141. Freda的投资札记第2集:Tokenmaxxing、把电机塞进蒸汽机、接力赛变篮球赛、孤独、人的连接
141. Freda的投资札记第2集:Tokenmaxxing、把电机塞进蒸汽机、接力赛变篮球赛、孤独、人的连接
Podcast1 hr 23 min
Listen to Episode
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights

Investors should prioritize NVIDIA (NVDA) as big tech capital expenditure continues to be revised upward, providing high revenue visibility through long-term chip and power supply agreements. Be cautious with traditional SaaS and software companies that rely on per-seat subscriptions, as these "human-centric" models are being disrupted by usage-based AI "intelligence" platforms. Keep a close watch on Anthropic and OpenAI for potential mega-IPOs, which may trigger a significant rotation of liquidity out of existing Magnificent 7 stocks. Meta (META) offers potential margin growth through internal AI efficiency, but investors should monitor for "waste-driven pullbacks" if their massive compute spending doesn't yield immediate revenue. Look for emerging opportunities in Agentic Infrastructure and "New Labs" that focus on coding agents and "dollar-per-outcome" pricing rather than simple raw data processing.

Detailed Analysis

Based on the podcast interview with investor Freda, here are the investment insights and analysis regarding the AI sector and the broader technology market.


Anthropic (Private)

Current Status: Identified as one of the most significant model companies of the year with explosive revenue growth. • Competitive Advantage: Freda notes a potential "recursive self-improvement" loop where better AI (specifically coding agents) helps train the next generation of AI faster, potentially creating an insurmountable lead. • Efficiency: Despite having only ~3,000 employees, it generates revenue comparable to traditional software giants with tens of thousands of staff. • Product Success: The "Artifacts" (visualization) feature and "Claude Code" are driving significant user migration from competitors like OpenAI.

Takeaways

Revenue Quality: Anthropic’s revenue is shifting from "per-token" to "usage-based," which Freda believes is more sustainable. • Valuation Perspective: While private valuations are high, the "Time to Value" (TTV) is accelerating; the company reached $1B in revenue much faster than traditional SaaS benchmarks.


OpenAI (Private)

Market Position: Remains a top-tier player but is facing intense competition from Anthropic in the coding and reasoning space. • Structural Shift: Recently restructured to prioritize "Coding" and "Reasoning" (o1/Strawberry themes) as the core pillars of their development. • Revenue Model: Currently uses a mix of API fees and subscriptions, but Freda predicts a shift toward "value-based" or "outcome-based" pricing in the future.

Takeaways

IPO Outlook: Freda suggests the market has sufficient liquidity (approx. $1 trillion in cash held by mutual funds) to absorb a massive OpenAI IPO, though it may cause a "rotation" out of existing "Magnificent 7" stocks.


NVIDIA (NVDA)

Sector Dominance: Continues to benefit from the "Tokenmaxxing" trend where companies prioritize raw compute power to drive model performance. • New Growth Vector: Freda highlights NVIDIA’s NVLM (open-source models) and its push into autonomous driving as a "car operating system" (similar to Android for phones) as a major trend to watch.

Takeaways

CAPEX Trends: Capital expenditure from big tech (Microsoft, Google, Meta) is being revised upward, not downward, due to long-term supply agreements for chips and power. This provides a high level of visibility for NVIDIA’s near-term revenue.


Meta (META)

AI Strategy: Meta is spending billions on compute (Llama models) but faces questions on how to monetize this directly beyond advertising. • Internal Efficiency: Using AI for internal coding (Llama-based agents) could significantly boost margins by reducing the need for massive headcount growth.

Takeaways

Risk Factor: There is a risk of "waste-driven pullbacks." If Meta’s multi-billion dollar investment in tokens doesn't translate to immediate "end-customer" revenue, the stock may face temporary pressure.


Software & SaaS Sector (General)

Bearish Sentiment: Freda is cautious on traditional software. She argues that the "SaaS" model (per-seat subscription) is being disrupted by "Usage-based" AI models. • Valuation Gap: There is a disconnect between public markets (where software has dropped 50%+) and private markets (where valuations remain high). • Vulnerability: * High Risk: Companies whose value is primarily a "pretty UI" or simple "point solutions" (e.g., project management, electronic signatures). * Medium Risk: CRM and ERP systems that act as "Excel wrappers."

Takeaways

Investment Pivot: Investors are moving away from "Software" and toward "Intelligence" (Models) and "Infrastructure" (Data/Compute). • New Opportunity: "Agentic Infrastructure"—startups building tools specifically for AI agents (e.g., AgentMail, AgentPhone) rather than human users.


Investment Themes & Sectors

1. The "Tokenmaxxing" Era

Context: We are currently in a "wasteful" stage where companies compete on the volume of tokens processed. • Insight: This is a temporary phase. The industry will eventually move toward "Dollar per Outcome" rather than "Dollar per Token." Investors should look for companies that prioritize efficiency and reasoning over raw output volume.

2. Coding Agents as the "Singularity"

Context: Coding is the first sector where AI is truly "replacing" or "massively augmenting" human labor at scale. • Insight: The "Total Addressable Market" (TAM) for coding is being redefined. It’s no longer just about the 5 million developers; it’s about any digital task that can be translated into code.

3. "New Labs" (Neo-Labs)

Context: A new wave of startups founded by researchers from top labs (OpenAI, Google DeepMind). • Insight: Silicon Valley VCs are "indexing" this sector—investing in 4-5 different "New Labs" simultaneously because they cannot predict which one will become the next $100B company.

4. Market Risks

Social Impact: Large-scale white-collar layoffs could trigger government intervention or social backlash. • Liquidity Drain: Upcoming mega-IPOs (SpaceX, OpenAI, Anthropic) could suck liquidity out of existing tech giants, leading to a period of high volatility for the Nasdaq.


Actionable Summary for Investors

  • Watch the "Cloud-Model" Revenue: The most important indicator for the market right now is the revenue growth of Anthropic and OpenAI, as they are the primary drivers of the current AI bull run.
  • Monitor CAPEX: As long as Big Tech continues to sign "long-term supply agreements" for chips and power, the semiconductor cycle (NVIDIA, etc.) has a floor.
  • Be Wary of Traditional SaaS: Public software companies with high "human-centric" sales costs are at risk. Look for companies transitioning to AI-agent-friendly architectures (e.g., using WebSockets/Discord-style real-time connectivity over traditional HTTP).
Ask about this postAnswers are grounded in this post's content.
Episode Description
今天是我们的系列节目《Freda的投资札记》第2集。 可能有听众是第一次听我们的节目,那还是先介绍一下——Freda Duan在湾区做投资,是Altimeter Capital的合伙人。Altimeter是一个硅谷科技基金,横跨一二级。在一级市场投资案例有OpenAI、Anthropic、字节跳动等,在二级市场投资案例有英伟达、存储芯片、Snowflake、Robinhood等。 在《Freda的投资札记》第1集,也就是我们的125集节目中,Freda深入分析了当下美国的明星公司,给他们的巨额投入算了算账;也探讨了美国资本新秩序,以及当时大家很关心的泡沫。 可以预告一下的是,在本集,我们除了一如既往聊AI和投资这些很干的内容——如,Tokenmaxxing、AI时代的组织架构;还会聊一些感性的东西,尤其在这个充满不确定、每个人都很焦虑和孤独的时代,人和人的关系剩下什么? OUTLINE: 02:13 Token的世界 Token是一个非常容易误导人的单位 同样一个task,不同模型消耗的token可能差几十倍,甚至上百倍 Token消耗是越多越好吗?肯定不是越多越好,而且是反过来的 Tokenmaxxing 现在有大量浪费 - 是的;Token用量未来还会继续上升 - 也对 行业会逐渐从按token收费,转向按效果收费 11:44 模型公司的赛点与竞争 Coding Agent成熟之后,出现了一个很重要的loop:better AI makes better AI 现在到了无法追赶的地步了吗?用一个不太恰当的类比 - 马车换汽车 最早的汽车经常抛锚,跑在前面的几匹快马还能跟一跟、追一追,那时候两者差距没有那么清楚;但一旦汽车可以稳定跑起来,没有汽车引擎的马,再快也追不上 OpenAI按GPT和Coding两条主线重组,把Coding放到非常高的位置;Google是Sergey直接管Coding;Meta内部也开始推自己的Coding model,目标是年底前做出SOTA;xAI公开收购Cursor 27:41 投资人是在抛弃软件公司吗? Anthropic Cowork这么成功的产品,是两个人做出来的,AI来了以后,值得重新思考“一个公司到底应该需要多少人” 投资人是在抛弃软件公司吗?对,是。最简单的投资逻辑是discount rate 哪些软件更脆弱?AI时代新的软件机会在哪里?软件是不是要为Agent重新设计? 36:32 组织架构变革 Dario提过两个概念:technology diffusion和economic diffusion AI现在在曲线的哪个位置?在“把电机塞进蒸汽机位置”这一阶段——每个人都在把AI加进自己的工作流,但很少有人退一步问:这个流程本身、这家公司本身,为什么长成现在这样?为什么需要这么多层级? 接力赛变篮球赛——过去是接力赛,一棒接一棒,每一棒都是一次翻译、一次等待;未来更像篮球赛,3到5个人的小分队,必要技能都在队里,同步推进,自己拍板,只有重大方向才往上升级 44:21 AI对投资行业的影响 投资行业是一个非常低效的行业,大量时间花在找信息、清洗数据、比较预期、判断positioning 我这几个月想的问题是:如果给Agent足够干净的数据和足够清晰的交易目标,它有没有可能做的比我好? AI时代会让更多人参与股市,散户行为会更复杂,也更值得tracking 投资本质上会越来越像一个解构市场玩家的游戏 50:08 应用创业公司还剩什么? AI这波创业公司,哪些方向已经跑出收入?一波快问快答 这几年经历了一个过程 - 23年大家特别担心模型公司吃一切,wrapper大家都不敢碰;过去两年大家不担心了,因为应用或者叫wrappers收入增长很快;但到了今年我现在又重新开始担心了 Anthropic先吃编程这一块,现在第二大业务吃金融领域;OpenAI马上自己推出音频模型,对ElevenLabs会是什么影响?OpenAI和Anthropic都和大的私募基金合作,把自己模型推到私募基金的被投企业里面,这个决心是很大的,真的adoption还没有开始 2026年了,你还投应用创业公司吗? 硅谷最火的是Neo Labs,另一个很重要方向是给Agent用的基础设施 01:00:42 大票和资本市场 01:08:33 AI会带来裁员和通缩吗? 01:14:17 焦虑、孤独与人的连接 LINKS: 我们的播客在小宇宙、Apple Podcast、Spotify等全音频平台播出; 我们的视频播客在Bilibili、小红书、视频号、抖音等全视频平台播出; 如果你想服用文字版,请搜索我们工作室的公众号:语言即世界language is world。 DISCLAIMER: 本内容不作为投资建议。 CONTACT: xiaojunzhang@lisw.ai Jump into the new world-and explore with us!😉
About 张小珺Jùn|商业访谈录
张小珺Jùn|商业访谈录

张小珺Jùn|商业访谈录

By 张小珺

努力做中国最优质的科技、商业访谈。 张小珺:财经作者,写作中国商业深度报道,范围包括AI、科技巨头、风险投资和知名人物,也是播客《张小珺Jùn | 商业访谈录》制作人。 如果我的访谈能陪你走一段孤独的未知的路,也许有一天可以离目的地更近一点,我就很温暖:)