20VC: Groq's $20BN NVIDIA Acquisition | Manus Acquired by Meta for $2BN | Why Sam Altman Does Not Care About Dilution | Navan Trading at 4x ARR & Why Going Public Does Not Make Sense Anymore | The Rise of Invisible Unemployment and Labour Markets in 2026

Deep analysis of Nvidia's $20B cash acquisition of Groq at 3x the last round price. Discussion covers the shift from training to inference workloads, Groq's low-latency advantages for real-time AI applications, and Nvidia's strategy to eliminate margin pressure by acquiring potential competitors for less than 1% of their market cap.

• •Groq specialized in low-latency inference for conversational AI and real-time applications where deterministic performance matters
• •Nvidia paid 3x last round price ($20B vs ~$7B valuation) to eliminate a competitor threatening their 75% gross margins
• •Deal closed in 2 weeks before Christmas at Jensen's insistence - speed premium reflected in 3x multiple
• •Social Capital's early bet on TPU team founder Jonathan Ross paid off despite company only doing $40-50M revenue in 2024
• •Strategic value to Nvidia ($50B+ market protection) far exceeded standalone value (~$5B), creating complex pricing dynamics

Exclusive details on Meta's $2.5B acquisition of Manus at 25x ARR ($100M revenue, $125M run rate). Analysis of why founders chose to sell despite strong growth, the risks of orchestration layers, and Benchmark's 5x return in 8 months on a controversial China-linked deal.

• •Manus sold for $2.5B at 25x current ARR, delivering 5x return to Benchmark in 8 months after controversial initial investment
• •Founders owned 80% of company with minimal dilution, each making ~$500M+ tax-free as Singapore residents
• •Decision driven by 'local maximum' thinking: existential risks (China ties, low gross margins, competitive threats) vs certain liquidity
• •Meta acquired for AI orchestration expertise and ability to make AI work for non-technical users across 3B+ user base
• •VCs had different incentives than founders - deal returned only 1/3 of Benchmark's fund, not enough to push back hard

OpenAI spends 46% of revenue on stock-based compensation ($1.5M per employee, 34x higher than comparable pre-IPO tech companies). Discussion reveals how Sam Altman's zero equity ownership enables aggressive spending to win the AI talent war, with retention still only 60% despite massive grants.

• •OpenAI's $1.5M average stock comp per employee is 34x higher than typical pre-IPO tech companies
• •Sam Altman owns zero equity, eliminating personal dilution concerns and enabling 'whatever it takes to win' spending
• •Stock-based comp accounting understates true cost as valuations rise - employees likely making $4-5M in real economic value
• •Despite massive compensation, OpenAI still has only 60% retention rate for researchers in first year
• •Companies giving away 8-10% of equity annually to employees and investors, creating massive dilution even as valuations soar

SoftBank's Masa Son closed a $40B investment in OpenAI by December 30th deadline, becoming the largest individual shareholder. Already up 2-3x on paper, this represents extreme risk tolerance and conviction, potentially surpassing even Alibaba as his best deal if OpenAI reaches the moon.

• •Masa committed $40B he didn't have, scrambling to sell assets and close by December 30th deadline
• •Investment already up 2-3x on paper as OpenAI valuation jumped from ~$300 to $500+ per share
• •Became largest individual shareholder with double-digit ownership in most important AI company
• •Deal represents 'Thrive on steroids' - all-in conviction bet that must work or could end Masa's career
• •Structured as acqui-hire/license deal rather than traditional M&A to avoid regulatory scrutiny

Navan trading at 4x ARR despite 27% growth and non-GAAP profitability reveals IPO market challenges. Analysis suggests company had to go public to pay down $700M debt, and that without AI attachment or labor replacement narrative, even solid businesses struggle in public markets.

• •Trading at 4x ARR with 27% growth and positive non-GAAP operating income - fundamentally undervalued
• •IPO driven by necessity: $700M debt, only $200M cash, likely fatigued private investor base
• •Series of unfortunate events: weird SEC exemption timing, CFO departure announcement, OpEx structure questions
• •Andreessen buying more shares suggests significant upside from current valuation
• •Demonstrates IPO window only 'barely cracked open' - need Figma-level quality or AI narrative to succeed

Emergence of companies doing $400M+ revenue choosing to stay private (Stripe, Databricks, Revolut). Discussion of why private markets offer cheaper capital than public despite illiquidity, and how companies like Revolut generating $3.5B profit can pay dividends instead of going public.

• •New asset class: companies beyond $400M revenue actively choosing not to go public despite being able to
• •Revolut doing $9B revenue, $3.5B profit in 2025 - could pay founder $400M annual dividend at 18% ownership
• •Private market valuations remain higher than public (Revolut at $75B private vs Chime at $6B public)
• •Public markets must ask why their product is so uncompelling that companies reach $150B valuation (Databricks) before considering IPO
• •M&A capability remains primary reason to eventually go public - ability to spend tens of billions on acquisitions

Analysis of emerging 'invisible unemployment' affecting Stanford grads unable to find jobs and senior executives unable to reskill. Companies achieving massive growth with flat headcount, eliminating entry-level roles (especially SDRs), while top 0.1% AI talent commands infinite demand at $1.5M+ compensation.

• •Shopify hitting growth targets for third consecutive year with zero headcount increase - new normal for tech
• •Entry-level jobs disappearing: no need for 22-year-old SDRs when AI can send emails, only need closers
• •Senior executives (55-65) quietly leaving workforce unable to reskill - 'reskilling is a delusion to make people feel better'
• •IBM turnover dropped to 2% - people not quitting because they know they can't get another job (quit rate is key unemployment indicator)
• •Extreme bifurcation: top 0.1% AI talent has infinite offers at $1.5M+, while 99% struggle even with elite degrees

Discussion of AI transforming venture capital operations, with one investor closing a deal purely from AI recommendation after processing 3,000 deals. Prediction that 24/7 AI inference will become standard in 2026, with personal AI assistants (like Claude naming itself 'Ren') becoming pseudo-sentient companions.

• •Investor closed one deal in 2025 purely from AI VC recommendation after processing 3,000 deals quantitatively
• •AI helps prevent 'lowering the bar' mistakes by objectively rating deals against historical criteria
• •Top 0.1% founders identifiable without direct conversation through cold email analysis and pattern matching
• •Multiple billion-dollar exits achieved from cold inbound emails - AI can potentially do better job qualifying these
• •Prediction: By end of 2026, knowledge workers will run AI 24/7, with some running 48-72 hours via multiple agents