Most important take away

We are in an AI bubble that is arguably frothier than 2021, with seed rounds priced at tens or even hundreds of millions, and AI revenue multiples at 100-200x while the best public software companies trade at 15-20x forward revenue. The only thing that ultimately matters is backing iconic founders in massive, secular markets — price and ARR milestones are far less important than founder quality and market gravity, because power-law outcomes dwarf any rational pricing discipline applied to the wrong company.

Summary

Saam Motamedi (Greylock GP) walks Harry Stebbings through the current venture landscape with a series of actionable, often contrarian insights. Below are the most useful takeaways, with emphasis on career advice and tech/investing patterns.

Bubble dynamics and pattern recognition

• AI is in a clear bubble. Seed rounds for newly-incorporated companies are routinely priced at $20M-$100M+ post; revenue-stage AI companies are pricing at 100-200x revenue while the best public comps (CrowdStrike) trade ~15-20x forward revenue. That dislocation only resolves if growth durability is unprecedented — Saam doubts it for most.
• Corporate/strategic investors (Amazon, Google, MSFT) distort pricing because they optimize for cloud spend/partnership, not financial returns.
• Pattern: many fast-growing AI prosumer apps spike then fade — run Google Trends on them. Strong early revenue is not the same as durable retention.

How to evaluate AI applications (tech pattern)

• Foundation model providers will ruthlessly own “foundational primitives”: pure text generation, raw code generation, basic chat.
• Big startup opportunity is in focused applications layered on top — vertical copilots (legal, medical), SRE/incident response, customer service AI, vertical workflow tools. These can be very large.
• Where the model and app must be tightly coupled: personal agents, horizontal enterprise agents, and (debatable) code generation.
• Pricing is shifting from pure per-seat to hybrid (seats + outcome/usage). Real pricing power only arrives when AI replaces a unit of work (e.g., AI BDR replacing a human BDR), not when it acts as a feature layered onto existing tools.

Why now is a great time to build SaaS

• Largest SaaS outcomes (Salesforce, Workday, ServiceNow, HubSpot, SAP) require disrupting one of: data model, delivery model, or interface. For ~15 years none of those were disruptable. Generative AI now disrupts all three:
  • Data model: agents can synthesize views from raw inboxes/call recordings without users filling a CRM schema.
  • Delivery/pricing model: outcome-based pricing that incumbents can’t adopt without cannibalizing.
  • Interface: dynamic, generated UIs and agents replace clunky enterprise UX.
• That opens the door to building the next Salesforce-scale company.

Seed and Series A pricing (actionable for founders and emerging investors)

• Bread-and-butter seed for strong teams: $20-$40M post, often pre-product.
• Series A with light product-market fit: $80-$200M post.
• Don’t pass on a great founder over a 20-vs-30 post-money debate — the only outcomes where that matters are mediocre ones; iconic outcomes make entry price irrelevant.
• BUT: at portfolio level you must enforce discipline, because if every deal is “the exception” you’ve blown construction. Greylock prioritizes ownership (target 20-25%+ on core early positions) over price within a range.
• Large seed rounds (e.g., a $26M seed for a repeat founding team) can actually be lower-risk than a $2M seed with an unproven team — the risk you’re paid for is execution risk, not just dilution.

ARR is a terrible Series A filter

• Most companies that hit $1-2M ARR never reach $10M, let alone $100M+. Using ARR as your primary screen filters for survivors of small markets, not iconic outcomes.
• Better lens: who is the founder, and is the market large/secular? Wiz raised its Series A at a $500M valuation with $0 ARR — and was right.

Founder vs. market trade-off

• Markets have gravity. Even iconic founders rarely overcome a market that’s too small once the plane is built.
• Heuristic: good zip code, bad street = back the founder; wrong zip code = pass.
• Saam’s biggest mistakes have been passing on iconic founders working on bad early ideas — they almost always pivot, and you should have been on the cap table.

Series B vintage warning

• Despite SoftBank/Coatue/D1 pulling back, large early-stage platforms with growth pools have filled the gap. Pricing remains exuberant. Saam thinks the Series B asset class broadly may not make money this vintage unless public multiples re-expand or growth durability surprises.

Signaling risk is “total bullshit”

• Founders raise the next round on the company’s merits, not on whether the seed lead led the A. Greylock’s seed companies almost always raise As, including from other top firms.

Reserves

• Reserves aren’t just for offense (winners). They’re for air pockets — Greylock has done many defensive inside rounds for companies it believed in.

Career advice for young VCs (most actionable section)

• Big mistake: junior investors think their job is to source deals and stamp their name on them. That’s actually optimizing for stepping-stone exits, not for scaling within a firm.
• To scale at a top firm: be judicious, have a clear answer to “why would a founder want to work with you?”
• Develop a lane. Spend a year mapping a sector (e.g., fintech: meet PMs/eng leads at Stripe, Square, Klarna) so you meet the next great founder six months before anyone else. Sourcing edge compounds into winning edge.
• Sourcing is the single most important skill; servicing is the least, but you need at least two of the three (sourcing, picking, servicing).
• When someone great (e.g., Elad Gil) offers to bring you into a round, don’t ask twice. Saam passed on Vanta at seed because “SOC 2, what is that?” — a lesson he won’t repeat.

Other tactical notes

• Investors can help pre-PMF: tightening ICP, sitting in customer calls, narrowing positioning — PMF isn’t pure magic.
• Incubations broadly don’t work (negative selection, too much cap table), but the exceptions are massive (Palo Alto Networks, Workday, Abnormal Security).
• Saam updated his view: focused/specialty models (e.g., ElevenLabs for voice) can win by quickly building application/workflow depth on top, even as foundation models commoditize the underlying capability.
• Market size discipline: pick a market where the concentric circles can keep expanding (Palo Alto started as a firewall add-on; Facebook started as a half-campus social network) — intentional sequencing matters more than the starting wedge.

Chapter Summaries

1. Childhood and wiring — Grew up in Houston; competitive debate and small-team biomedical research shaped him. Venture rewards the hyper-competitive/hyper-curious — at any age, if you’re wired for it.
2. Are we in an AI bubble? — Yes, arguably worse than 2021. 100-200x revenue multiples vs. 15-20x public comps; corporate strategic investors distort pricing; many AI apps show spiky non-durable usage.
3. How to evaluate AI apps — Use both the “data is exploding, why not invest?” lens and a fundamental PMF/defensibility lens. Refuse to chase if fundamentals don’t hold.
4. OpenAI risk and what’s defensible — OpenAI will own foundational primitives (text gen, raw code gen). Focused vertical and workflow applications are safe and can be huge.
5. Foundation model layer — Maybe 1-2 winners as API/cloud; real foundation-model value accrues where the model must couple tightly to the app (personal agents, enterprise agents, possibly code gen).
6. The end of per-seat / end of SaaS? — Per-seat pricing evolves to hybrid; SaaS is far from dead — generative AI is the first force in 15 years that disrupts data model, delivery model, AND interface, opening the door to new Salesforce-scale companies.
7. Pricing power from AI — Real pricing power comes when AI replaces work units (e.g., AI BDR), not when bolted onto existing tools.
8. Seed and Series A pricing — $20-40M post seed and $80-200M post A for strong teams; price discipline is irrelevant for iconic outcomes but matters at portfolio level.
9. Large seed rounds and founder caliber — A big seed for a proven team can be less risky than a small seed for an unproven one. Capital is an accelerant for the right founder.
10. Signaling and reserves — Signaling is overblown; Greylock seed-to-A conversion is high. Reserves serve both offense and defense.
11. Why $1M ARR is a BS Series A filter — Most $1-2M ARR companies never get to scale; founder + market should lead, ARR is secondary.
12. Founder vs. market trade-off — Markets have gravity; bet on iconic founders especially in good “zip codes.”
13. Series B is dead? — No. The market is arguably frothier than 2021 because early-stage platforms with growth funds have replaced retreating crossover players. Saam doubts the Series B vintage will perform.
14. Mistakes — Passing on Glean (Arvind Jain) and Codium (Varun Mohan) taught him: when you meet a truly iconic founder, default to yes regardless of the initial idea.
15. Sourcing, selecting, servicing — Sourcing is hardest and most important.
16. Career advice for young VCs — Don’t optimize for stamping your name on deals. Build a lane in a sector; develop a clear “why would a founder want me?” answer. Sourcing edge compounds.
17. Quickfire — Pre-PMF help is real; admires Elad Gil and Gili Raanan most outside Greylock; best first meeting was Ankur Goyal (Braintrust); incubations mostly don’t work; changed his mind about focused models (ElevenLabs as proof); market size is everything — start with a narrow concentric circle but make sure the circles can keep expanding.