#550. AI Capital Market Frenzy: Anthropic $90B Valuation, SpaceX Super IPO & SaaS Old Kings' Survival Transformation

Listen
问这期播客
会先在本集摘要、章节、转录和笔记里找答案。
TL;DR · AI Summary
AI hype drives capital market frenzy, with Anthropic reaching a $90 billion valuation, while traditional SaaS companies face transformation pressures and AI infrastructure/IPOs become key topics.
Key Takeaways
- Anthropic's $90 billion valuation reflects the intensifying balance sheet war am
- Salesforce spends $300 million annually on Anthropic tokens, making AI costs a m
- Traditional SaaS companies have lost their 2021 future glow and must now be re-p
Outline
Jump quickly between sections.
Discussion on how AI hype drives capital market sentiment and boosts AI company valuations.
Analysis of why Anthropic secured a $90 billion valuation financing and strategic implications.
Exploring how AI token spending becomes the largest external vendor expense for software companies.
Assessing survival prospects and valuation re-pricing trends for SaaS firms under AI disruption.
Analyzing growth in AI infrastructure companies and impact of AI-related IPOs on markets.
Highlighting legal disputes, conflicts of interest, and social backlash risks in AI.
Mindmap
See how the topics connect at a glance.
查看大纲文本(无障碍 / 无 JS 友好)
- AI资本市场狂热
- AI公司估值
- Anthropic 9000亿估值
- SpaceX超级IPO
- SaaS转型
- 收入增长重定价
- AI冲击下的网站业务
- 基础设施与IPO
- 算力商品化
- AI基础设施公司增长
Highlights
Key sentences worth saving and sharing.
Anthropic’s $90 billion valuation financing reflects that AI model companies are entering a balance sheet war.
Salesforce spends $300 million annually purchasing Anthropic tokens, making AI costs a major R&D expense.
SaaS companies have lost their 2021 future glow and must now be re-priced based on revenue, growth, and cash flow.
Chapters
开场 & 播客简介
开场 & 播客简介
Anthropic 开场:Karpathy 加入与 9000 亿美元估值融资
Anthropic 开场:Karpathy 加入与 9000 亿美元估值融资
为什么 Anthropic 可能是私有市场里“最划算”的交易
为什么 Anthropic 可能是私有市场里“最划算”的交易
为什么 Dario 愿意在看似便宜的价格上融资
为什么 Dario 愿意在看似便宜的价格上融资
AI 模型公司的本质:算力、CapEx 与资产负债表战争
AI 模型公司的本质:算力、CapEx 与资产负债表战争
Anthropic 与 OpenAI 融资风格对比:七十分交易 vs 极限定价
Anthropic 与 OpenAI 融资风格对比:七十分交易 vs 极限定价
现金到账、融资条款与 OpenAI 复杂融资结构的差异
现金到账、融资条款与 OpenAI 复杂融资结构的差异
Salesforce 每年花 3 亿美元买 Anthropic token,贵吗?
Salesforce 每年花 3 亿美元买 Anthropic token,贵吗?
每个工程师每年 1.5 到 2 万美元:AI coding 成本只是入场费
每个工程师每年 1.5 到 2 万美元:AI coding 成本只是入场费
为什么 token 正在成为软件公司最大的单项外部供应商支出
为什么 token 正在成为软件公司最大的单项外部供应商支出
一万亿美元 token 收入需要什么条件:吃掉研发工资的 20%
一万亿美元 token 收入需要什么条件:吃掉研发工资的 20%
Jason 的看空逻辑:模型更高效后,token 成本可能没想象中高
Jason 的看空逻辑:模型更高效后,token 成本可能没想象中高
Transcript
开场 & 播客简介
Anthropic 开场Karpathy 加入与 9000 亿美元估值融资
为什么 Anthropic 可能是私有市场里“最划算”的交易
为什么 Dario 愿意在看似便宜的价格上融资
AI 模型公司的本质算力、CapEx 与资产负债表战争
Anthropic 与 OpenAI 融资风格对比:七十分交易 vs 极限定价
现金到账、融资条款与 OpenAI 复杂融资结构的差异
Salesforce 每年花 3 亿美元买 Anthropic token,贵吗?
每个工程师每年 1.5 到 2 万美元AI coding 成本只是入场费
为什么 token 正在成为软件公司最大的单项外部供应商支出
一万亿美元 token 收入需要什么条件吃掉研发工资的 20%
Jason 的看空逻辑模型更高效后,token 成本可能没想象中高
当 token 支出从噪声变成裁员变量
Klaviyo 的 AI 执行力每个产品相关员工都必须提交代码
SaaS 公司还有时间吗创始人领导的公司可能仍能翻盘
Datadog、Figma 与 SaaS 从底部反弹
Rory 的“二十一岁”比喻SaaS 再也回不到 2021 年光环
Datadog 单季度十亿美元收入从亢奋估值回到现实估值
Figma 连续加速为什么 Jason 承认自己看漏了一点
Figma Make 不够好,但 Figma 仍受益于软件构建爆发
Lovable、Replit 与 Figma 工作流被绕开的风险
Figma 真正错过的机会一键从设计推到生产级 prototype
Wix 暴跌与 Base44 达到 1.5 亿 ARR
AI 业务是增量,还是对核心业务的替代?
Wix 和 Squarespace 被两头夹击vibe coding 与 Shopify
电商增长被 Shopify 吃掉,信息展示网站被 AI 工具冲击
Base44 能否拉动 Wix 重新增长?
股票回购争议为什么“抵消稀释”不是好理由
回购、激进股东与管理层的防守剧本
Wix 一倍收入估值终局状态还是反弹机会?
哪家老牌软件公司会靠 AI 重新变得乐观?
Nebius 增长 684%算力短缺还是泡沫狂热?
CoreWeave、Nebius 与算力商品化风险
除了传统软件,几乎所有科技基础设施都在火
Nvidia、电力、网络与 AI CapEx 的传导链条
为什么现在还看不到短期崩盘迹象
作为 VC,还会不会在 seed 投下一家新 CoreWeave?
投资人如何给创始人真实反馈
什么情况下应该解释拒投理由,什么情况下不该解释
Cerebras IPO 大获成功Snowflake 之后最大美国科技 IPO
“Better Than Figma”新 IPO 标准是否被抬高?
Cerebras 的独特位置半导体、inference 与 OpenAI 客户叙事
IPO 首日暴涨后还能买吗?Rory 提醒 base rate 很重要
SpaceX 计划史上最大 IPO1.75 万亿美元估值、750 亿美元融资
S1 可能讲不清真正的 SpaceXAI 化之后公司已完全不同
xAI、Cursor、Anthropic 交易可能不会完整体现在 S1 中
市场想要兴奋感,而 SpaceX 是地球上最令人兴奋的公司
散户会不会像 GameStop 一样冲进 SpaceX?
机构目标价、流通盘与 IPO 首日交易博弈
Jason 下注 SpaceX 可能交易到 5 万亿美元
Rory 的冷静提醒Facebook IPO 当年也是全民期待
三人打赌3 万亿、5 万亿,还是低于 3 万亿?
Sam Altman 给每家 YC 公司提供 200 万美元 OpenAI tokens
这是争夺开发者心智的聪明动作
Jason 判断这一定会推高 YC startup 估值
token 额度可能成为新融资锚点
这会不会让 VC 更难拿到 YC 公司股份?
对 Replit、Legora 这类公司,token 可能一年就烧完
token spend 就是 marketing spend:免费额度会改变早期增长策略
如果 OpenAI 有闲置算力,这就是非常聪明的资本配置
一个推论OpenAI 可能有富余 token,而 Anthropic 没有
OpenAI 与 Musk 诉讼结果Rory 认为判断基本命中
为什么陪审团快速驳回可能是按法律而非情绪判断
Elon 诉讼带来的连锁效应Sam Altman 财务关系被更多审视
Rory 对 Sam 的复杂同情善意结构反过来咬了他
Jason 反驳Sam 并非完全没有从 AI 中获得经济利益
复杂利益安排的代价当你把世界首富变成敌人
为什么美国公众可能越来越不喜欢 AI
Rory 的警告聪明科学家在政治上可能非常笨
Jason 为什么仍然站 Team Sam 和 Team OpenAI
从学生为 ChatGPT 鼓掌,到毕业典礼嘘 Eric Schmidt
“不是裁员,是为了机器减少岗位”企业话术的危险
Cisco、LinkedIn、Meta、Intuit 裁员与政治后果
Jason 的激进建议科技行业需要重新扩张招聘以避免社会动荡
AI 到底替代了多少岗位5%、20%,还是被夸大了?
被裁员工会如何看待财富税和科技富豪?
Jason 判断这可能是我们这辈子见过最严重的科技裁员潮
突发传闻OpenAI 可能最快周五提交 IPO 文件
OpenAI 如果启动保密提交,意味着钱站最后几班车正在开走
Show notes
#550. AI Capital Mania: Anthropic's $90B Valuation, SpaceX's Mega IPO, and the Life-or-Death Transformation of SaaS Veterans
📝 This episode’s podcast summary
This episode clones: the top-tier Silicon Valley venture capital podcast *20VC with Harry Stebbings* **Andrej Karpathy Joins Anthropic | SpaceX Files S1: How Does it Trade | Cerebras Smashes Day 1**
This is a high-density roundtable discussion on AI, SaaS, IPOs, and public market sentiment. Host Harry Stebbings and two regular guests, Rory O'Driscoll and Jason Lemkin, start by discussing Andrej Karpathy joining Anthropic and Anthropic’s $90 billion valuation financing, dissecting why AI model companies are entering a "balance sheet war." They further explore whether Salesforce spending $300 million annually on Anthropic tokens is expensive, how much AI token expenditure will eventually eat into R&D salaries, and whether these figures can support OpenAI and Anthropic’s trillion-dollar revenue expectations.
In the middle of the show, the topic shifts to public market SaaS companies such as Datadog, Figma, Atlassian, Klaviyo, Wix, and Squarespace—whether they still have time to complete their AI transformation. Rory presents a highly impactful judgment: SaaS companies have lost the "future glow" from 2021 and can now only be revalued based on revenue, growth, and cash flow. The second half dives into AI infrastructure and mega IPOs, discussing whether the market is chasing the most exciting companies on Earth. Finally, the three don’t shy away from political risks in the AI industry: the lawsuit between OpenAI and Elon Musk, Sam Altman’s conflict of interest, public backlash against AI, tech layoffs, and possible social backlash.
This is an information-packed, sharp-pointed, fast-paced roundtable on the tech and capital markets. It not only discusses which companies are rising and falling but more importantly, attempts to answer a bigger question: when AI becomes the most exciting narrative in capital markets, who is truly creating value, and who is just being carried up by the bubble?
👨⚕️ Guests
Rory O'Driscoll, Partner at Scale Venture Partners, has long focused on cloud software, SaaS, and AI investments. He is known for his deep analysis of public markets, valuation cycles, and enterprise software business models. In this episode, he repeatedly deconstructs the AI capital market from angles like ROI, capital allocation, and historical base rates.
Jason Lemkin, founder of SaaStr and a SaaS investor and commentator. He has long been active in the enterprise software ecosystem and excels at judging whether SaaS companies can really survive the AI shock from the perspectives of founders, operations, and product execution. His views on Figma, Klaviyo, Wix, SpaceX, and OpenAI in this episode are very direct.
Harry Stebbings, host of *20VC* and founder of the 20VC fund. In this episode, he continuously raises the hottest questions in capital markets, including Anthropic financing, Cerebras IPO, SpaceX S1, YC and OpenAI token deals, and the tech company layoffs trend.
⏱️ Timestamps
00:00 Opening & Podcast Intro
AI Giants' Balance Sheet War
01:35 Anthropic Opening: Karpathy Joining and $90 Billion Valuation Financing
03:10 Why Anthropic Might Be the Best Deal in Private Markets
06:49 Why Dario Is Willing to Raise at What Seems Like a Cheap Price
07:49 The Essence of AI Model Companies: Compute, CapEx, and the Balance Sheet War
08:39 Anthropic vs OpenAI Financing Styles: 70/30 Deals vs Extreme Pricing
10:37 Cash Receipts, Financing Terms, and Differences in OpenAI’s Complex Financing Structure
AI Token Economics
11:45 Is Salesforce Spending $300 Million Annually on Anthropic Tokens Too Expensive?
12:11 Each Engineer Spending $15K–$20K Annually: AI Coding Costs Are Just Entry Fees
12:56 Why Tokens Are Becoming the Largest Single External Vendor Spend for Software Companies
15:21 What It Takes for $1 Trillion in Token Revenue: Eating 20% of R&D Salaries
18:17 Jason’s Bearish Logic: As Models Get More Efficient, Token Costs May Not Be as High as Expected
22:11 When Token Spending Shifts from Noise to Layoff Variable
22:34 Klaviyo’s AI Execution: Every Product-Related Employee Must Submit Code
24:00 Do SaaS Companies Still Have Time? Founder-Led Companies May Still Turn Things Around
SaaS Veterans’ Revaluation
25:59 Datadog, Figma, and SaaS Rebounding from the Bottom
26:20 Rory’s “21 Years Old” Analogy: SaaS Can’t Return to Its 2021 Glow
28:18 Datadog’s $1 Billion Quarterly Revenue: From Exuberant Valuation to Realistic Valuation
29:11 Figma’s Continued Acceleration: Why Jason Admits He Missed Something
29:46 Figma Make Isn’t Good Enough, But Figma Still Benefits from the Software-Building Boom
32:23 Risks of Being Bypassed by Lovable, Replit, and Figma Workflows
33:14 Figma’s Real Missed Opportunity: One-Click Design-to-Production Prototype
Wix, Squarespace, and Website Business Under AI Impact
34:14 Wix Plunges and Base44 Reaches $150M ARR
34:35 Is AI Business Additive or Replacement for Core Business?
35:17 Wix and Squarespace Caught Between Two Fronts: Vibe Coding and Shopify
37:13 E-commerce Growth Eaten by Shopify, Information Display Sites Hit by AI Tools
37:41 Can Base44 Drive Wix’s New Growth?
40:03 Stock Buybacks Controversy: Why “Offsetting Dilution” Isn’t a Good Reason
41:19 Buybacks, Activist Shareholders, and Management’s Defensive Script
43:55 Wix’s 1x Revenue Valuation: Endgame or Rebound Opportunity?
46:43 Which Legacy Software Company Will Become Optimistic Again Thanks to AI?
Compute Infrastructure and CapEx Cycle
48:37 Nebius Growth 684%: Compute Shortage or Bubble Mania?
49:00 CoreWeave, Nebius, and the Risk of Compute Commoditization
50:52 Besides Traditional Software, Almost All Tech Infrastructure Is on Fire
52:41 Nvidia, Power, Network, and the Transmission Chain of AI CapEx
53:58 Why There Are No Signs of Short-Term Collapse Yet
55:40 As a VC, Would You Still Invest in a New CoreWeave at Seed Stage?
57:22 How Investors Give Founders Honest Feedback
59:36 When to Explain Rejection Reasons, and When Not to
IPO Window Reopens?
01:00:25 Cerebras IPO a Huge Success: Biggest US Tech IPO Since Snowflake
01:01:06 “Better Than Figma”: Has the New IPO Standard Been Raised Too High?
01:02:06 Cerebras’ Unique Position: Semiconductors, Inference, and OpenAI Customer Narrative
01:03:44 Can You Still Buy After IPO Day Gains? Rory Reminds Us That Base Rate Matters
SpaceX’s Mega-IPO and the Public Market Casino
01:04:41 SpaceX plans the largest IPO in history: $1.75 trillion valuation, $75 billion financing
01:05:00 The S-1 may not fully reveal the real SpaceX: the company has changed completely after AIization
01:06:20 xAI, Cursor, Anthropic deals may not be fully reflected in the S-1
01:07:30 The market wants excitement, and SpaceX is the most exciting company on Earth
01:08:25 Will retail investors rush into SpaceX like they did with GameStop?
01:10:28 Institutional target prices, float supply, and IPO opening day trading博弈
01:12:31 Jason bets that SpaceX could trade for $500 billion
01:14:26 Rory’s calm reminder: Facebook’s IPO was also universally anticipated
01:15:50 Three-way bet: $3 trillion, $5 trillion, or below $3 trillion?
OpenAI Grants Tokens to YC Companies
01:16:44 Sam Altman gives each YC company $2 million in OpenAI tokens
01:17:13 This is a smart move to win developer minds
01:18:38 Jason believes this will definitely boost YC startup valuations
01:19:58 Token allocation might become a new funding anchor point
01:20:37 Could this make it harder for VCs to get shares in YC companies?
01:21:31 For companies like Replit, Legora, tokens might burn through in a year
01:22:34 Token spending is marketing spending: free credits will change early-stage growth strategies
01:24:01 If OpenAI has idle compute power, this is very smart capital allocation
01:25:30 One inference: OpenAI likely has excess tokens, while Anthropic doesn’t
AI Litigation, Sam Altman, and Political Risk
01:27:10 OpenAI vs. Musk lawsuit outcome: Rory thinks his judgment was basically spot-on
01:28:25 Why the jury quickly dismissed it might be based on legal reasoning rather than emotion
01:31:19 Chain reaction from Elon’s lawsuit: Sam Altman’s financial ties are under more scrutiny
01:31:32 Rory’s complex sympathy for Sam: good intentions backfired on him
01:33:52 Jason counters: Sam hasn’t entirely benefited economically from AI
01:35:03 The cost of complex interest arrangements: when you turn the world’s richest man into an enemy
Public Backlash Against the AI Industry
01:36:37 Why Americans might increasingly dislike AI
01:37:00 Rory’s warning: smart scientists can be politically very naive
01:38:02 Why Jason still supports Team Sam and Team OpenAI
01:39:00 From students clapping for ChatGPT to graduates booing Eric Schmidt at graduation
01:39:48 “Not layoffs, but machine reduction of jobs”: the danger of corporate rhetoric
01:40:27 Cisco, LinkedIn, Meta, Intuit layoffs and political consequences
01:41:00 Jason’s radical suggestion: tech industry needs to rehire aggressively to avoid social unrest
01:41:59 How many jobs has AI actually replaced: 5%, 20%, or exaggerated?
01:43:38 How will laid-off employees view wealth taxes and tech billionaires?
01:44:44 Jason’s judgment: this might be the worst tech layoff wave we’ve seen in our lifetimes
01:45:09 Breaking news: OpenAI may file its IPO documents as early as Friday
01:45:45 If OpenAI starts confidential filing, the last few rounds of funding are about to end
🌟 Highlights
💰 Anthropic Is a Balance Sheet War
Rory believes that Anthropic’s fundraising isn’t about needing money—it’s about reducing risk for future massive compute commitments. Competition among AI model companies is no longer just about model capability; it's about having a strong balance sheet to convince hyperscalers to invest hundreds of billions in infrastructure.
“This is a very big balance sheet war.”
📊 AI token spending could determine whether Anthropic or OpenAI’s valuations succeed
One of the core models discussed is that if OpenAI and Anthropic want to support trillion-dollar token revenues, they must consume a large portion of enterprise R&D salaries. Rory estimates that to meet these revenue forecasts, they might need to take 20% of each engineer’s salary. If actual token spending stays at 1% or 5%, today’s valuations may be overly optimistic.
“You really have to start thinking about what percentage of total engineering payroll in software development markets they’ll take.”
🧊 SaaS Lost Its “Future Glow” from 2021
Rory describes SaaS in 2021 as being “twenty-one years old”: the market was willing to believe you had infinite potential. But once that glow fades, the market returns to pricing based on revenue, growth, and cash flow. Figma, Datadog, Atlassian can still be great companies, but they can’t return to the era of 50x ARR.
“Once you lose that glow, the market will price you for the rest of your life using some combination of revenue, revenue growth, and cash flow.”
🧩 Figma’s Missed Opportunity
Jason admits he underestimated Figma because even if Figma Make wasn’t a good vibe coding product, Figma still stood at the center of the “software building explosion.” But he also sharply points out that Figma missed the ability to instantly push designs into runnable production prototypes—something Replit and Lovable have already made native entry points for.
“Figma is about software building. So all companies truly helping others build software will benefit.”
🔥 SpaceX Is the Excitement the Market Craves Most
Regarding SpaceX’s largest-ever IPO, Rory’s judgment is clear: the S-1 may not fully reflect the post-AI SpaceX, but that won’t necessarily stop the market from embracing it. In today’s public market, what’s wanted isn’t cash flow—it’s excitement. SpaceX happens to be one of the most exciting companies on Earth.
“We’re selling the most exciting company on Earth when the market wants excitement.”
🧨 OpenAI Giving Tokens to YC Startups May Change Early-Stage Funding
Sam Altman giving each YC startup $2 million in OpenAI tokens is viewed by the three guests as a crucial move to win developer minds. Jason believes this will raise YC company valuations because it reduces early-stage risk and allows startups to treat tokens as marketing budgets, enabling more aggressive freemium products.
“Many startups, tokens are marketing.”
⚖️ The AI Industry Is Heading Toward Political Backlash
In the final segment, Rory makes a sharp social judgment: over the past few years, the AI industry has continuously told the public that AI will destroy jobs, increase electricity bills, and change the world—but without offering sufficiently transparent and mature political narratives. Now, as big companies begin laying off, public resentment could quickly turn into political pressure.
“We have some very smart scientists here, but they’re completely clueless politically; those who don’t understand AI but know politics well will eat us alive.”
🏭 Tech Companies Might Have to Rehire
Jason proposes a counterintuitive idea: perhaps the tech industry needs to expand hiring again after AI boosts efficiency, even establishing a new kind of social contract. Otherwise, mass white-collar job losses could trigger strong political backlash, ultimately harming the entire tech industry.
“Tech companies will eventually have to rehire thousands of people to prevent social unrest.”
🌐 Podcast Info Supplement
This podcast uses original voice tones for audio production, so some parts may sound odd.
AI translation is used, so there may be some awkward phrasings;
If you’d like to listen to other foreign-language podcasts in Chinese, feel free to contact WeChat: iEvenight