that's why we don't feel like we need to necessarily build a cloud and sell it directly in order for it to be a really positive business strategy for us. And part of what we are doing is working closely with AWS, I think, especially did great work for this release.Other companies like Databricks, Nvidia, of course, other big players like Microsoft with Azure, and Google Cloud, they are all supporting this. And we want developers to be able to get it anywhere. I think that's one of the advantages of an open source model like Llama is – it is not like you're locked into one cloud that offers that model, whether it's Microsoft with OpenAI or Google with Gemini or whatever it is, you can take this and use it everywhere and we want to encourage that. So I'm quite excited about that.The enterprise and business applications that we are going to be most focused on, though, in addition to just optimizing the advertiser experience like I talked about in my comments earlier, it is the business agent piece. I just think that there is a huge potential like I said earlier. I think pretty much every business today, it has an e-mail address. They have a website. They have social media accounts. I think in the future, they are going to have at least one, if not multiple business agents that can do the whole range of things from interacting to help people buy things to helping support the sales that they've done, if they have issues with the product, if they need to get in touch with you for something.And we already see people interacting with businesses over messaging working quite well in countries that have low cost of labor. But the thing is that in order to have someone answering everyone's questions is quite expensive in a lot of countries. And I think that this is like a thing that AI, I think is just going to be very well suited towards doing. And when we can make that easy for the hundreds of millions of businesses that use our platforms to pull in all their information and their catalogs, and the history and all the content that they've shared and really just quickly stand up an agent, I think that's going to be awesome.So we can combine Llama with a lot of custom work that we're doing in our business teams and couple that with all the other investment that the rest of the ecosystem is doing to make Llama good. And I think that's going to be huge, but that's just one area. I mean, this really goes across all of the different products that we are building, both consumer and business. So it is a lot of exciting stuff.Krista, we have time for one last question.Thank you. That question will come from the line of Ross Sandler with Barclays. Please go ahead.Great. Mark, so on Monday in your interview with Jensen, you said something along the lines of if scaling ended up stopping one day, you'd have five years of product work to do and ahead of you. So outside of agents or AI assistants, what other areas in AI are you thinking about or looking at in that five-year road map? And then the second question is maybe one more stab at the capacity question.You guys said that Llama 3.1 was trained on 16,000 H100s. You've also said that you're going to have 600,000 available by year-end. So even if we kind of go up to 160,000 GPUs for Llama 4, we have plenty of extra capacity for inference and future products. I guess how are you guys modeling out this entire kind of CapEx road map between training, inference, and future things? That's all. Thanks a lot.I can start with the first one, and then I'll let Susan answer the second one. It is an interesting question. It is a little hypothetical because I mean, I do think it's true that if, let's say, there were no future foundation models. I think there would just be a huge amount of product innovation that the industry would bring to bear, and that just takes time. But then at the same time there are going to be future foundation models, and they are going to be awesome and unlock new capabilities and we are planning our products around those.So I'm not really planning our product road map assuming that there isn't future innovation. On the contrary, we are planning what's going to be in Llama 4 and Llama 5 and beyond based on what capabilities we think are going to be most important for the road map that I just laid out for having the breadth of utility that you're going to need in something like Meta AI, making it set businesses and creators and individuals can stand up any kind of AI agents that they want, that you are going to have these kind of real-time, multimodal glasses with you all the time that will just be increasingly useful for all the things that you're doing.And that -- I guess this kind of dovetails what I'd expect Susan to talk about next. And we do have this huge set of use cases already about people wanting to discover content and interact with their friends and businesses reaching people, and all that stuff is getting better with this, too. So there is -- I guess my point there was its just – there is some lag between the technology becoming available and the products becoming kind of fully explored in the space. And I just think that -- that was kind of my way of saying that I think that this is just a very exciting area where there's just going to be a lot of innovation for a long time to come. I'll let Susan take a stab at the numbers around the GPUs and all that.Thank you. We are clearly in the process of building out a lot of capacity, and Mark has alluded to that in his comments about what we have needed to train prior generations and the next generation of Llama. And we are -- and that's driving sort of what we've talked about in terms of the significant growth in CapEx in 2025. And we aren't really in the position now to share a longer-term outlook.When we think about sort of any given new data center project that we are constructing, we think about how we will use it over the life of the data center. We think about the amount of capacity we would use in terms of training whatever the subsequent generations of Llama are and it is architected around that. But then we also look at how we might use it several years into its lifetime towards other use cases across our core business, across what we think might be future needs for inference, for generative AI-based products.So there is sort of a whole host of use cases for the life of any individual data center ranging from gen AI training at its outset to potentially supporting gen AI inference to being used for core ads and content ranking and recommendation and also thinking through the implications, too, of what kinds of servers we might use to support those different types of use cases.So we are really mapping across a wide range of potential use cases when we undertake any given project. And we are really doing that with both a long time horizon in mind, again, because of the long lead times in spinning up data centers, but also recognizing that there are multiple decision points in the lifetime of each data center in terms of thinking through when to order servers and what servers to order and what you will put them towards. And that gives us flexibility to make the sort of the best decisions based on the information we have in the future.Great. Thank you all for joining us today. We appreciate your time, and we look forward to speaking with you again soon.This concludes today's conference call. Thank you for your participation, and you may now disconnect.![]()
