20VC with Harry StebbingsWho Wins in the Next Decade of VC | Margin and Price Matter Less in a World of AI | Sameer Dholakia
20VC with Harry Stebbings
AI as a new market 0:00
Intelligence is framed as a market bigger than oil, because it powers every kind of user. The speaker says the feeling is like earlier software moments, when new layers of technology made whole new businesses possible. He points to the internet, VMware, and then Jasper on top of GPT-3 as moments that felt like magic and signaled a larger cycle.
Anthropic and the model split 5:01
The Anthropic investment is described as a Series D at about $15 billion pre-money and $18 billion post. The logic was that frontier AI would be capital intensive and would likely consolidate into a few major players, while many more applications would still be built on top. Open and closed models are both seen as important, and the exact token mix is treated as less important than the size of the market.
Entry price and outlook 15:00
Entry price still matters, but he judges it against what a company can become over the next 12 months. He says he is less focused on what the business looks like today than on whether it can compound into a venture return. That is why he was willing to back companies at prices that looked extreme, including Bridge at about $5 million of ARR and Anthropic at roughly $15 billion. He would rather think forward than stare backward at current multiples.
Quality, scale, and capital 19:02
He cares more when revenue is sticky, gross dollar retention is strong, and customers treat the product as strategic. He is much more cautious with consumer or high-churn businesses. Margins matter, but mostly over a five-to-seven-year hold, because scale and market position can improve them. He thinks legal AI can reach $5 billion to $10 billion in revenue by serving a global legal spend near $1 trillion, and that capital itself is now a moat.
Industry scale has changed 30:00
Seed checks once meant three to five million dollars on a twenty to forty million valuation. Now the dollars are much larger, and that makes picking far more important. The speaker says venture has become a professionalized asset class. It now spans the world, every sector, and stages from pre-seed to IPO. In that setting, you must be either a giant platform or a sharp specialist. Being in the middle is a hard place to make money.
Growth funds need conviction 35:02
He says late growth should still be concentrated, not spread thin. A fund can back a winner with hundreds of millions across a few rounds, and reserves are part of the plan. He does not worry much about signaling when one company gets more capital than another. To him, the business will show whether it is taking off. Early growth and late growth are different, but both can produce ten and twenty times returns, with five times as a bare minimum.
Culture and teamwork 45:30
A company needs a clear set of values that shape how people work together. Here, that meant happy, hungry, humble, and honest. The point was not that every company should choose the same words. The point was that the people you hire should fit the way your team wants to operate. That is how you build teammates, not just colleagues.
Liquidity and public markets 53:00
In venture, DPI matters, so some gains should be realized along the way and returned to LPs. Late-stage secondaries are usually understood as prudent, not disloyal, and founders are often rational about selling a little for family security. The view here is that companies will stay private longer, but more IPOs should open up later because public investors want pure AI exposure.
Retail and private access 1:00:30
Retail can belong in venture, but only if the access is good. The view here is that broad democratization is hard to oppose, yet retail should avoid weak funds and adverse selection. It fits best in large global platforms with brand and reach, or in highly specialized managers. Otherwise, public markets may be the safer place for compounding without a liquidity hit.
Fireworks and sovereign models 1:02:01
Fireworks is framed as a major opportunity because Lynn and George together combine deep technical judgment with strong go-to-market skill. The larger case is that enterprises will want to own some of their intelligence, not rent all of it. That logic also supports sovereign models, especially for India’s many languages and for government, regulated, or military uses where local control matters.
Robotics and autonomy 1:16:00
Robotics is moving faster than many expect. Sameer says the firm has looked at many deals, though he cannot name one offhand. He is more interested in autonomy beyond ride hailing: construction, mining, agriculture, personal cars, trucking, and logistics. He sees these as large markets, not just a rideshare play.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
