Overview of 20VC with Sameer Dholakia, Bessemer Venture Partners
Harry Stebbings interviews Sameer Dholakia of Bessemer Venture Partners about the AI supercycle, why capital is now a major moat in venture, how Bessemer is thinking about seed and growth in a more expensive market, and where the next decade of venture returns may come from. The conversation centers on a core thesis: AI/intelligence is becoming a platform shift larger than cloud, and the firms that win will likely be the ones with the most capital, the best distribution, and the strongest conviction.
Key Takeaways
-
AI is viewed as a “magic moment” comparable to the internet and cloud, but likely bigger than both.
-
Dholakia argues that intelligence will be a market larger than oil, because it powers every human and company, not just machines.
-
Capital intensity has fundamentally changed venture competition: the best AI companies now need huge war chests to train models, scale infrastructure, and defend share.
-
Bessemer’s view is that the market will be dominated by a small number of frontier model leaders plus a large ecosystem of AI applications built on top.
-
In today’s market, entry price still matters, but quality of revenue, retention, and the next 12 months’ growth matter more than current ARR multiples.
-
Bessemer believes seed and early-stage investing now require larger funds, because valuations are higher and only a few wins will drive most returns.
-
Dholakia sees venture firms as increasingly needing to be either:
- large global platforms, or
- deep specialists
—with the middle becoming a difficult place to make money.
-
He expects companies to stay private longer, but thinks the IPO window will reopen as public investors demand pure-play AI exposure.
-
He is especially excited about robotics/embodied AI and AI in therapeutics and drug discovery.
AI, Frontier Models, and the “Cloud Redux” Thesis
Intelligence as the next giant platform
Dholakia’s central framework is that AI is a cloud redux:
- Like cloud, frontier AI is massively compute-intensive and capital-intensive
- That means the market should consolidate into a handful of major providers
- But, just as cloud led to countless SaaS companies, AI will likely support hundreds of vertical and horizontal applications
He believes the opportunity is not just big — it is category-defining at global scale.
Why Bessemer backed Anthropic
Bessemer invested in Anthropic’s Series D at roughly $15B pre / $18B post. Dholakia said the logic was:
- AI models require enormous capital to scale
- Only a few players had the balance sheet and strategic positioning to win
- AWS distribution and compute support made Anthropic especially compelling
He admitted the deal was controversial internally, but Bessemer’s culture encouraged conviction-based action.
Open vs. closed models
Dholakia does not believe the debate over open vs. closed models should be framed as either/or. His view:
- Both will be huge
- The overall market is so large that precise token share ratios matter less than overall adoption
- Bessemer is investing across both closed frontier labs and open model ecosystems
He also noted that enterprise buyers often care deeply about model provenance, especially if the model is seen as “Chinese,” but argued that once hosted and supported by a trusted provider, the buying dynamic changes.
How Bessemer Thinks About Valuation, Margin, and Growth
Price matters, but future scale matters more
Dholakia repeatedly stressed that current valuation is less important than future compounding potential. He said he is willing to pay seemingly extreme multiples if he believes:
- the company can scale very fast,
- the business model is durable,
- and the company can still produce venture-scale returns.
He referenced prior investments like:
- Anthropic
- Legora
- Fireworks
- Field Guide
as examples where current ARR multiples looked absurd at the time, but the outcomes could still be exceptional.
Gross margins and near-term economics are not the whole story
His framework for AI companies:
- In the next 12 months, gross margins matter less than many investors think
- Over a 5–7 year holding period, margin expansion must be plausible
- The real questions are:
- Can this company win the market?
- Can it compound revenue?
- Will scale improve efficiency over time?
He’s especially focused on:
- quality of revenue
- gross dollar retention
- net dollar retention
- strategic importance of the product to the customer
Venture Fund Strategy: Why Bigger Funds Make Sense
“Do you need a billion dollars to do seed?”
Dholakia’s answer was effectively: maybe not literally, but seed has become so expensive that small funds are at a disadvantage.
His reasoning:
- Seed prices are much higher than in the past
- AI startups often need significant capital early
- A seed fund needs enough shots on goal to find the winners
- In many cases, seed is now an option on future ownership, not a standalone return engine
He described Bessemer’s current structure as:
- $1.75B seed/early
- $4B growth
and said the firm is intentionally built as a platform with reserves and the ability to “back up the truck” on winners.
Why only platforms or specialists survive
He argued that as venture matures and professionalizes:
- big winners will go to firms with massive capital and brand
- or to hyper-specialists who can see and support opportunities others can’t
Being in the middle, in his view, is increasingly hard.
Notable Portfolio Examples
Legora
Dholakia used Legora to illustrate how huge vertical AI opportunities can become.
His view:
- Legal services is a massive global market
- AI doesn’t need to take a huge share to build a very large business
- The company can become a major player even if it remains verticalized
Fireworks
He was especially bullish on Fireworks, saying the company has the right combination of:
- technical depth,
- enterprise relevance,
- and go-to-market leadership
He highlighted the pairing of Lynn and George Hu as a particularly strong founding/executive combination.
Fireworks’ thesis, in his words, is that enterprises will want to own some of their intelligence, not rent all of it from a single frontier provider.
Waymo / autonomy
Dholakia is very excited about Waymo and autonomy more broadly, arguing the opportunity is bigger than ride-hailing alone:
- personal vehicles
- trucking
- logistics
- construction
- mining
- agriculture
He sees autonomy as a potential trillion-dollar category.
Therapeutics and drug discovery
His second major excitement area is AI for therapeutics:
- automating wet labs
- accelerating discovery
- helping find cures for previously “undruggable” diseases
He framed this as one of AI’s highest-impact applications.
Culture and Leadership
Dholakia spent a meaningful portion of the interview on culture, emphasizing that:
- “Teammates and friends beat colleagues and co-workers every day and twice on Sunday.”
- Teams perform better when people genuinely care about each other
- Culture should be explicit, enforced, and aligned with values
At SendGrid, he used four core values:
- Happy
- Hungry
- Humble
- Honest
He also argued that directness in investment committees is a strength, not a weakness.
Liquidity, IPOs, and the Private Markets
Dholakia believes:
- companies will stay private longer as a structural shift
- LPs are increasingly allocating to private markets to capture the upside
- the IPO market will likely reopen more meaningfully later next year and into 2028
He expects the reopening to be driven by:
- investor demand for pure-play AI exposure
- the appetite for new public listings after a long window of inactivity
He also said it makes sense for companies and founders to take some secondary liquidity along the journey.
Final Thoughts
The interview’s core message is that AI is not just another software cycle. It is a capital-intensive, infrastructure-heavy, globally significant wave that will reshape venture, enterprise software, and the public markets. Bessemer’s posture is clear:
- invest early in frontier intelligence,
- support both open and closed ecosystems,
- concentrate capital behind true winners,
- and build a firm that can operate either as a global platform or a highly specialized edge player.
In Dholakia’s view, the next decade of venture will be defined by scale, conviction, and capital discipline — and the biggest returns will come from backing the companies that become foundational to how intelligence is built and distributed.