Every Seed Check Is an Option

Venky Ganesan is a partner at Menlo Ventures, a firm he describes as having a 50-year history and an anchor LP, the Washington State Investment Board, that has backed it since 1981. He has been in venture for about 28 years, starting as an associate at Globespan Capital Partners, where he learned his first lesson about selling from a 90% loss on IPO shares. Menlo was an early backer of Anthropic, and Ganesan says the position grew to 20% of one fund after the company 10x'd in each of the last two years. He calls himself a "blowhard" on Twitter, wears pocket squares to partner meetings, and closes by crediting a Ronald Reagan line about not caring who gets the credit. These signals are from Venky Ganesan's interview with Harry Stebbings on 20VC.

1. Going for Broke. Ganesan says whether to sell depends on who you are and what your balance sheet looks like. As a 24-year-old associate he put $5,000 into Avanex at its IPO, watched it reach $200,000, refused his fiancée's advice to sell some for a house down payment, and sold for about $8,000 to $9,000 after a 90% drop. The lesson he took: at some point you take chips off the table. Menlo is in the opposite position today, and he says the firm is "going for broke" on the defining AI companies because it has hit home runs before and can afford to swing.
2. Dancing While the Music Plays. Ganesan quotes Chuck Prince ("when the music is playing, you got to dance") because a professional investor never knows when a cycle ends. Several smart firms got into the dot-com boom in 1993 and 1994, stepped out in 1996 and 1997, missed 1997 through 1999, and came back in 2000 at the peak. His answer is to stay in the game and play it differently: be more selective, and use portfolio composition and position sizing to protect against the turn. He also warns against building a long-term strategy off today's rounds, since "a dot is not a line."
3. Seed as an Option. "Each seed investment is an option bet," Ganesan says, a way to buy a look at whether a company is an outlier. He admits that large funds, Menlo included, are "somewhat indifferent" to seed valuations because the seed check is a seat at the table, and the real goal is to invest more once a company breaks out. That indifference is part of why AI application seeds now raise $10 million to $20 million, up from the old $3 million to $5 million. His rule is to hold enough at-bats to catch an outlier, then size up only when revenue and other hard metrics prove it.
4. Laddering to 20%. Menlo owns less than 2% of Anthropic, and Ganesan says that one company reached 20% of a fund, the most Menlo would put in a single name. Getting there "was a hard conversation," made easier because Anthropic grew 10x this year and 10x the year before. He separates writing 20% of a fund in one check on day one from laddering up to 20% as new data arrives, and says the second is far better. With even 10% ownership now hard to get, he expects firms to win by establishing a position and adding to it as the company performs.
5. Ownership as Insurance. Ganesan says he would take 2% of a trillion-dollar company over 20% of a $100 million company, but ownership before a company is a proven outlier still matters. A fund that only swings for grand slams has no singles or doubles to fall back on if it misses the big winner, so early ownership lets mid-sized outcomes still move the fund. His example is Higgsfield, where partner Amy got Menlo 15% for a $5 million check and the firm has since invested more. Once a company is a known outlier, he says, "it's no longer a selection game," everyone wants in, and there is no alpha left.
6. Gamed Metrics. "Once a metric is measured, it can be gamed," Ganesan says, and he expects the next few years to expose "the accounting creativity of some of our founders." In the SaaS era, investors leaned on net revenue retention, so a founder could take a $10 purchase order, then a $50 one a week later, and report 500% retention instead of the 100% a single $100 order would show. He cites the economist's term "bezzle" for the hidden embezzlement that swells in every boom. His filter is whether founders and investors are focused on terminal value or on markups.
7. Reflexivity Until It Stops. Ganesan reads kingmaking as George Soros's reflexivity: fast revenue earns a quick markup, the markup brings capital, press and talent, and the company grows faster. That works while revenue is real, but copycats conclude "the secret is to have a markup," investors assume a marked-up company will be marked up again, and the loop runs "until it stops." He says equity rarely cracks a cycle, because equity losses just get written down. Debt does, since lenders expect to be repaid, and at 4x leverage, being right is not enough without the timing.
8. Innovators, Imitators, Idiots. Every cycle brings innovators, then imitators, then "eventually the idiots," and Ganesan places same-week tranche rounds (100 this week, 200 next week) on that path. The original logic was sound: bring in "build with me" investors at a lower price, then use them to raise plain capital at a higher price. Now everyone does it regardless of company quality. Menlo will still come in on the higher tranche if the founders are special, because "my only ego is to make money for my investors," and he says the biggest mistakes happen when VCs get caught in their egos.
9. Sins of Omission. Venture is asymmetric: you can lose the dollars you put in but make 10x if you're right, so Ganesan says the most expensive mistakes are the deals investors pass on. His came on the Plaxo board, alongside Mike Moritz, Ram Shriram and Tim Koogle. Sean Parker, who had just been pushed off that board, told Ganesan he was going to work with a college dropout and urged him to get involved. Ganesan didn't take the meeting and missed the chance to write roughly a $50,000 check into what became Facebook.
10. The Acquisition Floor. Ganesan distrusts any investment case that rests on an incumbent buying the company for at least its liquidation preference. Recent prices, which he lists as AMD paying $8.5 billion, Nvidia paying $14 billion for Hugging Face and Stripe "allegedly" paying $8 billion for OpenRouter, make a $1.5 billion floor feel safe. In the dot-com era Nortel bought Xros for about $3.5 billion and Lucent bought Chromatis for $4.5 billion, both pre-revenue and paid in stock, and Redpoint's Geoff Yang said winners would sell for billions and losers would still return the preference stack. That didn't hold after March 2000, and Ganesan notes that acquirers care about founders, so structured deals often hire the team and leave investors behind.
11. Velocity and the IRR Hurdle. Menlo assumes 60% dilution from its first check, so 10% at seed becomes 3.5% to 4% at exit, and Ganesan says time is what drives that number. Fast-rising companies dilute less: a $200 million company might give a senior hire 2%, while the same person costs about 0.1% at $2 billion. Speed also matters for returns, because every venture-backed company pays a tax to Nvidia, the hyperscalers and the model labs, all of which LPs can own in a no-fee index fund. A venture fund, he says, has to beat that by 1,000 basis points to justify private-market capital.
12. The Windshield Test. Ganesan's advice to LPs: "Look at the windshield, not the rear-view mirror," since fund performance lags by 5 to 7 years. He tells them to call founders of successful AI companies and ask which partners they respect, including partners whose money they didn't take. LPs want DPI over paper gains, and he says they can't sit out AI because their private equity books, often 3 to 4x their venture exposure, are full of software that AI is hitting. Menlo learned the cost of crowding a vintage firsthand: in its 50 years, the only fund that didn't return capital was Menlo 8, deployed in 10 months across 2000 and 2001.

Watch the full video at https://www.youtube.com/watch?v=EZWsOxpVWwU.