Most important take away

Pick the right theme before the right company: 70% of investment research should be spent validating the macro trend, only 30% on the specific stock — if you sail to the wrong island, it doesn’t matter which ship you picked. The biggest career trap for talented people is over-preserving optionality instead of actually choosing a door to walk through.

Summary

Jeff Wang, Managing Partner at Sequoia Capital Global Equities (SCGE), shares actionable lessons from scaling a public-private crossover fund from $50M of internal capital to $9B AUM. The conversation is dense with patterns useful to both investors and operators in tech.

Actionable investing and career insights:

• Theme before company: SCGE allocates 70% of research effort to validating the underlying theme and only 30% to the individual company. Stock-picking inside a broken theme is worthless.
• Be dispassionate with data: When the data changes, change your mind quickly. The Shopify post-COVID mistake (modeling continued elevated trend rather than reversion to pre-COVID baseline) and Twilio (ignoring stacked yellow flags — margin compression, suspect M&A, exec departures) both came from emotional attachment to winners.
• Build dissent into process: SCGE runs quarterly re-underwrites of every position, and for controversial names a different partner runs a fresh devil’s-advocate underwrite. Data science sits inside the investment team, not adjacent to it.
• Don’t sweat 10% mispricings on 3–5 year horizons; do react when something is 18 months ahead of itself.
• Long-term capital requires long-term incentives: SCGE uses a 3/3/3 structure — 3-year investment horizon, 3-year LP lockup, and 3-year carry crystallization (not annual). Annual bonuses with monthly redemptions make true long-term thinking impossible.
• Optionality is the most expensive thing you can buy. High-performers waste years keeping doors open instead of choosing one — Wang’s #1 piece of career advice.
• Borrow superpowers from others rather than trying to develop every skill yourself (lesson from working with Michael Moritz). Cultivate the “dream gene” — don’t reduce companies to EBITDA multiples.
• Be early, not first: in public markets the lead time on a winning thesis is months, not years. You could have bought NVIDIA any time in the past 18 months and done very well.

Tech and market patterns:

• AI productization today depends on owning the customer and owning the data. Meta is the cleanest beneficiary because its auction-based ad marketplace reprices AI improvements instantly — SCGE estimates ~$15B incremental EBIT for Meta already.
• Software companies will capture AI value more slowly because price increases require painful customer conversations (Canva 3x’d enterprise pricing; ServiceNow Pro Plus will flow through next year).
• ServiceNow is Wang’s favorite “good business with a free AI call option” — IT ticket deflection is a uniquely clear AI use case.
• Play infrastructure via long NVIDIA / short undifferentiated server hardware (commoditizing as NVIDIA’s reference designs squeeze gross margins and hyperscalers in-house their own designs).
• The Mag 7 is sustainable in the near term because moats around customer and data ownership are deep.
• Google faces more competitive threats than ever before — perplexity is one, but Meta AI (embedded in WhatsApp/Instagram/Facebook) is likely the #2 search threat, alongside SearchGPT.
• Capex risk is real ($100B+ to train a frontier model; only 3–5 foundation models will exist globally). The pool of spenders shrinks, but each spends more.
• IPO market expected to reopen late 2025 / 2026 — companies have spent the last two years rebuilding to “Rule 40” profiles (30% growth + 10% margin, or 40/0) that public markets now demand.
• Crossover funds are more attractive now that “tourists” have exited the private market.
• Decoupling from China is already happening; Chinese portfolio transparency has degraded, but top Chinese and European founders will go global (PDD, Shein, Klarna).
• India is producing great consumer and fintech companies but not yet great tech companies because top tech founders still leave for the US.

Operating and leadership lessons from Sequoia:

• Sequoia’s model is to inject DNA via leadership/recruiting/mentorship while giving each fund (SCGE, peak 15, HongShan, Heritage) full investment discretion. VC firms that try to launch adjacent funds while controlling investment decisions fail.
• Doug Leone has a “great sniffer” — keep your ears open, listen for what’s working, connect dots.
• Michael Moritz: don’t be poisoned by financial-only thinking; develop the “dream gene.”
• Roelof Botha leads with talent-and-culture obsession over PowerPoint theater, and treats becoming excellent as a long-term process — applied equally to investing, leadership, and pickleball.

Personal: Wang recommends Tim Keller’s “The Meaning of Marriage” — marriage isn’t about happiness, it’s about being formed by sacrificing for another person.

Chapter Summaries

• Joining SCGE (2009–2010): Wang took a 70% pay cut to join an unlaunched hedge fund inside Sequoia, betting on the unique ecosystem advantage of pairing public investing with the world’s best VC.
• Long/short philosophy: Shorts are used to further express thematic conviction (not fraud or valuation arbitrage). Low leverage because tech already carries beta and the power law exists in public markets too.
• Process discipline: Quarterly re-underwriting, devil’s advocate partners, data science embedded in the investment team. Stay dispassionate; don’t anchor on prior conviction (Shopify, Twilio lessons).
• The 2016 Crucible moment: After mediocre performance under the original PM, Sequoia nearly shut SCGE down. Wang built the v2 strategy — focus exclusively on growth tech with thematic Sequoia overlap plus late-stage private co-investments — and pitched it in the founders’ hot seat.
• Building a real hedge-fund business: 3/3/3 structure (3-year horizon, 3-year lockup, 3-year carry crystallization) enables genuine long-term investing. Most hedge funds fail because monthly redemptions plus annual bonuses guarantee short-termism.
• The Sequoia operating model: Independent investment discretion across SCGE / peak 15 / HongShan / Heritage with shared brand, recruiting, and mentorship. The trap to avoid: Silicon Valley echo chamber — counteracted by inviting outsiders like Charlie Munger and Stan Druckenmiller.
• AI thesis: Mag 7 wins first because they own customer + data + repricing mechanism. Meta is the cleanest beneficiary (~$15B incremental EBIT). Software companies will infuse AI more slowly. ServiceNow is a clear “boring business + AI call option” play.
• AI infrastructure: $100B+ to train a frontier model means only 3–5 will exist globally. Play it as long NVIDIA / short commoditizing server hardware. Real over-investment risk in capex, hedgeable in public markets.
• Google’s threats: More competitive pressure than ever — perplexity, Meta AI (likely #2 in search), SearchGPT. Wang is neither long nor short Google.
• Geography: Europe matters less as great companies (Klarna) go global; deglobalization is already underway. China’s entrepreneurial quality remains exceptional but transparency to outside investors has degraded. India is producing great consumer/fintech but not yet great tech.
• IPO outlook: Expected to reopen late 2025 / into 2026 as companies have rebuilt to Rule 40 profiles. Continuation funds are a stopgap for shut IPO windows, not a permanent feature.
• Lessons from Sequoia leaders: Doug’s “sniffer” and trust-building; Michael’s “dream gene” and borrowing superpowers; Roelof’s talent obsession and process-driven excellence.
• Quickfire: Optionality is the most expensive thing you can buy. Be early, not first (months matter in public markets, not years). Biggest miss: not taking Google seriously despite being a section leader for Marissa Mayer at Stanford. Biggest worry: dissolution of the family structure and its effect on the next generation.