20VC OGs: SpaceX Valued at $800BN & Harvey Raises $160M at an $8BN Price | Airwallex Raises $330M and The Battle with Keith Rabois | Netflix Acquires Warner Brothers | IPO Market Predictions for 2026: Anthropic, Stripe, Databricks and SpaceX

Analysis of SpaceX pursuing $800B valuation through secondary sale at 40x revenue for a company doing $15B growing 30%. Discussion covers whether this represents Elon magic premium versus fundamental value, comparison to typical SaaS multiples, and implications for the broader late-stage market pricing discipline.

• •SpaceX at $800B represents 40x run rate revenue for 30% growth - significantly higher than typical multiples
• •Secondary sales at high valuations create emotional barriers to IPO if public markets value lower
• •Late-stage investors accepting valuation risk as core business model - some will be wrong on price
• •2% dilution rounds ($160M on $8B) may not represent true market clearing prices

Predictions for blockbuster 2026 IPO year with SpaceX ($800B), Anthropic ($400B), and Databricks ($200B) potentially returning $700B to VCs. Discussion of whether private market valuations will hold in public markets and historical patterns of down-round IPOs in 2025.

• •Potential $1.4T in combined market cap from three IPOs would return ~$700B to venture capital
• •Every 2025 IPO was down-round on prior private valuation - private markets consistently overpriced
• •Power law math: one or two mega-IPOs dwarf dozens of smaller exits (like Facebook 2012, Alibaba 2018)
• •Risk: If markets normalize, these could IPO at 50% of private valuations and still be good outcomes

Netflix's $82.7B bid for Warner Brothers shows venture-backed companies consuming traditional industries. Netflix ($470B market cap) vs Paramount ($20B) demonstrates massive competitive advantage. Analysis covers regulatory hurdles, Hollywood creator concerns about monopsony power, and broader trend of digitization eating advertising, entertainment, retail.

• •Netflix can acquire WB with <20% dilution; Paramount needs debt + PE to compete - massive imbalance
• •Hollywood creators fear Netflix monopsony power - not consumer harm but supplier leverage
• •Digital platforms have now consumed advertising (Google/Facebook), entertainment (Netflix), retail (Amazon) - fintech and auto likely next
• •Regulatory challenge: narrow definition (streaming) shows monopoly, broad definition (entertainment) shows small player

Tiger raising $2.2B fund (down from peak) with only 9 deals in 2024 vs 100+ in 2021. 20% GP commit ($400M+) shows conviction. Discussion of adapting strategy to market conditions and taking responsibility for 2021 over-exuberance.

• •20% GP commit ($400M+) is exceptionally high - shows 'money is truth serum' and real conviction
• •Shift from spray-and-pray (100+ deals) to concentrated (9 deals) reflects market reality
• •2021 strategy was right for that moment (IPO-a-day market), 2024 strategy right for concentration on winners
• •At $400M personal capital, 20% GP commit means you're in because you love it, not for marginal returns

Harvey raises $160M at $8B (2% dilution) with $150M ARR growing 300% YoY, 98% GDR, 168% NDR. Debate over whether legal TAM can support valuation (no legal software company ever worth >$2B) and whether growth rates will sustain or decelerate rapidly.

• •Math: If growth goes 10x→3x→2x→80%, you're paying ~3 years ahead at public market multiples
• •Legal TAM expansion required: must dominate market AND expand beyond legal into professional services
• •Competition from Lagor (10x growth on smaller base, winning Europe) shows market fragmentation risk
• •Key assumption: AI companies maintain SaaS-like deceleration curves (85% of prior year) vs faster collapse

Debate on whether improving foundation models will obsolete application layer companies. Discussion of deep reasoning moving from 15 minutes to seconds, model switching/commoditization, and whether GTM/implementation provides defensible moat.

• •Current AI apps built on slow reasoning (seconds) - when deep reasoning hits milliseconds, could enable new generation of apps
• •Model switching is real: companies rotate between Anthropic/OpenAI based on quality, unlike cloud lock-in
• •Counter-argument: Implementation, workflow integration, customer relationships provide stickiness beyond model quality
• •Risk: We haven't seen deep reasoning in B2B apps yet - when 15min→1sec, current apps may seem 'quaint'

Analysis of Benioff's 'LLMs are commodities' comment. Discussion of whether API-level commoditization means bad business (like hard drives) or consolidation to profitable oligopoly. Consumer memory creates stickiness; B2B switching remains easy.

• •Hard disk analogy: 30 commodity providers→3 survivors who extract profits through consolidation
• •Consumer side: ChatGPT memory creates real lock-in, losing it would leave 'significant gap'
• •B2B side: Easy API switching means LLM providers must own apps (Anthropic→Claude, OpenAI→ChatGPT)
• •High fixed costs + easy switching = airline economics, but consolidation makes it viable business

Discussion of Andreessen portfolio using Chinese open source models (clarified: 20% use open source, 80% of that is Chinese). Debate over national security concerns, cost benefits, and whether inference costs negate model savings.

• •Cursor and other hot startups use Chinese open source models 30-40% of time - 'not that inferior'
• •Cost savings unclear: inference on AWS costs similar whether using Chinese model or paid API
• •National security: Government should set clear guidance; if legal, it's just business decision
• •Demand for cheap open source exists because Meta/others stopped providing state-of-art free models

Airwallex raises $330M at $8B doing $1B ARR - same revenue as Ramp but 1/4 the valuation. Debate sparked by Keith Rabois raising China data concerns. Discussion of what constitutes 'Asia discount' vs legitimate regulatory risk for companies with Chinese engineering presence.

• •Airwallex at 8x revenue vs Ramp at 32x - 75% discount despite similar metrics represents 'Asia discount'
• •Not a Chinese company (Australian HQ) but significant Chinese engineering presence creates regulatory exposure
• •Zoom, Microsoft, Apple have more Chinese engineers - question of consistent standards vs selective targeting
• •Sales objection added: Must make data sovereignty 'unimpeachable' to remove friction, likely precludes government contracts

Discussion of prediction markets, insider trading concerns, and whether VCs should participate. Debate over information advantages, market manipulation risks, and comparison to public market insider trading rules.

• •VCs have material non-public information on portfolio companies - creates unfair advantage in prediction markets
• •Counter: Prediction markets are small, self-correcting, and VCs betting would actually improve accuracy
• •Comparison to sports betting: coaches/players banned despite having legitimate expertise
• •Practical issue: Easy to manipulate small markets, and appearance of impropriety damages VC reputation