What Sequoia Is Thinking

Synthesized from 20 recent items.

AI & agents

AGI arrived: baseline knowledge (2022) + reasoning (o1, 2024) + iteration (agents, 2026) = done. 2026: This is AGI
Agents run 30 minutes unassisted; capability doubles every 7 months (METR). Day-long tasks by 2028. 2026: This is AGI
Shift from talkers (2023–24 chatbots) to doers (2026–27 agents) is the defining phase transition. 2026: This is AGI
Two killer apps — coding tools and ChatGPT — each approaching $10B revenue; $0→$1B club forming. AI in 2026: A Tale of Two AIs
LangGraph already in production at Uber, Klarna, LinkedIn, JPMorgan — agent infra is post-experimental. LangChain: From Agent 0-to-1 to Agentic Engineering
Specialist agents live: Harvey (law), OpenEvidence (medicine), XBOW (security), Juicebox (recruiting), Ricursive (chips). 2026: This is AGI
Chip design: AI cuts 18–36 months to hours; human labor is 50–70% of a $600M design cost. Partnering with Ricursive Intelligence
Voice is the next primary interface; Sesame bets emotional intelligence beats utility-focused assistants. Partnering with Sesame

For every $1 on software, $6 goes to services — the next trillion-dollar company captures the $6. Services: The New Software
AI autopilots (WithCoverage, Crosby) beat copilots (Harvey, Rogo) on value capture — outcomes, not tools. Services: The New Software
Labor TAM: insurance brokerage $140–200B, IT managed services $100B+, consulting $300–400B. Services: The New Software
Wedge into outsourced, intelligence-heavy work first — existing budgets, tolerance for external delivery. Services: The New Software
Enterprise DIY fatigue is accelerating; incumbents losing ground to AI-native startups. AI in 2026: A Tale of Two AIs
Organizational hierarchy — unchanged 2,000 years — finally obsolete; Block replacing managers with world models. From Hierarchy to Intelligence
Moat requires per-customer economic graphs (Block: Cash App + Square); without it, AI is cost optimization only. From Hierarchy to Intelligence

Best founders play opening, midgame, and endgame simultaneously — sequential play is the failure mode. The Opening, Midgame and Endgame in Startups
Bet founders before products: Sequoia backed Khimji as a partner in 2019, years before BlockIt existed. The Most Precious Resource
Top AI startups in 2025: >$1M revenue per employee. Market pull, not push. AI in 2026: A Tale of Two AIs
AI startups using AI for legal, recruiting, and sales build a self-improving flywheel incumbents can’t match. AI in 2026: A Tale of Two AIs
Bootstrapped to $1M ARR with 4 people signals product-market fit before raising (Juicebox). Why We’re Partnering with Juicebox

Data center supply chain can’t match hyperscaler CapEx; TSMC revenues +50% since 2022, CapEx only +10%. AI in 2026: A Tale of Two AIs
Abundance of AI content compounds scarcity premium on trusted IP — Netflix’s $82B Warner bet is the template. The Abundance Paradox: Why Netflix Paid $82B for Scarcity
USDC weekly volume: $5B (2022) → $23B (2024) — stablecoin rails replacing SWIFT for global remittances. Aspora: Cross-Border Banking Takes Center Stage
Physical AI addresses 80% of global GDP; hardware is the next major frontier after software-only AI. Building Tomorrow’s Transformational Companies

Last updated: 2026-04-15 15:01 UTC