Venky Ganesan on AI Venture Capital Ownership

In 2000, a young venture associate named Venky Ganesan put $5,000 into the IPO of a fiber-optics company called Avinex. The stock climbed so fast that his modest stake was worth $200,000, a fortune to a man about to get married. His fiancee suggested selling some for a house down payment. He refused: Avinex was going to a million. It dropped 90 percent instead, and he sold what remained for about $8,000 to $9,000 (05:40). Twenty-five years later, as a partner at Menlo Ventures, he still calls it the most useful loss of his career, not because it taught him to sell everything, but because it taught him that at some point, an investor has to take some chips off the table.

That lesson sits underneath almost everything Ganesan says about venture capital today, because the game, by his own account, barely resembles the one he joined 28 years ago (36:27). Stebbings' team now comes to him and reports they cannot find a seed deal under $100 million, meaning that is the size of the round, not the valuation (07:29). AI application startups are raising $10 to $20 million at seed, a number that would have been a Series A a few years ago (10:19). The frontier labs raise far more. Ganesan's response is not panic but a kind of discipline: a dot, he says, is not a line, and a snapshot of one crazy quarter should not reset an entire investment strategy (07:47).

Part of that discipline is treating every seed check as what he calls an option bet.

"Each seed investment is an option bet. You're buying an option to see if it's an outlier." — Venky Ganesan [09:34]

The point of the option is not ownership for its own sake. It is information. A small check buys a seat at the table and the right to watch a company's real numbers before deciding whether to size up. That is also why Ganesan is candid about why seed prices have become indifferent to logic: large funds, including his own, are often willing to overpay at seed because the real goal is simply to be in the room if the company becomes a monster (10:19).

Those real numbers, though, are getting harder to trust. Founders now show contracted revenue that is not yet collected, or run-rates extrapolated from a single best day multiplied by 365 (11:23). Ganesan has a blunt explanation: any metric that investors weight heavily will eventually be gamed, the same way net revenue retention was gamed during the SaaS years by splitting one big contract into several smaller ones to inflate the growth math (11:48). He reaches for a concept he attributes to Keynes or another famous economist, the idea of "the bezzle," the gap between when embezzlement happens and when it is discovered, which grows quietly during every boom (12:07). Layered on top of that is what Ganesan calls reflexivity, borrowing from investor George Soros: a company marks up, the markup attracts more capital and talent, which drives real growth, until copycats start chasing the markup itself rather than the growth, and the cycle eventually snaps (13:20). Historically, he says, the snap comes from debt, not equity, because equity investors simply write down a loss while lenders expect to be repaid regardless of what the business is actually worth (14:28).

Ownership Stops Being the Point

If seed-stage ownership is shrinking, as it is across the industry, Ganesan argues that is not automatically a problem. "I'd rather take 2% of a trillion-dollar company than 20% of a $100 million company." — Venky Ganesan [25:10]. Menlo owns less than 2% of Anthropic (25:22), a stake that is still larger in dollar terms than a dominant position in a mid-size outcome would be. But he is careful to separate two different games. Before a company proves itself an outlier, he argues investors should fight for ownership, because that period is still a selection game. Once a company is an obvious winner, ownership becomes a pure capital-allocation question, since everyone already knows it is going to work (25:50).

That does not mean concentration is unlimited. Menlo caps any single company at roughly 20% of a fund, and has only once gone that high, for Anthropic (27:13). The logic is about sequencing: laddering up to a large position as new data confirms a company's trajectory carries far less risk than betting big on day one, before there is any evidence at all (27:36).

Dilution complicates the ownership math further. Menlo assumes that a 10% stake bought at seed will shrink to roughly 3.5 to 4% by exit, a loss of about 60%, driven by follow-on financing rounds and the steady expansion of employee option pools (34:00). But the rate of dilution depends heavily on how fast a company grows. A company that scales quickly and exits quickly, the way Stebbings describes AI model interface platform OpenRouter doing, suffers far less dilution, because a fast rise in valuation means each new hire costs less equity. A $200 million company might give up 2% of itself to hire one senior executive; the same hire at a $2 billion company might get $20 million in stock that represents just 0.1% of the company (36:04). Time, in other words, is the quiet variable behind every dilution outcome.

All of this feeds into a shift Ganesan says the whole industry has made: from cash-on-cash thinking to a focus on internal rate of return, or IRR, a measure of how fast money grows annually rather than just by how much. He argues that because every major AI company now effectively pays a toll to Nvidia, a cloud provider, and a foundation model company, all of which investors can access through public, no-fee index funds, venture capital has to clear a much higher bar: roughly 1,000 basis points, or 10 percentage points, of extra annual return above the public market, just to justify its fees (37:13). Limited partners, meanwhile, are asking for something more concrete than IRR: distributions to paid-in capital, or DPI, which is cash actually returned rather than a paper markup (41:19). Ganesan says most LPs now have two complaints: they want real DPI, not just unrealized gains, and they cannot afford to skip the AI wave because their much larger private equity portfolios, often three to four times the size of their venture allocations, are directly exposed to AI disruption and need a hedge (44:09).

What Money Reveals

Ganesan is just as direct about the psychology of the business as the math. On the subject of founders who have made enormous sums very young, he offers a line he has clearly turned over before.

"Money doesn't change people, it reveals them." — Venky Ganesan [50:45]

People who were driven before stay driven, he argues, because for the best operators money is simply a scoreboard, not the game itself (51:04). But he is unusually candid about how much privilege plays into his own clean conscience, admitting that his comfort with doing the right thing now, as a wealthy and established investor, is not proof he would have made the same choices at 25, scrapping to build a career.

"Morality is sort of a privilege of the people who already succeeded. It's easy to be moral now when you already have the things you have." — Venky Ganesan [53:48]

He extends the same honesty to the structure of venture capital itself. Being richer, he says, makes him a better investor mainly because it removes fear: he can go for broke more easily, the way a poker player with a large chip stack can see more cards and take more risks (54:29). That, he concedes, is exactly why the rich tend to get richer in this business, and why funds between $30 million and $100 million are, in his words, the worst place to be right now, squeezed by firms with far bigger stacks, unless the manager is exceptional enough to beat the odds anyway, the way Sarah Guo's Conviction or Dave Tisch's BoxGroup have (55:09).

His closing advice to limited partners sizing up a fund is less about spreadsheets than about asking the right people. Performance, he says, is a lagging indicator, trailing the real decisions by five to seven years, so a track record tells an LP what a firm did, not what it will do (60:29). His suggested test is simple: call successful AI founders and ask which investors they respect most among the ones who never got to write the check. If a firm's name does not come up among the rejections, that is the real signal, more honest than any IRR on a slide deck.