20VC: Canva Slashes Growth: How Much is it Really Worth | Demis Hassabis and Jeff Dean: Talent Exodus at Google | Revolut's $50BN CEO Pay Package | Elon Musk's $55BN Terrafab

本周由 Harry Stebbings、Rory O'Driscoll 和 Jason Lampkin 主持的 "20VC" 播客，深入探讨了人工智能在科技领域带来的颠覆性影响，审视了公司业绩、人才流动和新兴趋势。 讨论始于 **Canva 宣布将其 2026 年的增长预测削减三分之一**，原因是人工智能服务成本飙升。Rory 指出，虽然 Canva 仍保持 20% 的健康增长，但增速放缓意味着集成前沿 AI 模型带来了巨大的成本负担。这引发了一场辩论：AI 对 Canva、Adobe 和 Figma 等创意软件公司而言，究竟只是一个功能，还是一种生存威胁？Jason 认为，AI，尤其是 ChatGPT 等代理式应用，正日益使 Canva、Notion 和 Airtable 等“无代码”工具过时。他分享了自己放弃这些服务的个人经历，因为 AI 代理现在可以处理以前由这些平台完成的任务，这表明在许多专业消费者（prosumer）场景中，“无代码时代已经结束”。小组成员质疑 Canva 开发自有 AI 模型的策略是否足够，或者是否为时过晚，无法对抗通用 AI 界面带来的“捆绑效应”。 这引出了一个更广泛的讨论：**私营公司面临的挑战**，例如 Canva，与像 Figma 这样的上市公司（后者也因 AI 使用而导致毛利率受损）之间的差异。文中强调了有限合伙人（LPs）和风险投资家（VCs）在快速变化的市场中评估这些公司的难度，特别是考虑到“23 年和 24 年的犹豫不决”将导致“26 年和 27 年的付出代价”。 随后讨论转向了 **谷歌显著的人才流失**，特别是 Jeff Dean 在工作 27 年后离职（并带走了其他三位传奇人物），以及 Demis Hassabis 从 Google DeepMind 淡出。Rory 和 Jason 都认为，这些离职反映了顶尖 AI 人才追求“登月计划”式科学发现的自然愿望，他们希望摆脱大型上市公司的企业官僚作风和营收压力。目前可用于独立 AI 研究的巨额资本，使得此类举动越来越具吸引力。这也凸显了 **精英 AI 工程师“神级”薪酬** 的出现，使得小型公司难以竞争，尽管对于那些不关心大型科技公司 AI 议程中“无聊”部分的人来说，仍存在机会。 小组成员还谈到了 **当地居民对新建 AI 数据中心的反对**，引用了众议员 Ro Khanna 提出的“数据中心权利法案”。虽然有人认为这种邻避主义（NIMBYism）可能会减缓 AI 的部署，但也有人相信，各州和各县为争夺经济利益（就业、税收）而形成的竞争格局最终会促成解决方案。Elon Musk 雄心勃勃的 **TerraFab 项目**，旨在建立自己的芯片制造能力以规避台积电（TSMC）的供应链限制，被视为对这些限制的大胆、垂直整合的回应，尽管其规模和雄心都令人瞩目。 讨论的一个重要部分集中在当前 AI 环境下 **创始人领导型公司与非创始人领导型公司之间的区别**。Jason 明确表示，他相信“我所做的任何非创始人运营的投资都将归零。在这个时代，它注定会归零。”他认为，只有创始人拥有驾驭 AI 带来的彻底变革所需的深厚承诺和远见，这引发了关于高激励高管薪酬方案的讨论，例如 Revolut 提议的 CEO 薪酬方案。虽然此类方案可能看起来过高，但在一个私人资本允许创始人避免公开市场压力的市场中，它们可能是留住创始人所必需的权宜之计。 最后，播客回顾了 **近期第一季度财报结果**，指出 WhatNot（一个直播购物平台）、Shopify 和 Atlassian 等公司表现强劲，而 Datadog 则略有放缓。WhatNot 的成功，实现同比两倍增长，被强调为一个提醒：并非所有盈利的企业都直接由 AI 驱动。小组成员总结道，那些迅速适应并保持强大基本面（例如 Atlassian，实现了自 2015 年以来的最大涨幅）的公司将获得回报，而那些未能正视 AI 颠覆性力量的公司则面临过时或被 Bending Spoons 等整合者收购的风险。核心信息很明确：在 AI 时代，适应能力、强大的领导力（最好是创始人领导）和清晰的价值主张对于生存和发展至关重要。

This week's "20VC" podcast, featuring Harry Stebbings, Rory O'Driscoll, and Jason Lampkin, delved into the transformative impact of AI across the tech landscape, scrutinizing company performance, talent shifts, and emerging trends. The discussion kicked off with **Canva's announcement to cut its 2026 growth forecast by a third** due to soaring AI serving costs. Rory noted that while Canva is still growing at a healthy 20%, the slowdown implies a significant cost burden from integrating frontier AI models. This sparked a debate on whether AI is merely a feature for creative software companies like Canva, Adobe, and Figma, or an existential threat. Jason argued that AI, particularly agentic applications like ChatGPT, is increasingly rendering "no-code" tools like Canva, Notion, and Airtable obsolete. He shared personal anecdotes of churning from these services because AI agents now handle tasks previously performed by these platforms, suggesting that the "era of no-code is over" in many prosumer contexts. The panel questioned whether Canva's strategy of developing its own in-house AI model would be enough, or if it's too late to counter the "bundling" effect of universal AI interfaces. This led to a broader conversation about **the challenges faced by private companies** like Canva versus public ones like Figma, which has also taken a hit on gross margins due to AI usage. The difficulty for LPs and VCs in valuing these companies in a rapidly changing market was highlighted, especially given the "hesitancy in 23 and 24" that will lead to "paying the bill in 26 and 27." The conversation then shifted to **Google's significant talent exodus**, particularly Jeff Dean leaving after 27 years (taking three other legends with him), and Demis Hassabis stepping back from Google DeepMind. Rory and Jason agreed that these departures reflect a natural desire for top AI talent to pursue "moonshot" scientific discoveries without the corporate bureaucracy and revenue pressures of a large public company. The immense capital now available for independent AI research makes such moves increasingly attractive. This also highlighted the emergence of a **"God tier" compensation for elite AI engineers**, making it difficult for smaller companies to compete, though opportunities exist for those not interested in the "boring" aspects of big tech's AI agenda. The panel also touched on **local opposition to new AI data centers**, citing Rep. Ro Khanna's proposal for a "data center bill of rights." While some argue this NIMBYism could slow down AI deployment, others believe that the competitive landscape between states and counties for economic benefits (jobs, tax revenue) will ultimately lead to solutions. Elon Musk's ambitious **TerraFab project**, aimed at building his own chip manufacturing capabilities to circumvent TSMC's supply chain, was seen as a bold, vertically integrated response to these constraints, though its sheer scale and ambition were noted. A significant portion of the discussion focused on **the distinction between founder-led and non-founder-led companies** in the current AI climate. Jason emphatically stated his belief that "any investment I've made that is not run by a founder is a zero. It's going to be a zero in this age." He argued that only founders possess the deep commitment and vision needed to navigate the radical shifts brought by AI, leading to discussions about highly incentivized executive packages, such as the proposed Revolut CEO compensation. While such packages might seem excessive, they could be a necessary evil to retain founders in a market where private capital allows them to avoid public market pressures. Finally, the podcast reviewed **recent Q1 earnings results**, noting strong performances from companies like WhatNot (a live shopping platform), Shopify, and Atlassian, while Datadog saw a slight slowdown. WhatNot's success, growing at 2X year-on-year, was highlighted as a reminder that not all profitable ventures are directly AI-driven. The panel concluded that companies that adapt swiftly and maintain strong fundamentals (like Atlassian, which had its biggest jump since 2015) will be rewarded, while those that fail to confront AI's transformative power risk becoming obsolete or being acquired by consolidators like Bending Spoons. The core message was clear: in the age of AI, adaptability, strong leadership (preferably founder-led), and a clear value proposition are paramount for survival and growth.