Horowitz Says AI Infrastructure Requires Complete Rebuild

霍洛维茨称AI基础设施需要彻底重构 当前AI硬件和基础模型存在根本性低效，需要从底层彻底重建整个技术栈，以支持不可预测的推理工作负载。 a16z联合创始人本·霍洛维茨（Ben Horowitz）与主持人詹姆斯·诺曼（James Norman）展开对话，回顾了他从网景（Netscape）、Loudcloud、Opsware到风险投资这一创业历程中历经磨难汲取的教训。霍洛维茨阐述了他的核心评估框架——支持那些拥有世界级、突出的单项极致优势而非毫无缺点的创始人，并详述了a16z如何围绕对创始人的同理心、包容性人才发掘以及高杠杆运营网络进行构建。转向现代技术领域，他们对比了互联网泡沫时代与当前的AI浪潮，探讨了代码智能体如何对传统SaaS企业构成威胁、为何算力平权改变了初创企业的护城河，以及为何现有AI基础设施必须从底层彻底重构。 不同于增加人手会导致边际收益递减的传统软件开发，资本和大规模计算集群如今让AI竞争对手能够以创纪录的速度，通过强力硬拼与市场领导者达到技术平权。 (40:00) 企业级SaaS的估值倍数面临结构性压力，因为开发者正利用Claude和Cursor等AI编程工具，在数天内用自研的内部软件替代价值数千美元的订阅服务。 (43:08) 本·霍洛维茨认为当前的AI技术栈存在根本性缺陷，必须彻底从头重建，他指出源于视频游戏的GPU、不可预测的长达数天的推理工作负载，以及不可持续的训练数据需求都是主要问题。 (48:35) 根据是否毫无弱点来评估创始人是一个关键的风投陷阱；具有变革意义的公司是由某些单项具有世界级突出优势的创始人打造的，即便他们存在明显的运营盲点。 (20:00)

Ben Horowitz认为，卓越的创业者并非没有弱点，而是具备世界级的单项特长与极强的心理抗压能力；在此基础上，风险投资机构应转变为全方位支持创始人的赋能平台，并在AI重塑计算范式、颠覆传统SaaS的产业周期中重新定义技术壁垒与底层基础设施。 创始人特质与极端压力考验 (12:35) 评估创业者时不应盲目寻找没有短板的平衡型人才，而应考察其是否在关键领域具备世界顶级的能力。所有颠覆性企业的建立，都依赖于创始人极具突出的单项优势，即便他们在财务或营销等常规领域存在明显盲点也是可以接受的。在实际评估中，创意往往被视作了解创始人的诊断镜头，投资团队通过创始人构想创意的经历和思考深度来反推其个人特质。过分苛求消灭缺陷，常常会导致投资机构与极具天资的顶尖创业者失之交臂。 创业历程呈现剧烈的波动画卷，任何长青企业都会遭遇创始人面临极限考验的“WFIO（濒临崩溃）”绝境时刻。企业之所以失败，许多并非源于商业逻辑漏洞，而是创始人在底层重压下心理防线瓦解。极度高压往往会剧烈改变人的生理与心理反应，因此创业者的临场抗压韧性至关重要；正如Nas在处于低谷时以一曲《Ether》吹响反击号角一样，绝地求生的反叛意志是撑过企业存亡关口不可或缺的情感动力。 重塑风投机制与拓宽人才评价标准 (15:20) 传统风投机构主要围绕有限合伙人（LP）回报展开，擅长挑选项目却缺乏协助年轻CEO运营实战的组织结构。针对这一痛点，Andreessen Horowitz（a16z）放弃了部分短期管理费分配，将资本重投到投后服务平台中，自建企业级销售渠道、高管招募网络、政策战略支持体系以及促进初创团队融入产业生态的CEO聚会，致力于打磨一款专门服务于企业家的风投产品。 在组织招聘与项目发掘上，硅谷长期存在以狭隘刻板印象筛选人才的“人才盲区”，惯于招揽与面试官技术背景和文化特征重合的候选人。通过介入Angela Benton创立的NewME加速器等生态，a16z打通了连接非洲裔科技人才的独特网络，走出了Chris Lyons（历经约8次晋升成长为总合伙人）等优秀代表。团队应当拓宽人才的筛选维度而非降低任职门槛，精准围绕岗位真正所需的稀缺特长（如人际网络构建能力或助人特质）开展考核，进而将原本被系统性忽视的群体转化为机构的差异化竞争优势。 政企沟通受阻与产业政策博弈 (22:45) 针对外界对a16z介入政治竞选政策所产生的质疑，Horowitz说明这主要是应对监管政策阻力的务实策略。在拜登执政期间，加密货币领域遭遇了绕过法定程序的非正式监管，包括企业被切断传统银行服务接入以及收到法律定义不明的威尔斯通知（Wells notices）；与此同时，针对人工智能的一项行政命令提案甚至拟规定出售GPU必须预先获得政府批准，这对于芯片级计算生态构成了巨大威胁。 在关键行业亟待明确规则的背景下，包括谷歌的Sundar Pichai、礼来制药的Dave Ricks以及苹果的Tim Cook在内的多位企业高管，在其四年任期内均未获得直接同拜登会面的机会。由于寻求常规制度沟通无果，a16z选择直接介入大选政策的对话以保护底层创新环境。除了在科技政策层面与唐纳德·特朗普建立接触，Horowitz亦曾同卡玛拉·哈里斯进行政策沟通并为其撰写科技政策提案草稿，认为大型风投有责任为新兴科技企业的生存争取合理合规的外部空间。 算力平权冲击与新型AI护城河 (40:00) 虽然互联网泡沫时期全球仅有5000万至6000万拨号上网用户且基建高昂，但当前AI周期背靠数十亿联网人群、高速宽带以及年均下降约100倍的推理成本，因此并非单纯的概念炒作。然而，AI技术的突破重构了传统的研发壁垒：过去单纯依靠堆砌资金和招募上千名工程师无法快速赶超行业先驱，但在当前范式下，只要手握充沛资本和大规模数据集，企业便能通过采购庞大GPU集群迅速在短时间内追平前沿水准，xAI快速缩小与OpenAI的技术差距即是例证。 在资金能快速买来技术对等的环境下，AI企业的护城河转向了开发者嵌入深度与真实数据反馈回路。以ElevenLabs为例，通过嵌入开发者底层架构并服务于Netflix等企业级客户，其模型得以针对多语言的笑声、语调与节奏等长尾边缘场景不断迭代，形成了超越一般基础模型的专属数据壁垒。与此同时，Claude Code等AI编程工具的成熟促使企业用户能够以极低成本自建定制化CRM等工具，从而取消HubSpot等传统软件订阅，使得缺乏核心数据资产与深度定制能力的传统SaaS企业估值倍数持续承受下行压力。 Horowitz预测，在未来8至12个月内，现有的AI基础设施将经历颠覆性重构。当前基于游戏渲染改造的低良品率高功耗GPU、不可预测的推理运行时长，以及必须耗尽全网公开数据的低效基础模型架构，都属于过渡性质的错误栈，全栈层面的技术推倒重来将给新一代底层建设者带来决定性机遇。

## Background 本期内容由 Ben Horowitz 与 James Norman 于 2026 年 10 月 10 日在播客平台发布，为《Vintage x Vintage》播客的首发集。节目简介显示，a16z 联合创始人兼普通合伙人 Ben Horowitz 与主持人 James Norman 展开深入对话，探讨涵盖创业哲学、风险投资生态、嘻哈音乐文化、当前超越互联网规模的 AI 变革周期，以及底层 AI 基础设施可能面临的彻底重构。 ## TL;DR Ben Horowitz 回顾了自己从工程师成长为初任 CEO、带领 Opsware 绝境重生的创业历程，阐述了优秀创业者贵在拥有极致长板而非没有短板的核心投资哲学。两人深入探讨了 a16z 颠覆传统风投、以赋能创始人为核心构建人才网络的过程，以及破除硅谷刻板招聘与拓宽人才识别门槛的必要性。针对当下狂飙突进的 AI 周期，Ben 对比了 2000 年互联网泡沫的历史差异，指出 AI 门槛降低正彻底瓦解传统 SaaS 护城河，并断言现有以游戏 GPU 和暴力吞噬全网数据为基础的脆弱 AI 技术栈，在未来必将自下而上迎来全方位的底座重构。 ## Atmosphere & style 整场对话沉浸在醇厚的年代感与硬核科技洞察之中，既有开瓶加州名庄葡萄酒的闲适叙旧，又有触碰商业与技术本质的锐利思辨。Ben Horowitz 展现出硅谷老兵历经死局后的从容韧性，将街头嘻哈的野蛮生命力与公司生死战场的生存哲学无缝交织。两人的互动松弛而默契，在荒诞传奇的行业轶事、政治博弈的生存现实与前沿计算架构的深度推演之间自如切换，营造出一种既富文化底蕴又充满硬派实战哲学的深刻氛围。 ## Bulletin - `00:00` Ben Horowitz 认为伟大的创业者并非毫无短板，投资人应关注其核心强项是否达到世界顶尖水平，而非纠结于缺陷。 - `35:00` 本轮 AI 变革的影响力远超当年的互联网泡沫时期，Token 推理成本正以每年约 100 倍的速度快速下降，硬件需求具备真实的物理支撑。 - `43:10` AI 编程工具大幅降低了软件开发门槛，导致传统 SaaS 估值体系承压，企业和个人用户已具备根据特定需求自行编写替代软件的能力。 - `47:43` 当前的 AI 底层技术栈在底层芯片、任务调度系统以及依赖全网数据的暴力训练机制上存在根本缺陷，未来 8 到 12 个月及更长周期内将面临自下而上的全栈重构。 ## Key takeaways - `00:00` 评判伟大创业者的核心标准在于其长板是否达到世界顶级水准，而不是看他是否存在财务或管理等维度的技能缺陷。 - `10:00` 任何成功的创业公司都会经历生死存亡的“彻底完蛋”时刻，能够在极限压力下保持稳定发挥是优秀创始人最核心的特质。 - `12:30` Nas 在职业生涯谷底绝地反击写出的《Ether》所展现的心态，是创业者在遭遇全盘唱衰时必须具备的复仇与求生意志。 - `15:00` a16z 创立的核心逻辑是将风投机构定位为服务创业者的产品与顶级网络平台，将管理费重新投资于招聘、政策与战略赋能体系。 - `21:40` 拜登政府在加密货币领域的法外去银行化举措及对 GPU 销售行政审批的设想对行业构成威胁，迫使 a16z 下场展开跨党派政治博弈以保障科技企业生存空间。 - `30:00` 人才选拔的哲学应当是“拓宽门槛”而非“降低门槛”，打破基于刻板印象的自恋式招聘，才能真正识别出不符合传统画像的世界级人才。 - `33:14` 生成式 AI 将软件工程从核心壁垒降级为底层通用工具，纯粹的创造力与特定行业洞察成为新一代构建者的关键竞争力。 - `35:00` 与 2000 年互联网泡沫中用户少、成本高的资本狂热不同，当前 AI 周期伴随着营收暴增、Token 成本骤降及真实的物理算力消耗，不属于系统性泡沫。 - `40:00` 在 AI 周期中巨额资本与 GPU 采购可以迅速抹平纯算力代差，初创公司的真正护城河转向了深度的开发者生态、专有数据反馈闭环以及对真实场景的微调。 - `43:10` 随着 AI 个人化开发工具的普及，中小型场景下的专属软件开发门槛骤降，传统标准化 SaaS 面临订阅流失与估值乘数被严重压缩的危机。 - `45:26` 投资人留下的最核心遗产应当是将投资文化推向帮助人们构建伟大公司，并对早期激励过行业的先驱者予以物质与精神回馈。 - `47:43` 现有的 AI 基础设施在应对长周期推理负载、高功耗游戏架构 GPU 以及暴力吞噬全网数据的训练机制上面临极限，整个底层硬件与系统面临重写。 ## Detailed walkthrough ### `00:00` 伟大创业者的特质：不在于没有弱点，而在于拥有极致强项 节目开篇以一段关于创业者核心本质的精辟论断切入。在 a16z 联合创始人 Ben Horowitz 看来，评判一位创业者是否伟大，关键在于其强项是否达到了世界级水准，而不是看他是否存在短板。 > “Every great entrepreneur has got something horribly wrong with them, but that's okay. Are they truly great at what they're great at?” —— Ben Horowitz Ben Horowitz 指出，几乎每一位伟大的创业者身上都有极其严重的缺陷——比如有些人可能对财务一窍不通。但正是这种在某一维度的极致长板，让其他所有缺点都变得次要，这也是真正造就伟大公司与伟大创始人的核心要素。 本期节目为《Vintage x Vintage》播客的首发集，主持人 James Norman 与 Ben Horowitz 展开深入对话。内容不仅涵盖 Ben 从工程师成长为初任 CEO 的心路历程、创立 a16z 背后的理念支撑，还延伸至当前的 AI 浪潮（Ben 认为这一轮浪潮的影响力甚至超越了互联网）、软件开发门槛骤降对创业生态的重塑，以及两人共同的精神纽带——嘻哈音乐文化。 --- ### `00:44` Vintage x Vintage：Continuum 葡萄酒与 Opsware 的浴火重生 主持人 James Norman 借由节目名称“Vintage x Vintage”（年份与年代），拿出一款 2017 年份的加州著名葡萄酒 Continuum（傲云），串联起硅谷传奇的创业历程。 ```mermaid flowchart TD accTitle: Robert Mondavi 到 Continuum 的历程与 Opsware 创业轨迹对照 accDescr: 展示 Mondavi 家族因危机转折创立 Continuum，与 Ben Horowitz 带领 Opsware 绝地反击的平行对照。 subgraph WinePath["葡萄酒家族传承：Continuum"] Mondavi["Robert Mondavi 确立美酒基准<br/>(1993 葡萄根瘤蚜危机转折)"] Crisis["华尔街资本介入与品牌稀释<br/>(2004 家族重组)"] Continuum["幼子创立 Continuum<br/>历经困境沉淀出顶级酒庄"] Mondavi --> Crisis --> Continuum end subgraph StartupPath["硅谷技术突围：Opsware"] Loudcloud["Loudcloud 奠定云基础设施雏形<br/>(遭遇 2000 年互联网泡沫破裂)"] Pivot["被迫卖掉核心业务转型<br/>(股价跌入谷底，市场唱衰)"] Exit["绝境求生打磨出 Opsware<br/>16.5 亿美元被惠普收购"] Loudcloud --> Pivot --> Exit end ``` James 讲述了 Robert Mondavi 家族的过往： - 1993 年前后，加州纳帕谷遭遇严重的葡萄根瘤蚜病害（phylloxera），酒庄在巨大资金压力下被迫上市。 - 随后华尔街资本入局，对品牌过度开发，最终家族在 2004 年分崩离析。 - 家族幼子汲取了父辈积淀的酿造精髓与惨痛教训，在逆境中另起炉灶创立了 Continuum，打造出世界顶级佳酿。 James 指出，Ben Horowitz 的创业历程与这款酒的诞生如出一辙： - Ben 创立 Loudcloud，开启了云基础设施的先河。 - 遭遇互联网泡沫破裂时，不得不通过剥离并出售核心业务断臂求生。 - 公司上市后股价暴跌至几毛钱，几乎被整个华尔街判了死刑。 - 凭借极强的韧性与团队运作底蕴，他们逆风翻盘将 Opsware 打造成企业级运维软件巨头，最终以 16.5 亿美元的高价出售给惠普（HP）。 > “And so this was born out of struggle. And similar to my guest today, did the same in Silicon Valley.” —— James Norman James 将这款酒誉为装在瓶子里的《创业维艰》（*The Hard Thing About Hard Things*），向在极度逆境中生存下来的创业韧性致敬。 --- ### `04:08` 2017 年 CEO 烤肉派对与 Kanye West 的深夜录音室奇遇 两人的回忆随后被 2017 年份的葡萄酒拉回了一段极具戏剧色彩的真实往事：当年在 Ben 家中举办的 a16z 年度“CEO 烤肉派对”（CEO Barbecue）。 半夜时分，Ben 的手机突然响起，来电显示为未知号码。Ben 接起后发现对方正是 Kanye West（Ye）。Kanye 在电话里提出想来参加隔天的烤肉聚会，并陆续询问是否能带 Kim Kardashian 和 Kris Jenner 一同前来，Ben 均爽快答应。然而，Kanye 紧接着提出一个突兀的要求：“我想在你的录音室里录音。”Ben 当即愣住并明确告知：“我家里根本没有录音室。”Kanye 只是随口应了声“好的”，便挂了电话。 第二天清晨 6 点半，电话再次将 Ben 吵醒。对方自称 Noah，是 Kanye 的随行录音工程师，直接询问录音室的设备规格。Ben 只得再次解释自己根本没有录音设备。工程师顿时陷入恐慌：“你说没有录音室是什么意思？Kanye 以为这里有录音室，这下该怎么办？” 情急之下，Ben 紧急联络团队成员寻找救场方案。最终他们找到了 James（当时外号为 Howie）。接到电话时，James 正躺在家里看 Netflix，被问到是否有一套便携式移动录音设备后，James 立刻开车载着整套设备疾驰到 Ben 家中，在一间空房间内临时搭建了一座专业录音棚。 当天下午，Kanye 如约而至，还在现场为大家播放了正在制作的经典专辑《*The Life of Pablo*》。这场原本寻常的烤肉派对，不仅汇聚了各路科技创始人与嘻哈名人的激烈摩擦与恩怨互动，也成为硅谷与嘻哈跨界历史上的一段荒诞而真实的传奇插曲。 --- ### `06:33` 嘻哈精神、盲人说唱与商业洞察的源泉 两人的话题自然转向了对两人影响至深的嘻哈文化。Ben 回忆起自己最初接触嘻哈音乐的年代：大约 12 岁时，Sugarhill Gang 的《*Rapper's Delight*》问世，那是大多数非纽约本地孩子接触嘻哈的起点；而真正让他感受到颠覆性冲击的，则是 Run-D.M.C. 的经典曲目《*Hard Times*》。 真正让 Ben 将嘻哈音乐刻进生命底色的，是一段充满温情与痛苦的青春往事： - Ben 在纽约念大学期间，儿时同甘共苦的挚友 Seth 遭遇了一起极其离奇的枪击事故，面部中弹并不幸双目失明。 - 事故发生后，Seth 陷入重度抑郁与绝望之中，数月闭门不语，家人极其担心他会走上自杀绝路。 - 远在纽约的 Ben 试图拯救朋友，他意识到嘻哈音乐中那种“即使一无所有，也能大声炫耀自己球鞋是全场最帅”的野蛮生命力和绝对自信，或许正是 Seth 最需要的精神支撑。 - 当时才 18 岁的 Ben，每周六晚上放弃所有社交聚会，守在收音机前录制 DJ Red Alert 和 Chuck Chillout 的电台直播节目，随后把录音带一盘盘寄回给病床上的 Seth。 - 这一举动奇迹般地唤醒了 Seth，他终于开始重新开口说话，而且满嘴聊的都是嘻哈音乐和 DJ 节目。 大一暑假回乡后，Ben 更是与失明的 Seth 组成了一支名为“The Blind and DEF Crew”（双关词：失明与聋听 / 绝妙）的说唱组合。Ben 亲自为团队操刀写词，围绕盲人视角写出了许多戏谑自嘲却充满生命力的押韵段落。 Ben 总结道，早期的黄金年代嘻哈音乐本质上极具商业驱动思维（以 EPMD 的“Strictly Business”为代表）。他在日后撰写科技与管理类商业博客时，常常发觉许多深刻的商业哲学早已被写在那些经典说唱歌词里。他在每篇博文开头引用一句说唱歌词，不仅是一种风格标识，更是为了“注明灵感出处”： > “I started just putting the lyrics at the top of the post mostly to give credit where credit was due, because if I hear an idea and then I write an idea... you know it came from Rakim, or it came from Rick Ross, or whatever.” —— Ben Horowitz James 补充指出，由于所处年代（Vintage）和地域差异，早期西海岸年轻人缺乏流媒体分发渠道，很难像纽约本地或后来的孩子那样在幼年就浸润于纯正街头嘻哈中；而 Ben 则笑谈起好友、哈佛大学经济学家 Roland Fryer 近期在《华尔街日报》上发表的文章，两人由此拉开了关于嘻哈文化社会影响的更深层探讨。 ### `10:00` 从工程师到意外成为 CEO：早年经历与“彻底完蛋”时刻 Ben Horowitz 回顾了嘻哈音乐对其人生的深刻影响，提到曾有一项社会学研究，专门追踪各城市在嘻哈音乐尚未被主流电台播放时的发展轨迹，结果表明拥有嘻哈文化的城市表现明显更优，其注入的灵感与动力是无可比拟的。随后话题转向他的早期职业生涯。Ben 直言自己最初压根没想过要当创业者：“到我 25 岁的时候，我已经结婚了，还有了三个孩子。我当时只想挣钱养家糊口，维持生计（keep the lights on）。”在那个年代，创业远不像今天这样系统化，是极其罕见的选择。 Ben 最初作为一名年轻工程师加入了 Silicon Graphics（SGI），由 Ken Coleman 招募入职。在 SGI，他摸清了行业门道，学习如何做工程开发以及产品管理。随后，他跟随许多前 SGI 同事加入了 Netscape，并在那里结识了 Mark Andreessen。伴随 Netscape 的成功上市，Ben 分到了大约 100 万美元，这在当时让他觉得自己“巨富无比”。然而 Netscape 被收购后，Mark 突然提议合伙再做一家公司。Ben 当时毫无准备，便问谁来当 CEO，Mark 直接指着他说：“你来当。”Ben 只能硬着头皮接下挑战。 这段历程也让他领悟到了创业公司的典型起伏。他指出，创业不是平滑向上的，任何走上正轨的公司都会经历正弦波式的巨幅震荡，每个创始人都会迎来最黑暗的时刻： > “没有哪家公司是不经历那些低谷的，甚至不仅仅是低谷，而是我们常说的‘彻底完蛋了，全结束了’时刻（existential 'we're fucked, it's over' moment）。如果不经历那个时刻，你根本不可能把一家公司做起来。” 许多创业项目的夭折，正是因为创始人在面对这类生死存亡的压力时无法承受。 --- ### `12:30` 极限压力下的人性考核与嘻哈复仇金句 主持人 James Norman 谈到自己从事投资五年以来的体会：挑选公司容易，但看准人却需要持续演化，是一件极其困难的事。Ben 对此深表赞同，并将识人比作“谈恋爱相亲”——创始人在见投资人时总是竭尽所能展现最好的一面，但日常生活中的一个人与在极限压力下的人完全是两副面孔。当巨大的生存压力袭来，人的生理与心理机制会被完全重塑，谁也无法预料对方会作何反应。正因如此，Ben 高度评价了 Elon Musk： > “无论你喜欢他还是讨厌他，愿上帝保佑他，那家伙在压力之下的表现极其出色。完全没有任何压力能让他发挥失常，那正是创业者所需要的气质，这也是为什么他可能是我们这个时代最伟大的创业者。” 当被问及自己在跌落至最低谷、必须强撑着走下去时脑海中会浮现哪句嘻哈歌词，Ben 毫不犹豫地选择了 Nas 在巅峰 diss 曲目《Ether》中的著名歌词：“把我看作死人，僵硬且腐烂（Left me for dead, stiff and rotten）。”Ben 回忆道，那首歌诞生时 Nas 处于职业生涯的绝对谷底，而 Jay-Z 正如日中天，携手 Kanye West 制作的伴奏推出了极其凶狠的《Takeover》。在全世界都以为 Nas 已经彻底出局时，Nas 绝地反击写出了《Ether》。Ben 认为创业者需要的就是这股心气：在你以为我死透了的时候，我会带着致命的《Ether》卷土重来。James 补充说，Kendrick Lamar 在对抗 Drake 时推出的《Not Like Us》堪称《Ether 2.0》，Ben 则分析 Kendrick 当时虽处于内省与低谷期，但心态更接近于“看不过你把说唱带向歧途，我要走出房间收拾你”，与 Nas 当年绝境复仇的悲壮感略有不同。 --- ### `15:00` a16z 创立理念：颠覆传统 VC，做面向创始人的网络与赋能平台 两人将话题拉回科技圈，探讨 2008 至 2011 年间 Andreessen Horowitz（a16z）的诞生过程。当时 Ben 刚把公司（Opsware）卖掉，便迅速与 Mark Andreessen 着手组建一家前所未见的风投机构。传统风投机构人员精简，极为擅长为 LP（有限合伙人）赚钱，却极少真正懂如何运营公司。Ben 在亲身经历创业的艰难后发现，新手 CEO 极其孤独，他们不知道如何管理组织，需要能够随时致电有真才实学的人请教，需要有随时待命的招聘军团，在今天还需要能应对政策监管问题。因此，a16z 的底层逻辑彻底颠覆了行规： > “Andreessen Horowitz 背后的核心想法其实是：如果我们不是为 LP 做产品，而是为创业者打造一款更好的产品，那会怎样？” 为了践行这一理念，他们主动放弃了传统风投合伙人直接分掉的管理费，转而将管理费资金用来招募各路职能专家，搭建一个庞大的服务平台来辅助创业者成长。同时，a16z 将自身定位于一个顶级网络（network）。Ben 策划了著名的“CEO 烧烤派对（CEO Barbecue）”，让初出茅庐的创业新手在派对上与 Mark Zuckerberg、Larry Page 甚至 Kanye West 面对面交谈。当创始人发现自己能身处这群名流之中时，内心的自信与自我认知会发生质的飞跃。 ```mermaid flowchart TD accTitle: a16z 平台服务与赋能模式 accDescr: 对比传统 VC 仅专注 LP 投资回报与 a16z 以创始人为中心搭建投后赋能网络的机制差异。 subgraph TraditionalVC["传统风投模式"] LP1["有限合伙人 (LP)"] -->|"专注资金回报"| VC1["传统 VC 合伙人"] VC1 -->|"仅负责筛选与投钱"| Founder1["创始人（独自面对管理/招聘/政策等难题）"] end subgraph A16ZModel["a16z 平台赋能模式"] Fees["管理费资金再投资"] --> Platform["搭建投后服务平台"] Platform --> Recruiting["专属招聘团队"] Platform --> Policy["公共政策与战略支持"] Platform --> Network["顶级人脉网络 / CEO 烧烤派对"] Network --> Founder2["赋能创始人 / 树立顶尖信心"] end ``` 在审视硅谷的投资格局时，Ben 还敏锐地捕捉到了一个巨大的结构性空白：当时硅谷没有任何一家主流风险投资机构与黑人圈子有实质性连接。Ben 意识到这是一个不可多得的破局点，决心在硅谷建立连接黑人创业者与娱乐界领军人物的决定性网络。为此，他顺藤摸瓜找到了深藏于底层的孵化加速器 NewME（由 Angela Benton 创立并运营）。 Ben 亲自前往 NewME 进行演讲和挖掘，在那里结识并网罗了一大批顶尖人才，包括主持节目的 James Norman 本人、Chris Bennett、Chris Lyons、Nate Jones 以及 Fred Hudson 等。这一布局为 a16z 带来了巨大回报：机构成为了 Wonderschool（由 Chris Bennett 创立）的主要投资方，Chris Lyons 留任 a16z 成为重要合伙人，Nate 也曾在机构任职；此外，后来创立 645 Ventures 的 Aaron Holliday 和 Nnamdi Okike 也曾现身该圈子。Ben 在回顾这段经历时感叹，那笔看似微小的投入，为整个机构沉淀了难以估量的长期人才与网络价值。 ### `20:00` 挑选赢家：关注顶级长处而非补足短板 投资人究竟如何挑选赢家并管理投资中的情绪？Ben Horowitz 坦言，这其实非常微妙且充满挑战，因为归根结底起决定性作用的是“创业者本身”。如果有完美的创业者评级系统，让人在“优秀创业者”和“优秀想法”之间二选一，投资人毫无疑问每次都会选创业者。但现实的难点在于，创业者走进会议室时大家对其知之甚少，更无从预判他们在巨大压力下的真实表现。因此，创业者的“想法”反倒成了审视他们的一面镜子——他们经历了什么才形成这个想法？做了多少原创性的工作？想法本身的深度如何？投资人常犯的致命错误，要么是因为不喜欢某个想法而错失了伟大的创业者，要么是因为误以为某个想法很好而押注了平庸的人。 此外，投资机构内部极易陷入过度分析的陷阱。分析能力极强的专业人士往往能精确指出任何想法、公司或创业者身上的所有毛病，但 Ben 强调这根本不是投资的关键。针对这一点，Ben Horowitz 总结道：“What matters is are their strengths world class? So are they the best in the world at something? And then you overlook everything else.”（关键在于他们的优势是否达到了世界顶尖水平？他们是否在某件事上做到了全球最好？如果是，你就可以忽略其余的一切缺陷。） 以 Elon Musk 为例，他虽有性格瑕疵、常常惹怒身边的人，但他展现出的绝对能力之强，足以让任何投资人每次都坚定地下注。每一个优秀的创业者身上都会有极其致命的短板——或许不懂财务，或许欠缺营销技巧，但这完全没关系。一家伟大公司的诞生，靠的是创业者在核心长处上绝对碾压对手的卓越表现，而不是四平八稳的无缺陷状态。 --- ### `21:40` 游说困境与政治博弈：为科技企业争取生存空间 谈到支持特定政治候选人所引发的硅谷震荡与文化反弹时，Ben Horowitz 深入剖析了 a16z 卷入政治决策的底层逻辑。他强调机构一直试图保持无党派立场，但拜登政府在加密货币和人工智能两大前沿领域对创新生态构成了毁灭性打击： ```mermaid flowchart TD accTitle: Regulatory Pressure Flowchart accDescr: Flowchart illustrating Biden administration regulatory hurdles on crypto and AI leading to political engagement subgraph Regulation["政策打压 (Regulatory Pressure)"] Crypto["加密货币：法外排挤与去银行化"] AI["AI 产业：拟实施 GPU 销售行政审批"] end subgraph Deadlock["沟通断绝 (Communication Deadlock)"] Block["闭门不见：科技巨头皆无法会见总统"] end subgraph Strategy["应对策略 (Political Engagement)"] Endorsement["技术路线背书与政治接触"] Support["建立双边对话渠道，保障行业生存"] end Regulation --> Deadlock Deadlock --> Strategy ``` 在加密货币领域，监管部门没有立法规管，而是采取了法外手段：切断加密创业者的银行服务通道，发出缺乏清晰定义的 Wells 通知（虚假诉讼威胁）。机构多次呼吁出台明确法律监管，均遭到封锁。更严重的是沟通渠道彻底关闭，Ben Horowitz 四处联络试图争取与拜登会面，不仅被外界嘲讽“狂妄”，更发现无论是 Google 的 Sundar Pichai、礼来的 Dave Ricks 还是苹果的 Tim Cook，在拜登的整个四年任期内都没能获得会见机会。在 AI 领域，拜登政府的行政命令甚至企图要求出售每一张 GPU 都必须获得政府预先批准，这将直接让美国芯片与算力企业退出底层竞争。 面对生死存亡，a16z 决定必须与下任总统建立沟通渠道： - **特朗普阵营**：在科技政策上给予支持背书，当时其选情展现出压倒性优势。 - **哈里斯阵营**：哈里斯成为候选人后，鉴于过往私交（哈里斯曾多次拜访 Ben 家中并参加 CEO 烧烤聚会），Ben 亲自致电并为竞选团队撰写了科技政策立场文件，同时也提供了竞选捐款以维持对话渠道。 Ben Horowitz 直言参与政治必然招致各方愤怒（就像公众对 Nicki Minaj 或 Kanye West 的政治表态反应激烈一样），但他指出：“From my position, I have to make the world work for new tech companies like that. That is, you know, our responsibility as kind of the big VC.”（从我的位置来看，我必须让外部世界为新兴科技公司留出生存和发展的通道。作为一家大型风投机构，这是我们的责任。） --- ### `26:24` 科技圈的包容性演变：从嘻哈文化对比看破除刻板招聘 James Norman 提出，早期的硅谷虽然带有嬉皮士色彩，但在某种程度上是“表演性包容，本质上排外”；而如今 AI 将开发门槛彻底平民化，让没有 MIT 计算机背景的人也能成为顶级构建者。Ben Horowitz 纠正了这一历史视角：在 20 世纪 80 年代，科技行业既不光鲜也不性感，主流人群根本不愿涉足，行业内绝大多数是犹太人、华人与印度裔男性，女性极少。这种圈子的形成并非刻意排外，而是基于最朴素的自恋式招聘逻辑——“我知道自己擅长什么，我高度看重这项技能，我在面试里考这个，我招募和我一样的人”。 这就像嘻哈文化：白人涉足并非绝无可能，但必须达到极致天才的标准——就像 Eminem，必须强悍到无可挑剔才能打碎偏见立足，而平庸的白人根本没有机会；科技界也是同理，刻板印象会让外界预设戴眼镜、不擅言辞的技术人员才是好开发者。 为了彻底打破这种基于背景的自我复制，a16z 重新设计了人员招聘评估机制： - **引入多元视角审视岗位**：由不同背景的人共同审阅职位描述，发现那些传统管理者甚至不知道该如何考核的核心能力。 - **Chris Lyons 的经典案例**：该岗位本质是“维护人际关系”，但原招聘主管曾是投资银行家，便套用了极其死板的投行财务建模标准，导致 Chris Lyons 初试完全不及格。Ben 介入并明确指出这是关系型岗位，而 Chris 具备世界级的人际连接能力。以真实所需的正确标准录用后，Chris 在机构内凭借卓越表现连续晋升了大约八次。正是因为明确其核心优势，才没有让优秀的非裔人才被困在不相干的技能考评陷阱中。 ### `30:00` 拓宽选拔门槛：从“城市音乐”的偏见到识别真正的人才 谈到人才选拔与招聘机制，Ben Horowitz 指出，招聘的真正哲学应当是“拓宽门槛（widen the bar）”，而不是“降低门槛（lower the bar）”或盲目抬高门槛。拓宽门槛意味着拓宽视野，识别出候选人胜任工作所需的各项综合素质，哪怕这些特质不在传统的条条框框里。James Norman 联想到自己早期作为软件工程师的亲身经历：作为公司唯一的黑人工程师，他经常遭遇同事带着偏见的打量与冷嘲热讽，大家互不尊重且缺乏协作基础，团队甚至根本不明白为什么要招他进来。 为了说明硅谷在人才识别上的盲点，Ben Horowitz 分享了他与著名音乐经纪人 Steve Stoute 之间一段令人啼笑皆非的对话。Steve Stoute 曾致电 Ben Horowitz，提起自己早年担任 Sony Urban Music（索尼城市音乐部）总裁的经历： - **概念局限与地理歧视**：Steve Stoute 指出，唱片公司为了显得“不种族主义”，把黑人音乐冠以“城市音乐（Urban Music）”的代名词，但这直接导致他无法在乡村或偏远地区进行宣发，仿佛黑人全住在城市里一样。 - **自我设限的商业定义**：Steve 当时愤怒地反问：“我管辖的歌手包括迈克尔·杰克逊（Michael Jackson），难道白人就不听迈克尔·杰克逊吗？那根本不是什么他妈的黑人音乐，那就是音乐本身！”这种刻板分类没有让市场变大，反而将其人为压缩了。 Ben Horowitz 直言，硅谷过去的人力资源管理同样充斥着这种“城市 HR”的短视逻辑。人才就是人才，没有所谓的“黑人人才”，关键在于管理者能否听到那首“动人的音乐”。当行业高喊所谓的“精英唯才制（meritocracy）”时，往往只是在照镜子选人——“我就招跟我一模一样的人”。Ben 强调，a16z 能在风险投资领域占据巨大优势，恰恰是因为很多传统机构患有“人才盲视（talent blindness）”，看不见那些不符合传统画像的优秀人才。 以 a16z 负责市场营销的资深合伙人 Margit Wennmachers 为例，她手下的团队长期清一色全是女性。Ben 曾半开玩笑地问她为什么没有男性能通过她的面试标准，Margit 的回答只有简单的两个词：“乐于助人（helpfulness）”。在风险投资行业，每一位创始人都希望投资人真正能提供帮助，但这极少被写入传统的招聘标准。a16z 将“提供帮助”确立为核心评价维度后，没有设立任何硬性的性别配额，女性比例却自然上升至 54% 左右。Ben Horowitz 总结道：“If you can see more talent, you're going to win.”（如果你能看到更多的人才，你就会赢。） --- ### `33:14` AI 降低创造门槛：从“技术壁垒”到“创意即产品” 随着话题切入当下的技术周期，Ben Horowitz 与 James Norman 对生成式人工智能带来的爆发力表现出极度的兴奋。James Norman 坦言自己现在每天不上手写几行代码、构建点东西就浑身难受，甚至常常捧着手机用 Claude Code 熬夜开发。他提到身边一位名叫 DeSean 的朋友，过去从不写代码，但在得知如今完全无需雇佣工程团队就能做产品后，瞬间迷上了 AI，整天倒腾智能体（Agent），将原本四人团队缩减至两人，整个人像个“AI 瘾君子（AI fiend）”。两人笑称，照这个趋势发展下去，未来甚至需要成立“AI 戒毒所”，因为晚上躺在床上如果不大喊“快去睡觉”，根本无法停下手中的开发。 这种工具的革命性在于彻底拓宽了创业可能性的天花板。过去，一个来自石油化工行业的传统产业专家，哪怕对行业痛点与工作流了如指掌，只要他从未做过软件、从未涉足科技公司，投资人几乎不可能相信他能把一家软件公司做起来。而如今，从一个构想到一家运转正常的初创企业，整个孵化周期被急剧压缩，以往横亘在非技术人才面前的人力与技术壁垒荡然无存。 正如 Ben Horowitz 所言，人们不再需要先“进入科技行业”才能改变世界，科技已退居为底层工具，真正的核心竞争力变成了纯粹的创造力：“Creativity is the thing.”（创意才是核心所在。）只要懂得操作 AI 工具，脑海中的任何构想都可以立刻显化并落地。 --- ### `35:00` 超越互联网的技术巨浪与理性看待 AI 泡沫 回顾早期的互联网萌芽时代，两人感慨万千——那是充斥着 BBS 论坛、Prodigy、CompuServe、AOL 寄来的海量免费试用软盘、甚至从麦片盒里翻出软盘来刻录存储的纯真年代。然而，尽管互联网塑造了今天的世界，Ben Horowitz 斩钉截铁地表示，当下的 AI 变革不仅超越了加密货币这类单一赛道，其规模更是远超当年的互联网，是真正的第四次工业革命： > “AI is a new computer, right? Like it's a—it's like the reinvention of the computer. Imagine the entire computer industry got invented. But instead of it developing over 80 years, it developed over two years. And then like that's AI.” > —— Ben Horowitz 从 Waymo 正在解决的交通事故致死问题，到 AI 在癌症攻克上的最新进展，AI 正在直接正面解决人类最棘手的宏观命题。针对市场上甚嚣尘上的“AI 泡沫论”，经历过 2000 年互联网泡沫破裂全过程的 Ben Horowitz 提供了深刻的历史对比： - **泡沫的内在机制**：泡沫往往诞生于三者的脱节——技术底层演进、大众实际采纳度、以及极度“躁郁（manic-depressive）”的投资人情绪。 - **2000 年互联网泡沫的成因**：当时资本盲目笃定互联网绝不会跌价，然而全网仅有五六千万用户，其中半数还在使用高延迟的拨号上网，开发成本极其高昂，市场规模根本无法消化海量商业构想，崩盘在所难免。 - **当前 AI 周期的本质差异**： - **用户与收入飞速增长**：OpenAI 在极短时间内营收规模从零暴增至近 200 亿美元。 - **推理成本断崖式下跌**：Token 的单价正以每年约 100 倍的速度下降。 - **真实且无限的硬件需求**：当前市场对英伟达（NVIDIA）算力芯片的渴求是绝对真实的物理需求，而非虚构的资本空转。 Ben Horowitz 澄清道，目前科技巨头争相免费提供模型或拼命追赶，并不意味着没有商业价值。虽然在周期的极早期，很多目前看似领先的创业公司可能会因为底层技术的剧烈迭代而在下一轮洗牌中覆灭，但这属于正常的风险投资规律，而非整个赛道的系统性泡沫。不可否认，未来可能出现类似 2001 年电信带宽泡沫那样的局部产能过剩，但就目前的技术演进与采纳速度而言，其底层驱动力毋庸置疑。 ```mermaid flowchart TD accTitle: 互联网泡沫与当前 AI 周期的驱动结构对比 accDescr: 展示2000年互联网泡沫与当前AI周期在用户基数、技术成本及资本需求上的核心差异。 subgraph InternetBubble ["2000年 互联网泡沫"] IB_Users["用户基数小<br/>(5000-6000万人 / 多为拨号上网)"] IB_Cost["开发与带宽成本极高"] IB_Sentiment["资本极度狂热，远超实际市场承载力"] IB_Users & IB_Cost & IB_Sentiment --> IB_Result["市场崩溃 (泡沫破裂)"] end subgraph AICycle ["当前 AI 周期"] AI_Adoption["用户采纳速度史无前例<br/>(OpenAI 营收爆发增长)"] AI_Cost["Token 成本每年下降 ~100倍"] AI_Hardware["算力需求近乎无限<br/>(NVIDIA 芯片供不应求)"] AI_Adoption & AI_Cost & AI_Hardware --> AI_Result["技术与产品真正落地<br/>(存在局部迭代淘汰，非系统性泡沫)"] end ``` ### `40:00` 瓶颈转移与算力追赶：为什么资本与 GPU 能打破壁垒 Ben Horowitz 与 James Norman 探讨了技术周期的演进逻辑。Ben 回忆起互联网发展初期，曾有人抛出“带宽过多”的荒谬论调，但事实证明，那只是因为系统瓶颈发生了转移：当带宽不再是阻碍时，服务器性能不足的问题就暴露了出来，无法足够快地输出数据以支持视频播放。在过去几代科技创业浪潮中，人们无法单纯通过“砸钱”来消灭竞争差距——如果一家公司落后了，即便明天立刻新招一千名工程师也无济于事。然而在当下的 AI 周期中，规则彻底改变了：只要拥有数据并买下足够多的 GPU，巨额资金就可以直接换取技术成果。Ben 举例指出，OpenAI 的年化收入达到了约 100 亿美元，而 Elon Musk 迅速创办了 xAI，并凭借巨大的算力投入在短短几年内就追平到了接近产品对等的水平。 ```mermaid flowchart TD accTitle: 技术演进中的瓶颈转移与算力突破 accDescr: 展示从互联网早期的服务器与带宽瓶颈，演化至当今 AI 周期通过数据和 GPU 快速弥补竞争差距的机制。 subgraph InternetEra["早期互联网时代"] BW["带宽瓶颈消除"] --> BottleneckShift["瓶颈转移至服务器性能"] BottleneckShift --> FailThrowMoney["无法仅靠资本与堆人解决差距"] end subgraph AIEra["当前 AI 时代"] CapAndGPU["巨额资本 + 大量 GPU 采购"] --> TrainScale["海量数据规模化训练"] TrainScale --> QuickCatchup["快速抹平与巨头的技术/产品代差"] end InternetEra -.->|"范式转变"| AIEra ``` 面对这一变革，James Norman 提出投资逻辑必须随之重构。许多投资人目睹了这种巨变，却依然沿用老套的评估标准来决定投资标的与创始人。Ben 强调，在 AI 时代，单纯的护城河很容易被海量算力冲垮，但防御壁垒会出现在意想不到的地方。他以 a16z 投资的语音合成与克隆公司 ElevenLabs 为例：在投资之初，外界普遍认为大模型巨头一旦下场做语音，这类初创公司将脆弱不堪。然而 ElevenLabs 凭借敏锐的产品设计构筑了真正的壁垒： - 他们尽早切入了开发者工具链（Developer Stack），为开发者打造了体验绝佳的产品，从而获得了庞大的原生用户群。 - 深度嵌入口语使用场景构建起实时的数据反馈闭环（Data Feedback Loop），利用真实的在途数据持续微调模型。 - 雕琢精细的产品细节，例如不仅能说多国语言，还能将英语演讲者的微颤、笑意和重音精准保留并映射到西班牙语等其他语言中，以此先后拿下了 Netflix 和 Televisa 等头部客户的大单。 Ben 总结道，即便对手拥有更多 GPU，也很难复制这种建立在高质量专有数据与场景微调之上的产品壁垒：““And that, yeah, you could throw more GPUs at yours, but you may not be able to do the things I do because I've got better data. I've got, you know, like I've got a model that's really tuned to the use cases that have value, as opposed to like every damn thing that everybody wants to do.”（面对这种情况，是的，你可以给你的模型砸更多 GPU，但你可能根本做不到我能做的事，因为我拥有更好的数据，我的模型针对那些真正有价值的用例进行了深度微调，而不是去硬凑每个人想搞的各种乱七八糟的杂事。）” --- ### `43:10` SaaS 的重估危机与个人化定制软件的兴起 随之而来的深层冲击是对传统应用层软件的估值颠覆。Ben 提到，如今全球的 SaaS 公司都在以极低的市盈率/市销率倍数交易，因为市场开始意识到：客户现在完全有能力自己重新编写软件。James 对此深有共鸣，并分享了自己的亲身经历：大约一年前他开始接触 Claude Code、Cursor 等 AI 编程工具，体验之后震撼地意识到“我再也不想肉眼看代码了”。第二天，他就果断退订了 HubSpot，并表示“这玩意儿太烂了，我要的是适合我自己的东西”，仅花了一周时间就完全用 AI 搭出了属于自己的专属 CRM；接着他又退订了 Affinity，自己做出了募资管理平台。 James 坦言，现在面对订阅服务时，心态发生了彻底的逆转：面对一个月费 400 美元的工具，他会反问自己“这真的值 400 块一个月吗？还是我花四天时间自己搓一个出来？”Ben 指出，虽然目前自己手搓工具的现象主要集中在工程师或产品导向的人群中，但未来这一门槛会降到极低——任何人在网页上觉得用着不顺手，都可以直接要求 AI 按照自己的习惯重新生成系统。虽然大型企业（如美航运行 SAP）因为复杂业务流和遗留系统数据难以在一夜之间被推翻，但在中小型与个人场景中，任何人都能够调用 Anthropic 等巨头拥有的 GPU 算力，在极短时间内复刻出符合特定需求的专用系统，SaaS 的防御边界正在被全面瓦解。 --- ### `45:26` 遗产、文化改造与 Hip-Hop 基金会 Paid in Full 当访谈进入尾声，James 拿出了标志性的收尾问题，询问当下一代风险投资人谈论 Ben Horowitz 时，他希望给这个行业留下怎样的烙印（Legacy）。Ben 直言自己并不常思考“遗产”这种宏大词汇，他引用了拳王迈克·泰森（Mike Tyson）对该问题的经典回答——人死后不过是一抔黄土和尘埃，谈遗产显得荒谬。但他确实怀揣一个愿景：““You know, for me, I hope that people feel like I moved the culture of investing to a better place, to one that's more about helping people build companies than just making money.”（对我来说，我希望人们觉得我把投资文化带向了一个更好的境地——一个更注重帮助人们打造公司、而不仅仅是为了赚钱的境地。）” 当 James 追问作为个人的成就感来自何处时，Ben 借用了 NFL 的橄榄球“教练树”（Coaching Tree）理论：最让他感到欣慰的，是 a16z 培养出了怎样的人才，以及他们在这里工作期间是否成为了更好的人。在商业之外的个人领域，Ben 最重视的项目是他与妻子创立的 Paid in Full 基金会。该基金会致力于为那些在嘻哈音乐史上做出了开创性贡献、却未能获得相应经济回报与养老保障的老一代先驱艺术家提供经济支持和荣誉。Ben 严肃地反思道：““If we forget the people who kind of sacrificed their lives to inspire us, then you get into a very nihilistic existence.”（如果我们遗忘了那些牺牲自己的人生来激励我们的人，我们的存在就会陷入极度虚无。）”外界过去充斥着大量空谈，抱怨电台和产业机制不公，却没有人实际行动；因此他决定停止抱怨，直接下场设立基金会，这带给他无与伦比的精神富足。 --- ### `47:43` AI 基础设施的代际缺陷与底座重构预言 在谈及未来 8 到 12 个月科技格局的走向时，Ben 作出了一个坚定的前瞻性判断：当前的整个 AI 底层基础设施不仅严重过时、脆弱不堪，而且在根本逻辑上就是“错的”。这与云计算发展初期的弯路如出一辙——早期大家天真地认为，只需把现有的软件直接搬上云端、放在网线另一头就能搞定云计算；但事实证明这完全行不通，整个技术堆栈最终从底层存储、网络交换、操作系统、虚拟化机制到容器技术全部彻底重写。 ```mermaid flowchart TD accTitle: AI 基础设施重构演进图 accDescr: 对比传统工作负载与 AI 工作负载的差异，并推导从底层芯片、系统到模型架构的全栈重构。 subgraph WorkloadMismatch["工作负载的根本失配"] DBQuery["传统数据库查询<br>(毫秒级 / 负载明确)"] AIQuery["AI 推理与生成<br>(可能持续运行数天至一周)"] end subgraph HardwareLimits["硬件与数据瓶颈"] GPU["NVIDIA GPU 架构<br>(高功耗/低良率/起初为游戏设计)"] DataSuck["模型预训练吞吐<br>(必须吸干全网公开数据)"] end subgraph FutureRebuild["未来 8-12 个月及更远的重构方向"] NewSys["专为长程 AI 任务设计的新型底座系统"] NewChips["高良率、高能效专用 AI 原生硬件"] NewModels["跳出暴力堆叠全网数据的新型模型结构"] end AIQuery --> NewSys GPU --> NewChips DataSuck --> NewModels ``` Ben 深入剖析了当下的几大核心症结： - **工作负载特性的本质差异**：传统的数据库查询任务具有明确可控的时延与资源边界；而一个 AI 推理任务的运行周期可能是不可预测的，甚至连续运行长达一周。目前根本没有一套成熟的底层系统能够精确调度和管理这种超长、超重的工作负载。 - **底层算力硬件的先天气质不符**：英伟达固然伟大，但其 GPU 本质上是高功耗、低良率、最初为了运行视频游戏而设计的芯片。随着行业对模型运行机理的理解加深，继续依赖游戏卡架构绝非长久之计。 - **模型训练机制的暴力极限**：当前前沿模型需要强行抓取并吞噬互联网上的全部数据才能完成训练，这种依赖暴力堆叠的路径已经逼近物理和数据极限。 基于这些系统性缺陷，Ben 预言，从模型架构、调度系统到最底层的硅片硬件，整个 AI 技术栈未来都将自下而上经历彻底重构，而这正是未来孕育巨大创业与投资机会的沃土。在畅想技术底座重塑的激辩中，两人互致感谢，结束了 Vintage x Vintage 首期节目的录制。 ## Notable quotes > `00:00` Every great entrepreneur has got something horribly wrong with them, but that's okay. Are they truly great at what they're great at? — Ben Horowitz > `10:00` 没有哪家公司是不经历那些低谷的，甚至不仅仅是低谷，而是我们常说的‘彻底完蛋了，全结束了’时刻（existential 'we're fucked, it's over' moment）。如果不经历那个时刻，你根本不可能把一家公司做起来。 — Ben Horowitz > `15:00` Andreessen Horowitz 背后的核心想法其实是：如果我们不是为 LP 做产品，而是为创业者打造一款更好的产品，那会怎样？ — Ben Horowitz > `35:00` AI is a new computer, right? Like it's a—it's like the reinvention of the computer. Imagine the entire computer industry got invented. But instead of it developing over 80 years, it developed over two years. And then like that's AI. — Ben Horowitz > `45:26` You know, for me, I hope that people feel like I moved the culture of investing to a better place, to one that's more about helping people build companies than just making money. — Ben Horowitz > `45:26` If we forget the people who kind of sacrificed their lives to inspire us, then you get into a very nihilistic existence. — Ben Horowitz ## Mentioned Ben Horowitz, James Norman, Andreessen Horowitz (a16z), Mark Andreessen, Elon Musk, Kanye West, Kim Kardashian, Kris Jenner, Nas, Jay-Z, Kendrick Lamar, Drake, Eminem, Steve Stoute, Margit Wennmachers, Chris Lyons, Chris Bennett, Angela Benton, Ken Coleman, Roland Fryer, Joe Biden, Donald Trump, Kamala Harris, Sundar Pichai, Tim Cook, Dave Ricks, Silicon Graphics (SGI), Netscape, Loudcloud, Opsware, Hewlett-Packard (HP), Robert Mondavi, Continuum, OpenAI, xAI, NVIDIA, ElevenLabs, HubSpot, Affinity, Claude Code, Cursor, Anthropic, Waymo, Netflix, Televisa, The Life of Pablo, Ether, Takeover, Not Like Us, The Hard Thing About Hard Things, Paid in Full 基金会, NewME, Wonderschool, 645 Ventures ## Video description a16z co-founder and General Partner Ben Horowitz joins James Norman on the inaugural episode of Vintage x Vintage for a conversation spanning entrepreneurship, investing, hip-hop, and what changes when AI makes building dramatically easier. Ben reflects on going from Netscape to becoming a first-time CEO, why every founder eventually hits a “we’re fucked, it’s over” moment, and what he learned from building a16z around the needs of entrepreneurs. He explains why great investors should look for world-class strengths rather than an absence of weaknesses, and why expanding the ways we recognize talent can raise the bar rather than lower it. They also look at the current AI cycle, why Ben believes it is bigger than the internet, how AI changes who can become a builder, and why the infrastructure beneath today’s models may need to be rebuilt from the ground up. [Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/e56d3db2-edb1-4f5a-a918-c165823f4160/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=e56d3db2-edb1-4f5a-a918-c165823f4160&feed=JGE3yC0V)

[0:00] Ben Horowitz: Every great entrepreneur has got something horribly wrong with them, but that's okay. Are they truly great at what they're great at? [10:00] Ben Horowitz: 没有哪家公司是不经历那些低谷的，甚至不仅仅是低谷，而是我们常说的‘彻底完蛋了，全结束了’时刻（existential 'we're fucked, it's over' moment）。如果不经历那个时刻，你根本不可能把一家公司做起来。 [15:00] Ben Horowitz: Andreessen Horowitz 背后的核心想法其实是：如果我们不是为 LP 做产品，而是为创业者打造一款更好的产品，那会怎样？ [35:00] Ben Horowitz: AI is a new computer, right? Like it's a—it's like the reinvention of the computer. Imagine the entire computer industry got invented. But instead of it developing over 80 years, it developed over two years. And then like that's AI. [45:26] Ben Horowitz: You know, for me, I hope that people feel like I moved the culture of investing to a better place, to one that's more about helping people build companies than just making money. [45:26] Ben Horowitz: If we forget the people who kind of sacrificed their lives to inspire us, then you get into a very nihilistic existence.

[20:00] Ben Horowitz: Every great entrepreneur has got something, like, horribly wrong with them. But that's okay, you know, it's what are they truly great at. [30:00] Ben Horowitz: If you can see more talent, you're going to win. [37:12] Ben Horowitz: Because we're so early in the cycle, there are companies that are in a good position now that there will be a technological change and they'll get wiped out. But that's not really a bubble to me. That's just like investing. [45:20] Ben Horowitz: If we forget the people who kind of sacrifice their lives to inspire us, then you get into a very nihilistic existence.

[0:00] I think it was midnight and my phone rings. Unknown caller. So I'm like, oh, it's Kanye. I go, yeah, what's up? And he goes, hey, Ben, I heard you're having a barbecue tomorrow. And I was like, yeah, yeah, I am. I want to record in your studio. And I'm like, yeah, I don't have a studio. Are their strengths world class? So are they the best in the world at something? And then you overlook everything else. That's what makes a company great. And that's what makes an entrepreneur Wow, they know nothing about finance. There's so many things that are wrong with every entrepreneur. Every great entrepreneur has got something horribly wrong with them, but that's okay. Are they truly great at what they're great at? What separates a great entrepreneur from everyone else? For Ben Horowitz, it isn't the absence of weaknesses. It's having a strength so exceptional that everything else becomes secondary. Today's episode comes from our recommended listening series. where we feature standout conversations from people across our broader network. This one comes from Vintage and Vintage, where host James Norman sits down with A16Z co-founder Ben Horwitz for the show's inaugural episode. Ben shares what he learned going from engineer to first-time CEO, how those experiences shake the founding of A16Z, and why evaluating founders means looking for world-class strengths rather than trying to find someone who checks every box. this moment as bigger than the internet. As the cost and difficulty of building software collapses, creativity becomes increasingly important, opening the door for people with deep knowledge of a problem to turn that knowledge into a product. And throughout the conversation, James and Ben returned to a shared influence, hip-hop, from the lessons Ben took from Nas at his lowest point to why he created Pated Full to recognize the pioneers who shaped the culture. the first episode of Vintage and Vintage. This is the first one? The first one. All right. Well, that's good. We're getting off with a bang. Right. You know, step into the game. But to start the conversation is Vintage and Vintage. You are a certain vintage of entrepreneur. We also have a certain vintage. We have a certain vintage of wine as well. So like, And all the wines are connected to like The relationships, this is a continuum, right? But it's actually like Mondavi. So Robert Mondavi, you know, they kind of set the tone for American wine. But by 93, though, there was like this like virus vineyard situation in Napa. Yeah. They had to go public. You're under duress. Like, oh, we got to get this money. We got to keep going. Didn't turn out that well because Wall Street directors came in and abused the name. And now you have the Robert Rondi wine that people would know today. But it kind of fell apart by 04. And the family had to pick up the pieces. But the youngest son came along. and decided to make Continuum. And he wanted to shoot for the stars, take all the learnings he had, the inner workings of how they made great wine, and make this. And so this was born out of struggle. And similar to my guest today, did the same in Silicon Valley. Set the tone for cloud computing. Had to sell the heart of the company to keep the thing going. Even after going public, stock dropped as a result and had to crawl out of that. But like, took the knowledge... one had in the team and the inner workings of how the thing was managed, Opsware, and, you know, clawed your way back from people's like, they're counting them out. Nah, they're not going to make it. It's a .com boom bust. It's like, you know, got the $1.65 billion exit. And so Continuum is special and legendary, my guests, special and legendary, but it's great because of the struggle. And so this is the hard thing about hard things in the bottle. All right. And this is my guest Ben Horowitz. All right. That's a very nice introduction. Thank you, Ali. Or so I call you James. Most people know you're James. I'm talking from the new me days. Right. So... Actually, hold up. Actually, take a look at the year in Hawaii before that. 2017. Okay. Yeah. We did something pretty funny and crazy. We did something pretty – Was that the CO barbecue with Kanye? It was. Oh, boy. It was a day in the studio with Ye. Yeah, that was wild. So – For you listeners, what happened was we were having the CEO barbecue, which is a big event at my house and so forth. And... It's like... I think it was midnight and my phone rings. And, you know, I heard, I guess I didn't turn off the ringer and I look at it, unknown caller. So I'm like, oh, it's Kanye. And so I answered the phone and I go, yay, what's up? And he goes, hey, Ben, I heard you're having a barbecue tomorrow. And I was like, yeah, yeah, I am. He's like, can I come? I was like. Sure, you can come. He goes, oh, can Kim come? And I'm like, yeah, Kim can come. Of course, you can come. Well, can Chris come? Yeah, Chris can come. And he goes, okay, great. And I want to record in your studio. And I'm like, Yeah I don't have a studio And he goes Okay And so like Okay Then [5:00] like, I wake up, 6.30 in the morning, I get a call. Hey, Ben, this is Noah. I'm Kanye's engineer. He's like, I'm just calling about the specs on your studio. And I'm like, I don't have a studio, Noah. He's like, what do you mean you don't have a studio? What are we going to do? Kanye's expecting a studio. We've got to have a studio. I'm like, oh, shit. So I call Chris and Nate. I'm like, who do we know who has a studio? And they're like, Howie. And I'm like, okay, well, let's call him. Because Kanye's coming over. We need a studio. And so we call you. We're like, okay. what are you doing how he's like watching netflix and i'm like okay i was like is that studio you have portable and he's like yeah yeah it's portable and uh so long story short uh you drove the studio down to my house and uh we set it up in the room and then a lot more crazy stuff happened right uh yay was playing us the pablo pablo album yeah we went through a level we went through uh we went Yeah, so he was, that was a legendary day. That was a crazy thing. That was a legendary day, no doubt. There were a lot of beefs at that barbecue, too. There was conflict, which we could get into later, but we could actually take up the whole podcast with that. Yeah, that would be the hip-hop champ situation. Hip-hop's important. We're both hip-hop heads. When did hip-hop become a thing? When did it be like, oh, I can see myself in that, and it's motivating me to do things? How did that come to be? So I, you know, I was probably like 12 years old when Rapper's Delight came out. That was kind of the first one that a lot of us heard. You know, if you if you weren't in New York, I think the first one. where I was really like, Ooh, this is going to be... very different was when, uh, run dmc hard times came out and i was like this has a whole nother feel to it than anything i've heard um so i was like really into it but the thing that made me like locked into hip-hop so i went to school in new york i have this friend who i grew up with who um had gotten shot in the face like in like some kind of weird uh gun accident and he was blind um and so you know i was very worried about him we were super close uh And, you know, I would call home to my mom. I was like, how's Seth doing? Is he all right? Is he going to be okay? She's like, you know, he hasn't spoken to anybody since he's been shot. He's just, you know, sitting in the hospital. Everybody's worried he's suicidal, this and that and the other. And so I got the idea of being like a dumb person. 18 year old kid, uh, I was like, you know, hip hop is so great. Like, it's like the most, like, you can have nothing and you'll rap about your shoes. And like, my shoes are the best. And I'm like, you know, he needs that feeling. And so I would record the DJ Red Alert and Chuck Schiller shows. That's where all the good stuff was. Yeah, yeah. You know, I'd sit there on Saturday night. Everybody else would be out and I'd be just like recording. And I would send him like the tapes of the show. And then I called home to see how he's doing. he's talking now, but all he'll talk about is hip hop, you know, and DJ Red Alert and, you know, rock the bells and all that. And I was like, oh, cool. So then when I went out that summer, we started a hip hop group called the Blind and Deaf Crew, blind and DEF. And we rapped about being blind. I wrote most of the lyrics. I wrote things like, the blind deaf crew, rapping about being blind. And so then I was very attached to it after that. And hip-hop is very business-oriented, as you know, particularly in the early days. It was all about business, strictly business. And so... When I started writing on business, so many of the ideas came from those old hip-hop songs. And so I started just putting the lyrics at the top of the post. mostly to give credit where credit was due because if like, you know, I hear an idea and then I write an idea. This is my inspiration, so I'm going to, you know. Yeah, just put it out there so like, you know it came from, I didn't come up with this idea. Right. It came from Rakim, or it came from Rick Ross, or whatever. And so then I got really associated with it and all that kind of thing. Okay. It was funny, though, because knowing where I'm coming from, I'm of a different vintage, though, right? Yeah. So... you, it would be impossible for you to be growing up on the West coast and be of your vintage and actually like grow up in hip hop because it just didn't travel. There's no, there's no DSPs and stuff. So it's like, versus like, you know, I'll be six years old listening to, you know, wherever it's hammer or whatever. Like I'm nine listening to 36 chambers. You know what I'm saying? Yeah. Yeah. Um, so it's like, it's like a different, yeah, no, no, no, no. It was actually, it's so my friend Roland Fryer, the economist at, uh, Harvard just wrote a piece in the Wall Street Journal's kind of on, you know, did hip-hop had a bad influence [10:00] in society and the way he did this study, which turns out, no, it had a positive influence, was because No radio stations played it, so you could track the cohorts from the cities that had hip-hop and see how they did versus the cities that didn't have hip-hop. And it turns out the cities with hip-hop outperform. The inspiration level is just like their level. I guess that's the drive. Bye. You're getting into college, you're doing your thing. And now you're getting into entrepreneurship. How do we get into, I'm going to work with Mark and do all these things? I own a position to be... an entrepreneur very easily in that, you know, by the time I was 25, I was married, I had three kids. Like, so I was just trying to keep the lights on. That's normal life right there. Yeah, yeah, yeah. Keep the lights on. That was my main thing. And, you know, it was a little, you know, starting a company was a much more unusual thing in those days. You know, it's kind of more systematized now. So I started as an engineer at a company called Silicon Graphics. And it was a kid. Yeah, Ken Coleman. Yeah, yeah. Ken hired me in there. You know, I kind of learned the ropes, learned how to be an engineer, learned product management, that kind of thing. And I ended up... going to Netscape. A lot of XSGI guys were there. And that's where I met Mark. And, you know, I was still just trying to, like, I was happy to be, like, making money. And, you know, Netscape went public. I think I got a million dollars out of that, which was crazy. I couldn't even believe it. I was so rich. Yeah. And so you thought, yeah, yeah, yeah, yeah. So I thought, um, And then, you know, once Netscape kind of got sold, Mark is like, hey, we should start another company. You know, I didn't cross my mind to do something like that. And I was like, well, who's going to be the CEO? He's like, you. And I'm like, all right. I'll try, you know, I don't know anything about being CEO, but that's kind of how I got started. I mean, I also, actually speaks to a lot of, you know, the sinusoidal trajectory of, say, a company that finally gets to an opportunity. So, like, the first time you really step into something, there's going to be downs, right? Always. Like, I don't know anybody who doesn't have, like, not just downs, but existential, what we call the Wiffio moment, where fucked it's over. Like, you can't build a company without that. Like, that happens. And a lot of, look, a lot fail just because the founders can't handle that moment. It's a tough moment. Yeah, it's definitely, I find it to be like, This whole, I mean, I'm new to the investment game over the past five years, but like, I do find that like the people is everything. I'm very good at picking a company. Yeah. All right. I feel like it's probably an ever going evolution to like identify great people. Yeah. Yeah. Yeah, well, it's really hard because you only know – so much about them and it's a little like dating right they're coming they're putting their best foot forward when they meet you so you really have to get deep into their background but then like it's also it's one of those things where you're one person in life and you're another person under extreme pressure like those are two different people yeah your whole chemistry starts working differently yeah so like how people are going to perform in that situation is um really it, you don't know what's going to happen, to be honest with you. And I mean, this is like one of the reasons, you know, Elon love him or hate him. God bless him. That guy is very good under pressure. Yeah, like he is like there's no amount of pressure where he's not going to perform. You know, that that is what you want. And that's what you know, that's why he's probably the greatest entrepreneur of our time. So when you're when you're all the way down to extreme down, what is one one bar that comes to mind as you know, it was popping your mind, like, well, I got to keep going. I Well, it's probably from ether, you know. What is it? Left me for dead, stiff and rotten, the Nas line. You know, like that's the feeling like… Okay. Because you remember, Ether came out when Nas was at his lowest. He was way down. And Jay-Z was literally the biggest rapper in the world at that point. And he got hit with... the takeover where Kanye did the beat. Right. So it was that fucking, you know, everybody was like, wow. And then Jay-Z wrapped his ass off. That was a great fucking song. song Takeover, and Nas at the bottom, bottom, bottom that any rapper could possibly be at. And he said it, you know, you left me for dead. And so that's the feeling you want, like, "No, I'm going to come out with ether." When you think I'm dead, I'm going to come out with ether. I think the same thing, and now because of Kendrick, we have the evolution of Not Like Us, which is like ether 2.0. It is. It's a little... [15:00] A little different one. to me in the way that went down because Kendrick's career was also on the down during that battle. But he wasn't as far down As Nas was at the time. But Drake was like Jay-Z. There's no question. But that was more... You're taking rap in a direction I don't like. Right. Right. You know, so I'm going to have to get out of my room. Stop talking to myself. You know, just did a whole album of introspection. I'm going to come out of that and I'm going to deal with you right now. I don't like you. That was, that's a little different feeling for me, but I got you. All right. So we digress. Back to the tech thing that all y'all want to hear about. So build the company, sell the company. Thank you. I don't know why nearly immediately after that, you decided to go build a venture firm and then mess around with NuMe. But tell me about this 2008 to 2011 time where you said, I'm going to build a venture capital firm. And when you did it, It was nothing like the other stuff, really. I mean, the way you approach people, all these people running around the building. Who's a partner? Who's not a partner? Like, what's going on in this building? You know? Yeah. Yeah. Well, so the idea was it kind of came out of how hard it was. to build the company. And I always felt like, wow, If you're a founder, And you start something... You know, you don't know how to be a CEO. You don't know how to build a company. And you kind of think... at least in those days, I thought, oh, the venture capital firm would help me with that. And they weren't really set up for that idea. They were kind of like really, really good at picking entrepreneurs. They made a lot of money for their LPs. And they were very smart guys. But, To build a company, you need to feel. like you can run a company. And that means you've got to be able to call important people. You've got to be able to kind of have somebody to talk to who really knows what you're going through, who can, you know, help you with things. You need to be able to have, like – turn on a recruiting army. You've got to, you know, these days, you've got to be able to deal with, like, policy issues and this kind of stuff. And so, you know, the idea behind Andreessen Horowitz was really, like, wow, what if we made a better product, not for LPs, but for entrepreneurs? The first thing you had to do was, like, forego, like, a lot of your... because you have to take the fee monies and build this platform to help entrepreneurs kind of build their company. And that was kind of the initial idea with it. And then we always thought the thing that we could provide entrepreneurs that they would value most would be this network. And we always thought of the firm as a network. And so how does the, you know, how can you, like if I could call anybody, if that kind of bleeds into everything you do because now like you feel like somebody. This was the idea behind the CEO of Barbecue. Oh, if I go to a barbecue. Yeah. And I'm me, just like this little entrepreneur. And then I see Mark Zuckerberg, Larry Page, Kanye West. okay, I must be important. I feel like I have a whole other feeling about myself, and that was a lot of the idea of the firm. Because as a network, one of the things I thought, I was like, I'm looking at all these venture capital firms, And not any of them are networked like literally into any black people. And I know that's like a crude way of saying it, but I was like, well, that's an opportunity for me. I'm going to fucking build that network. I'm going to build the definitive network in Silicon Valley for like black entrepreneurs, black entertainment people, everything. And that's going to give me a huge advantage because that's a whole like important part of the world. And so that brought me to NuMe. Okay. I was looking under every rock. I mean, NuMe was definitely under a rock. This woman, Angela, Ran it and it was an accelerator for black entrepreneurs. There's a lot of talent and knew me So there was you there was Chris Bennett Chris Lyons, right? Hey, hey Jones Fred Fred Hudson. Oh, yeah, Fred Hudson. Yeah, yeah a lot of talent and a lot of ambition You know how to knew me so so that ended up being like a really good idea at least for me I think for the firm we got a lot of value out of it and We're a major investor in Wonder School. We had Chris. Chris still works here. He's done a lot of things. Nate worked here for a while. All those kinds of things. So, yeah, it's been… Yeah, for the small investment. I went. I spoke at Numi twice. Right. Aaron and Nnamdi do 645. Aaron came out. Oh, yeah, yeah, yeah, yeah. Aaron popped up. They popped up to do. I want to be bent. I want to see what's going on. They started their thing. I should have hired Aaron. That's the one that Chris Lyons got. winners. [20:00] Tell me something about picking winners. How do you manage emotion? How do you become great at picking winners? It's a tricky... thing because Really what ends up mattering is the entrepreneur. Like if you just, if you had a perfect job, rating system, and you could exact go, like, do you want this entrepreneur or this idea? You would take the entrepreneur every time. The problem is, as you were saying, it's hard to evaluate entrepreneurs. You don't know that much about them when they come in, and you don't know how they're going to perform under pressure. And so then you need the idea to help you evaluate the entrepreneur. Like, what did they go through in their life to get to that idea? And how good an idea is it? And like, you know, how much original work did they do to get there? And that kind of thing. So like that ends up being the lens at which you see the entrepreneur. Now, the mistake to make is like, there's something you don't like about the idea, and you pass on a great entrepreneur, like that's a mistake, or, you know, vice versa. You're like, this is a great entrepreneur, but you don't. realize his idea is no good. So, not as good as you think. So, it is tricky in that sense. And then, yeah, you don't know how they're going to perform under pressure. And then the other thing that, like, the other trap that, you know, we'll get into from time to time is, We have all these great analytical people. and they'll be able to figure out everything that's wrong with any idea or any company or any entrepreneur. But that's not what matters. What matters is Are their strengths world class? So are they the best in the world at something? And then you overlook everything else. Like kind of going back to Elon, like he's got some personality issues. You know, like he upsets people from time to time. But like his strength is so ridiculous that you would put money on that every time. And I think you want to. you want to look at the magnitude of the strength. And I think that like every entrepreneur we see who gets to us is pretty good. But are they outstanding? Are they better than everybody else at this? And that's what makes a company great, and that's what makes an entrepreneur great. Not, oh, they don't know how to do this, or wow, they know nothing about finance, or they're, you know, whatever, they're done with marketing. Like, there's so many things that are wrong with every entrepreneur. Every great entrepreneur has got something, like, horribly wrong with them. But That's okay. you know it's what are they truly great at what they're great at when you think about you know picking the person and be like well those things not so good but this person's gonna make some moves definitely sent waves through silicon valley but blow back from the culture unlike you know the firm kind of getting behind the trump campaign to some degree you know um and so when you think about something like that Do you think about it as, well... I know something some people don't. So this person's going to you know, a rise is the winner or this person is the winner. And so I'm going to play the game with the winner. Is the former or the latter? How do you think about the decision-making? With the Trump thing... We try to be as apolitical as possible, but we had run into – Massive problem with the Biden administration. on kind of two fronts. One was crypto and then the other was AI. So on crypto, they basically They didn't outlaw. So they did these extra legal things. So they kicked all the crypto entrepreneurs out of the banking system. They hit them with Wells notices that were like fake lawsuits, you know, on things that – We're just undefined. And so we were trying to get them to like, can we just have some laws? Can we have some regulation? And they blocked all of that and they just kind of attacked the industry because they didn't want it. And so we were stuck there. But the bigger problem around that was we didn't have a relationship with. with the administration, I mean, we couldn't get one. And it drove me crazy. Because like, I mean, I called everybody I knew to try and like, can we get a meeting with with Biden? And people call me arrogant, like, for you to ask for a meeting with a meeting with Biden. Yeah. But you know, then I come to find out, like, so I spoke to Sundar at Google, I spoke to Dave Ricks. Yeah. Yeah. Tim Cook. Nobody got a meeting with Biden, the whole four years. And then the same thing happened to us. And AI, where they went as far as like the last executive order from the Biden administration was to say it's illegal to sell a GPU with [25:00] prior government approval for that sale. So it would have literally taken us out of the AI industry you know, competition at the chip level. The thing that we decide is, okay, like we've got to have a relationship with who's ever the next president. And at the time we endorsed Trump. on tech, um, He was clearly going to beat Biden. There was no question. Then Kamala came around. You know, Kamala came around. And so it was like, OK, what if she wins? We knew we weren't going to have a relationship with Biden, so there was nothing to lose. With Kamala, I had known Kamala. Kamala came to the CEO barbecue. She'd been in my house many times, so forth. So I knew her. And I called her up. And I actually wrote a tech policy for the campaign. a tech policy position, which that made me nervous. I got to be quite honest with you on. But on that, you know, we made a donation to her just so that we would have a relationship there. Anytime you're involved in politics, people are going to be mad. You know, they're mad at Nicki Minaj now. They're mad at Kanye. They're mad. From my position, I have to make the world work for me. new tech companies like that. That is, you know, our responsibility as kind of the big VC. I mean, for me, like, Probably not a decision I could make, but I'm not sitting in your position either. I also don't run a big building as a money machine. You know, there's two different sides to that. Let's look at it this way. From your vintage, you come from like real hippie. This is Silicon Valley. People's running around the streets, probably no shoes on. Like that Silicon Valley, right? Yeah, no, it used to be different. Yeah. Yeah. And so like that Silicon Valley was like, let's call it. performatively inclusive. it was exclusive since it's why we did new me and all these things like you could be involved but kind of not because it was a thing but now We're kind of in the space of performatively exclusive, but it's quite the special time for it to be inclusive because AI has commoditized the ability to build. Yeah. You know what I mean? Like people who were otherwise not in the game can now be the expert and be the best. I'm going to be with MIT. Yeah. Okay, so I have a little different version of the history. And you got the history, so I'm not going to argue the history. So this is why we're here. Yeah, so I think in the early days, the only people who wanted to be in tech in the 80s were like, you know, Jews, Chinese and Indians men. Like there were just very few women who majored in computer science who did anything like that or anybody else. You know, and Ken was like super exceptional. Tech wasn't sexy. Yeah, it wasn't built to be exclusive. It was like these are the people who are participating. I know what I'm good at. I value it highly. I can test for it in an interview. I'm hiring me. Like, that's just generally how it goes. So, like, once you start that way, that's how it grows. And, like, not that nobody can get in, but the people that get in are, like, hyper-exceptional, like Ken, right? It's not that different than hip-hop, right? Eminem can get in. That's a very high level. Yeah, that's a very fucking high level to be Eminem. I mean, Jesus Christ, that guy is. But if you're a mediocre white guy, you've got no chance. And some of that is because you get tuned. It's the media. It's the culture. I always tell people this. I say, when I'm explaining how people self-select and how the media has shaped our worldview, protect her in glasses and can't talk well, who's the better developer? Yeah, yeah, right, exactly. Oh, it was me. Yeah, yeah, yeah. Right? So, like, you know. Yeah, you have to overcome that. You have to overcome that, yeah. So I think that what we did is we said, look, You're... You know you. But you don't know... the skills you don't have and you have no idea how to test for them and you have no idea that they even exist. So like, how do you how do we get around that? And so, you know, we created this hiring process where we have people from. different backgrounds, look at every job profile and go, hey, if I was hiring that, I would look at these things. You need a relationship person. Actually, how we got Chris was – It was really a relationship job, but the hiring manager had been an investment banker. So he had all these investment banker criteria. And if anybody knows Chris Lyons knows he flunked that test. But I was like, well, this is a relationship job. So let's look for people who are relationship people. And I was like, he is a world-class relationship guy, really good. And so because we hired him on that criteria, which was the right criteria for that job, got on board, we knew why we hired him. If we had thought that we hired him for financial modeling and he was black, he would never got promoted. He's been promoted like eight times. And that's because we know what he is, right? Like we know the, [30:00] special thing that he has. Philosophically, I think the right answer is You really want to widen the bar, not lower the bar, raise the bar. You want to widen the bar so that you can see all the criteria that you would need in that job, even if you don't have it. You have to get hired on the same criteria. Right. I remember when I went to get a job as an engineer. So there was a moment I only got a job for an engineer for like a year. Yeah. And I'm the only black person there. I was like, cool. And I was like, all these people are talking crazy here. This person is talking sideways to me at work. Now I understand why this doesn't work. Like these people actually can't work together. They don't respect each other. Yeah. Yeah. They don't know why they hired you. You know? Yeah. It's crazy. Well, there's a funny story that I had this crazy conversation with Stout. And, you know, he calls me up. He goes, Ben, I used to be president of Sony Urban Music. And I was like, Yeah, Steve, you know, it's Steve. You know, he's talking loud. And I was like, I know. And he goes, yeah, but it wasn't urban music. It was black music. They made me call it urban music to not sound racist. And I was like, oh, shit, that was stupid. And he's like, no, no, no. That was the stupid part was because we called it urban music. I wasn't allowed to market in rural areas, like all black people live in the city. I was like, that's a dumb shit. And he's like, no, I don't even think you're listening very carefully. I was president of Sony. urban music. I had Michael Jackson. What white people don't like Michael Jackson? It wasn't fucking black music. It was music. But the program for it made it smaller than it was, not bigger than it was. And I was like, you know, that's what we did in Silicon Valley. We had urban HR. It's like, no, it's talent. It's not black talent. It's talent. You just have to be able to see the talent. You got to be able to listen to the music. And then now it's gone the other way, as you said. Like, I agree with that. It's like, oh, it's a meritocracy. Well, what does that mean? Well, I'm going to hire me. It's like, OK. But I think, like, it's been a big advantage for us. venture capital. Like, I mean, you know this. Like, we get to people that they're never getting to. Right. Because we've got people who can do those things that they won't hire because they can't. Not because they – it's not a racist thing. It's they can't see them. They can't see it. They can't see them. They're blind. They're talent blind. And, you know, like – And all of us are talent blind. That's why you need a process and a methodology. You need a process to open your aperture and stuff. Margaret, the legendary Margaret Wenmark, is running marketing for us in I was like, she's around forever. Yeah. Yeah. Yeah. Yeah. She, she's here at the beginning, right? Yeah. Yeah. From the very beginning. I remember when I met her, I was like, oh man, I need, I need someone like this one day. Yeah. She had all women working for her. I was like, Margaret, like, why can't any men like past year, like what's in your profile? They can't, they can't fucking qualify for it. She's like helpfulness. It's like, she's right. And, but, you know, like that, think about that as a criteria, like how valuable is that in venture capital? Like everybody wants to like venture capitalists who's helpful. But I, you know, interview for that and so forth. And so like by developing that, you know, at the time, We were, you know, predominantly men now, I think for like 54% women, but we didn't have any women criteria. We just had like helpfulness as a criteria. And, you know, these are the things that change that make you stronger. Um, Not like you're not compromising. You're actually getting better by being able to see. If you can see all the talent, like we're all in a talent war. We're all competing for talent. If you can see more talent, you're going to win. We're getting to the climax here in terms of like, all right. Now we have AI. I'm actually feeling a little bit of an itch, like I'm not building right now. I literally can't make it through a day without generating some code and building something, because it's crazy. I'd be on cloud code on my phone, and it's like, You know, now, now, like, for instance, by the way, I was talking to Deshaun. He became an engineer. Oh, yeah, no, no, no. But I told you, Sean, I was like, hey, man, you don't have to hire anybody to build this stuff. He's like, all right, I'm hopping. And now he's addicted. Yeah, he's addicted. He's out of his mind. Right. I got, I got, I got an agent here. I got an agent here. I only got four people. Maybe I can make it two. Like, you know, like. He's almost like a AI fiend. He's a AI fiend. I think there's like, literally, I think there's gonna be AI rehab at some point. Yeah. Yeah. Because it's like, if you right now, if you can't force, if you're not like in your bed at night, like, go to sleep, go to sleep. then you're not on it. Because like, there's so much that can be done. It's like... It's crazy, but... I guess my point is that going back to like the larger aperture of opportunity here, like now we can like open up the, the, the stratosphere for like people who would not have been a fit before. Like if I come to you, I've never built a product in my life. I've never worked at a tech company. Oh, yeah, I worked at this oil and gas company. I really know this process works and we build this company. It'd be hard to accept that this might work. Oh, yeah. Right? But now you could be like… You can go from idea to company very fast. Right. Yeah. Without being able to hire all the barriers that were there before… aren't there? I 100% agree. I mean, like, that's why I think it's like, it's more important to get people into tech, because they're not even getting into tech. They're just getting into ideas. [35:00] It's like, okay, here's By the way, it's happening in every... Like creativity is now the thing. Creativity is the thing. Like, yeah, if you can use the tools, all you need to do is be creative. This opportunity is probably even bigger because, right, any idea you have, you can make. You can manifest. We both were around when the internet began, was on BBS boards and Prodigy.net and all these things. I mean, like. CompuServe. Yeah, all that. You know, a million little discs sitting around. Yeah. AOL used to fucking send a disc to you. Like you couldn't open your mail without getting in the AOL. Yeah. Cereal box. I was expecting a toy. You know, memory was hard to come by back then. I used to go to the store. I used to get like a box of them and just like can't put tissue in that. And I record on that. Um, but, uh, The internet was big. Like, I mean, that's how we both got to this conversation right here. But, like, to me, it's like, this is bigger. This is like the fourth industrial revolution. Like, what's your. People talk about. I think it's clearly bigger than the internet. People talk about it like. It's a sector. Like we're talking about, oh, it's a bubble like crypto. I'm like, crypto is a thing. It's a conduit, but like it ain't AI. You know, AI is a new computer, right? Like it's a it's it's like the reinvention of the computer. Imagine the entire computer industry got invented. But. Instead of it developing over 80 years, it developed over two years. And then like that's AI. And then it's more powerful on top of that. Anything you think you can solve, you can solve. Mm-hmm. We've already seen it. People are taking it so for granted because they're so worried about Skynet or whatever, but traffic deaths, huge problem. solve with Waymo. You know, cancer, we're solving that for sure. Any idea, any like major problem Like, is getting figured out with AI right now, like it is wild. Um, so like as an entrepreneur, it's like, just find a problem, you know, and you can go like solve it and people will be like fired up and willing to pay you and you've got a business. I mean, it is like that powerful. So you've been through a bubble, like a real bubble. you know, there's money going out here. It ain't coming back to people like where, where, where is the bubble S part of this? I don't be concerned with it, but where is for everybody who's talking about a bubble, where's the bubble component to this? I mean, you got to remember right there. Oh, the internet bubble felt like, The internet's here. It's kind of important. You could have made a lot of money on it. Look, where bubbles come from and where the internet bubble came from is like there's technology and how that advances and then there's consumer adoption and that's like goes more in fits and starts and then there's uh investor sentiment right and investors are like manic depressive um they get overexcited and then they get totally it's never gonna work and they'll actually so there there's a kind of an opposite of the bubble going on right now i think a little bit in uh with investors. But in the internet time, what happened was, um, investors were like the internet, it's the biggest thing ever. I'm putting all my money on it and it's never going to go down in price. Okay. The problem was the time they decided to do that, there was like, 50 or 60 million people on the internet. Half of them were on dial-up. It was very expensive to build these things. There wasn't a big enough market to make money in selling a lot of the ideas that people had. So the thing crashed down. I mean, right now, though, ChesubT ain't making no money. These guys are like, Gemini, they're giving away for free. How can we catch up with Anthropic? Just keep running this computer. Well, kind of. OpenAI went like zero to $20 billion 2022. And by the way, the price of a token is dropping like 100-fold a year. Yeah, wait for that crossover. Look, consumer adoption is... going faster than anything we've ever seen. The technology works and is advancing as fast as everything we've seen. And are people going to get Because we're so early in the cycle, there are companies that are in a good position now that there will be a technological change and they'll get wiped out. But that's not really a bubble to me. That's just like investing. That's just how it goes. That's part of times. People were trying to say, oh, the demand for NVIDIA chips is artificial. They're just investing in companies. And that's where – like the demand for NVIDIA chips right now is just unlimited. Like they can't make enough chips. Like that is actually true. I just think, all right. Correct. Now, will we hit a bubble at some point? Yeah, quite possibly. I mean, look, we had a bandwidth bubble, a bandwidth bubble in 2001. Mm hmm. We have, [40:00] too much bandwidth. Like, how could that even be possible? Like, we're at the beginning days of the Internet. How could you have too much bandwidth? Well, the bottleneck moved. Right. Like we had a bandwidth bottleneck and then that became like, oh, the servers weren't good enough. We can't even put the bits on fast enough to serve the video. Thank you. in my generation and in every generation, you could not throw money at a problem. Right? Like, if I want to catch you in your company and I said, I'm going to hire a thousand engineers tomorrow, now... I have data and I buy enough GPUs, like I can solve shit. And the greatest example to me is, Right, like OpenAI got to $10 billion in revenue. And Elon started a cult. Oh, yeah. And he, like... He's almost at product parity with them You know, in like a couple of years. This is an important position, though. So how does this change how you choose what you're investing? The world, to me, they've seen the change happen, but they're not changing the way they're thinking around like, this is the deal I should do. This is the person I should back. You have to reshape how you think about investing. It's hard to calculate. So there's ones that you won't expect. to sustain that too so like one of my favorite examples is uh 11 labs oh yeah those guys 11 labs Okay, of all the things, if you rewind back to when we made the first investment, that you would go, okay, they're probably vulnerable to the big guys is – Well, the big guys aren't going to be able to make their models talk. Really? They will be able to. Of course they'll be able to make their models talk. But it turns out... that 11 Labs, a very, very clever company. One, they got in the developer stack they built a better product for developers so What does that do? Well, they've got a built-in set of customers, and then they're in the data feedback loop. Oh, they're getting every voice case and so forth, and they're training their model on better proprietary, like real world, in-use data. Then they built the product kind of... subtleties around it. So, oh, yours can talk, but I can talk. in Spanish. Exactly like I'm talking in English with the exact intonations, laughing in the right way, everything perfectly. Oh, guess what? I get the Netflix deal. I get the Televisa deal. And now I've got more feedback. So I'm kind of building up. a voice product that's very, very special. And that, yeah, you could throw more GPUs at yours, but you may not be able to do the things I do because I've got better data. I've got, you know, like I've got a model that's really tuned to the use cases that have value, as opposed to like every damn thing that everybody wants to do. And so like there are things like that. On the other hand, you know, like every software company, SaaS company in the world, is trading at a very low multiple right now because there's a real belief that, hey, Even as a customer, I can rewrite. No, it's real life. I was like, you know what happened? I got in cloud code. right this must have been a year ago i got in cloud code i was at our admin cursor you know i'd already been you know before that i'm copying from jvc yeah so i got in cloud code i was like this goes crazy. I'm never going to look at code again. And then the next day I was like, cancel HubSpot. I'm building my own CRM. Like, this thing sucks. I want it to work for me. I could just build my own. One week later, got my own CRM. Yeah. There you go. Affinity? Cancel. Got my own fundraising platform. Yeah. Like why? You have to build something that's very special for him to be like, I'm not going to build it myself. And right now, obviously that's unique to engineers or like product-oriented people. But like the time may come where you're on a website, And you're like, I don't want it this way. Can I have it this way? So I think the world, or at least Wall Street, is anticipating that a lot. Now, look, there are other things that those companies have for your company. building something for you. Right. If it's at the individual level, probably a little different than, you know, American Airlines running whatever SAP or something. Like that's a little harder to figure out what the hell that is. And then, you know, they've got data and, you know blah blah blah blah blah but still yet and still like that's a real thing where you can chase you by using gpus that you don't even own that Anthropic owns can chase down HubSpot for your use case like that. Oh, right. Like that. You wouldn't even – consider doing that no it wasn't a consideration but now yeah i i reverse yeah i look at buying some [45:00] buffer, I'm like, is this really worth $400 a month or should I spend four days? and just have my own. It's a reverse consideration. Cheers, sir. I'm going to actually bring out my paper for my signature question. I want to get my signature questions right. All right. This is about legacy. when the next vintage of VCs talk about Ben Horowitz. Thank you. What do you want them to say? What's your mark on the game? Thank you. Thanks. Thank you. That's something I really don't think. Mike Tyson had a great question on legacy. He's like, what are you talking about? I'm just going to be dust, dirt. It's a crazy question. I don't think about that. You know, for me, I hope that people feel like I... move the culture of investing to a better place, to one that's more about... helping people build companies than just making money. So that would be my hope. Okay. So before we finish up that question, I was talking about VC. Think about Ben as the human. Because your name's on the firm. They made you have your name on the firm. Yeah. Like, you know, you have this direct association all the time. Yeah. Me as a, you know, Ben is a human. When you leave, it's like... What makes you feel accomplished? What makes you feel like, oh, I did my thing? Yeah, I really feel like it's the... The people who work here, going back to the NFL analogy, the football coaching tree idea and that kind of thing. Like, I mean, to me, You know, how... Many, you know, who are the great people who worked here? And they're like, did they become better people while they were here? And like, that's the main thing on the business front. You know, I think. You know, on the personal, personal front, like paid in full to me is really, really important. And I intimidate everybody into coming, as you know. Thank you for your contributions. You know, if we forget. the people who kind of sacrifice their lives to inspire us. Um, then you get into a very nihilistic existence. We're going to make those guys feel as important as they should feel. And then – in return, like we just get so much more meaning out of who we are. And so that's super important to me. Yeah. I think as that can grow, that can become – you know, the SAG, the pension plan, the, the, the, you know, the base of, you know, hip hop continues. I think it's a, I think it's a important thing that probably should have done a long time ago. Um, people are too busy talking about urban radio. Yeah, you know, like it's so funny how many, like one of the things that got me to start Payton Full is so many like podcasts and this and that. Somebody ought to do something about this. Someone should give them their flowers before they're gone. Yeah, yeah. Well, let me go back to my my back. Yeah, yeah, yeah, exactly. So I was just like, well, why don't I just do it? And, you know, like I said, I get way more out of it, I think, than everybody. So, you know, it's an easy thing to do in that sense. What's something that you know, you think it's kind of Happen in the next, say, 8 to 12 months as a result of the current landscape in technology? So I think the... Infrastructure for AI is Just out of date is not strong enough. It's just like wrong. And it is very reminiscent to me of the infrastructure for cloud computing being wrong. Like everybody in the beginning of cloud computing is like, oh, we'll just take – you know, our software, we'll put it in the cloud, we'll put it on the other side of the wire, and then that'll work. But, you know, it didn't work at all. And eventually, we rebuilt everything from storage to networking to, you know, the way operating systems work and virtualization and, you know, containers and this and that and the other. And, like, you know, the whole thing got reconstructed. If you look at AI, I mean, just start with the workload, right? So, somebody makes a database query, you kind of know what that is. Somebody makes an AI query, like I think it'd run for a week. I mean, who knows what, you know, so like what is the underlying set of systems that handles those requests that manages that correctly and so forth? Like it doesn't exist yet. If you look at, I mean, NVIDIA, God bless them, but like, You want to talk about a power-hungry, low-yield chip built for something entirely different. For playing video games. For playing video games. And now that we know a fair amount about the models, [50:00] the right answer probably not um then the models themselves like in these models you literally have to suck down all the data on the internet to get them right like that seems like a lot of training is that going to be where we end so i think that the you know from the model on down everything's going to change and there's going to be a lot of uh opportunity in that all right man well thanks for joining us all right oh no great great to be here Like great to kick it off. Yeah, for sure, man. No, it's going to be a great series. Appreciate you kicking off with me. Yes, sir. All right. Thanks again for listening, and I'll see you in the next episode. Testing Z does not guarantee its accuracy. you