Overview of 20VC: Anthropic's Monster $6BN Revenue February | OpenAI Kills Sora and Hits $100M on Ads Business | Oura Going Public & Whoop Raises $500M at $10BN Valuation | Manus Founders Trapped in China
A 50th-episode discussion on the week’s biggest tech and AI stories with Harry Stebbings, Rory O’Driscoll and Jason Lemkin. Main topics: Anthropic’s leaked Mythos & enormous February revenue; OpenAI’s strategic retrenchment (Sora shutdown) and early ads traction; cybersecurity market reactions; accounting and ARR skepticism across AI companies; major consumer wearables funding and IPO chatter; SoftBank leverage to buy OpenAI stock; and geopolitical risk after Manus founders were unable to leave China post-acquisition by Meta.
Major news summaries
Anthropic
- Reported ~ $6 billion revenue in February (28-day month) — framed as an extraordinary monthly figure.
- Accidental leak of “Claude Mythos” assets: reportedly a ~10T-parameter model focused on cybersecurity; leak exposed ~3,000 unpublished assets.
- Irony/impact: Mythos is pitched as a cybersecurity tool but leaked via a security lapse. Leak caused a short-term sell-off in cybersecurity stocks.
Key takeaways:
- Mythos is claimed to be extremely powerful and costly to serve (so higher customer pricing).
- Leak highlights operational/security risks in fast-moving AI dev cycles — agents and automation may accelerate such mistakes.
OpenAI: Sora shutdown & ads
- OpenAI killed Sora (consumer image/video product) after limited revenue and high compute cost; viewed by hosts as likely the right strategic retreat.
- OpenAI’s ad business reportedly at ~$100M ARR — early stage but necessary for a consumer monetization path.
- Hosts argue OpenAI’s realistic long-term consumer options: either massively improved paid conversion or a huge ad business (targets cited: tens of billions to be meaningful vs. Google/Facebook).
Cybersecurity market reaction
- Cybersecurity equities fell after Mythos leak (CrowdStrike, Palo Alto, Zscaler, Tenable, etc. down 4–9%).
- Hosts mostly view the sell-off as knee-jerk / overbroad: agentic AI will likely increase security threats and overall create demand for new security products (a “golden age of security”).
- Some application-security and code-review players may be impacted; many vendors will need to integrate LLM/agent defenses.
SoftBank / Masayoshi Son
- SoftBank received a $40B bridge loan to buy more OpenAI stock. Hosts note the extreme leverage and risk exposure — SoftBank is highly levered (1.5–2x equity), so big downside risk exists with meaningful drawdowns.
Wearables & health: Oura & Whoop
- Whoop raised $500M at a $10B valuation.
- Oura (wearable sleep/health ring) is in IPO conversations — consumer health/wearables are growing, but hosts debate durability and degree of defensibility vs. enterprise/B2B software.
Manus founders / geopolitical risk
- After Meta’s acquisition of Manus, reports say founders were prevented from leaving China (government pushback on talent/brain drain).
- Implication: raises investor/founder caution about China-based teams doing exits to US buyers — political risk and founder mobility must be examined.
Accounting, ARR & product metrics — why to be skeptical
- OpenAI’s ARR method: trailing 4-week average × 13 (smooths weekly variability).
- Anthropic and others reported metrics raise questions: gross vs. net reporting (gross-through vs. partner-net); possible double/triple counting of token usage/resales across platforms.
- Emergent Labs controversy: claim of going “0→$100M ARR” in ~8 months questioned. Common dubious practices cited:
- Recognizing paid ARR from free trials or heavily discounted first-month offers (e.g., marking $0→$20/mo trial as $240 ARR immediately).
- Tranched funding and headline valuations being used for optics rather than reflective pricing basis.
- Hosts emphasize: until profitability and cash flow matter more, ARR games and inconsistent metrics can mislead investors and markets.
Product & go-to-market strategy debates
- OpenAI’s Sora: high compute + low revenue = unsustainable in a compute-constrained world; refocus on enterprise/coding or ads is pragmatic.
- Ads: $100M ARR is only a first step—hosts argue that to justify a big consumer valuation, ads need to scale to tens of billions annually.
- Consumer hardware/subscription businesses (Whoop/Oura): can build valuable businesses but have different dynamics than enterprise SaaS — higher product risk, potential churn and faddishness vs. long-term high-margin contract businesses.
Notable quotes and highlights
- “We may be at the stage where we throw the humans under the bus, not the AI anymore.” — on shifting blame for operational security errors.
- “Get the fuck over it.” — in defense of consumer products that aren’t enterprise SaaS; founders should accept different business models/metrics.
- “This is the golden age of security.” — agents increase threats and create new market opportunities.
Market & macro notes
- Layoffs and industry adjustments continue (Epic Games layoffs cited as an example of permanent structural shifts in entertainment/gaming employment).
- Wealth migration/taxation: ultra-high-net-worth mobility (example: Steve Jervison moving out of California) discussed in context of state tax policy and potential loss of tax base.
- Tranched rounds and valuation optics: hosts call out the gamification of startup valuations and the opacity/inequity of staged/dual-priced closes.
Practical advice / action items for founders & investors
- Branding: if building a tech startup, consider a .tech domain (Harry’s recommendation).
- Metrics transparency: be explicit about how ARR is calculated; avoid misleading recognition practices.
- Product-market fit over VC optics: “Conform your company around your customers and your model, not your VCs.”
- Security-first mindset: as agentic AI adoption rises, invest in security and prepare CISOs — demand for agent-security products will surge.
- Due diligence on geopolitical exposure: avoid structural surprises in cross-border company domiciles, and assess founder mobility/legal risk when dealing with China-linked teams.
- Beware valuation optics: don’t over-rely on headline valuation numbers driven by tranched rounds; consider blended basis and long-term implications.
Quick “what to watch” list
- Anthropic: official product/availability of Mythos and its pricing/usage model.
- OpenAI: evolution of ad revenue growth and enterprise coding products; any further consumer product pivots.
- Cybersecurity vendors: strategic M&A activity as incumbents acquire agent-defense startups.
- Manus/Meta: any diplomatic/contractual resolutions for trapped founders; implications for future China-related exits.
- Emergent Labs / other LLM-based startups: more scrutiny on ARR claims and billing practices.
Final takeaway
The week underscored two forces: (1) rapid technical progress and market disruption (Anthropic’s Mythos and agentic workflows), and (2) growing operational, accounting, and geopolitical complexity that demands sharper discipline — in metrics, security, and strategy. Founders should focus on customer-aligned monetization, transparent reporting, and security postures; investors should dig beneath headline numbers and account for non-market risks.