Diversified seed investing beats concentration, argues Box Group's David Tisch

1. Micro outweighs macro — ▶ 3:18

   Tisch rejects the question of whether today resembles 2021. Each market moment, he says, is a standalone condition, and for an early-stage investor the macro matters less than whether a given company is the good, great or most important one. Frothy periods hold public blowups and companies that die quietly, but also a handful to many great, long-term businesses. He sees AI as a horizontal technology advancing in mini-cycles of two to three months, driven by model releases, vertical AI revenue growth and large M&A deals. The founder's ability to navigate that churn is what he tries to judge.

2. Why high entry prices still work — ▶ 7:22

   Tisch names two conditions under which 10-20 million dollar seed rounds do not damage venture economics. First, outcome sizes have grown: a $10 billion company was far-fetched 10 years ago and is now a normal outcome one must squint to reach. Second, dilution has fallen. In companies that hockey-stick quickly, Series B-D rounds cost founders and early holders 3-8% instead of 12-20%, so early stakes shrink less. He concedes that a portfolio made entirely of $20 million seeds at $200 million pre-money would change the business, so he wants balance across the portfolio, not on every deal.

3. Avoiding head-to-head seed fights — ▶ 10:29

   Tisch doubts that any manager can reliably buy 10-20% of a true outlier at seed against the best firms. He says multi-stage firms are now more concentrated, accessible and efficient at seed, and founders know how to run processes. Competitive rounds that once drew 3-5 term sheets now draw 10-15, or stay at 3-5 firms competing hard against each other. He says the host got two things wrong: the business math of a manager who must return outlier capital to LPs, and the assumption that early stakes lose most of their value to dilution.

4. Adverse selection in concentration — ▶ 14:34

   Concentrated managers claim they win 90% of the deals they want. Tisch calls that total BS and estimates 20%. A fund that needs 10% of 30 companies, he reasons, must offer terms to 120-150 companies to land its 30. The best companies, those the firm most wanted, may therefore be missing from the portfolio. His model gets into more of the companies it is excited about. This is his estimate, not measured data, and the host agrees the win rate is overstated but worries that cheaper-ownership rivals will re-price rounds upward.

5. Lean startup as fiction — ▶ 17:16

   The host worries that founders handed double the price and far more money lose focus. Tisch calls that a babysitting problem, noting that some founders benefit from runway, recruiting buzz and credibility with early customers. He says the lean startup was a total fake and challenges anyone to name a company that succeeded by running that playbook. When the host cites 37signals, he allows that some bootstrapped companies work but says anyone on the venture flywheel will end up fat. The difference between lean and well-funded, he argues, is not discipline, and funded companies must simply operate differently.

6. Tranches and the long game — ▶ 18:27

   Tisch dislikes tranched rounds but separates the emotional reaction from the business decision, and says he would still take a later tranche. He holds that the right to invest is earned, and that if a founder refuses his preferred terms it is his problem. He objects more to the other VC who structured a deal that pushed him into a different product, and says he has a long memory. In his framing, some investors play a short game in hot markets while others protect long-term relationships. The host argues that the cutting-out itself is the short game.

7. Losing on speed, not price — ▶ 23:53

   Tisch puts Box Group's win rate between 70% and 90% depending on market conditions. Its main losses come from reaching a deal too late and from the speed of its own decisions, both within its control. When it arrives late, he says, it is begging for scraps. He now accepts checks as small as 1-2% if the founder is exceptional, and initial ownership ranges from 1% to 10%. He admits the firm turned down great businesses over ownership in the past. The team of 10 investors meets 50-100 companies a week and sees an estimated 50-75% of what it wants to see.

8. Lead-only seed squeezed — ▶ 26:36

   Asked whether pure-play, lead-only seed funds still have a business, Tisch says the remaining examples have moved to more flexible styles and that he does not want to be in that business. His reasoning is that the greatest firms run the strategy as part of their offering. The host adds that those firms have bigger checkbooks and less price sensitivity, and says he has lost deals to firms offering treble the price and check. Tisch adds that the old habit of waiting to see a company at the A or B is gone, so losers must now find new entry points with minimal progress.

9. Decoupled founders, GPs, LPs — ▶ 29:03

   Tisch sees founders, GPs and LPs with decoupled incentives. Founders hold the leverage, GPs can only say yes or no, and LPs want exposures that are not available. He says LPs claim to dislike multi-stage firms' multiple products, yet most venture allocation goes there for scaled capital and exposure to top companies. Box Group runs a $275 million seed fund investing in around 150 companies. Tisch says some LPs are confused by the math and do not believe the model, but insists the math works. He offers no portfolio-level figures to support that claim, which stays his assertion.

10. The messy middle of outliers — ▶ 32:23

    Cursor, in which Box Group was in the earliest round, is Tisch's best liquid return, and he agreed it returned multiple times. He calls it the biggest venture M&A deal in history, with the host noting four years to that value. He stresses it is atypical. Clay was a bumpy four-year road, and Baseten needed inference to become a category. Early signs of a great founder include hiring outlier people and reaching inaccessible customers. But with undergraduates unable to see beyond two years, Tisch says no one can know immediately; the skill is real-time navigation.

11. Rogo and 39 passes — ▶ 43:20

    Tisch describes the founders' market as one of halves and have-nots, with a small number of hot categories attracting nearly all the interest. His example is Rogo, an AI-for-finance company. After the seed round, he put his name behind it and sent it to roughly 40 Series A lead investors, who passed 39 times. Keith then led the A, and Sequoia and Kleiner later invested. Tisch reads the episode as evidence that early-stage information is imperfect: companies off the zeitgeist can be misread at one stage and rewarded for execution at another.

12. Pro rata is earned — ▶ 48:25

    Tisch dismisses early-stage VCs who say they will give up pro rata as marketing garbage. He holds that the right to invest in later rounds is earned whether or not it is in the documents. Having it in the documents, he says, helps the founder more than the firm, because it gives leverage against a later investor demanding, for example, 22% when 6% of the cap table needs pro rata. He adds that when large firms say they need 20%, he has seen them take 5% a week earlier, and need means hope. Box Group's follow-on fund makes 20-30 checks of 5-30 million dollars.

13. Why he doesn't sell early — ▶ 52:31

    Tisch works for LPs, but says his customers are founders, and the two relationships are dislocated. He does not sell early, for two reasons: reputational risk with founders, and the belief that the best companies compound most late, since going from $2 billion to $4 billion to $8 billion is easier than going from zero to $1 billion. He rejects the idea that founders selling secondary justifies investors doing the same, because founders face different economics such as debt, stress and risk appetite. He supports employee tenders for retention, and says that de-risking is not right for a firm in the business of risk.

14. Fund four reveals ambition — ▶ 58:54

    When Tisch meets new managers he asks what fund four looks like, because that is their actual vision and starter funds rarely stay the same. New firms, he says, mostly want to become multi-stage, a very competitive part of the market, and many multi-stage leaders wish they ran smaller firms. The host notes that Box Group just raised 550 million across two funds. Tisch replies that the investing style is the same and the DNA is the dream stage. He sees consolidation, with 20-30 competitive firms and about 10 outliers among them.

15. No grenades at early stage — ▶ 1:05:58

    Box Group holds investment committee meetings twice a week and does not vote. Someone must want to do a deal, and anyone who has not met the founders is not allowed to be negative. Only ethical or reputational issues about a founder are fair to raise, and Tisch calls this the no-grenades rule. His reasoning is that any early company can be shown to fail, and with 70-90% of early-stage companies not working, listing risks is obvious and irrelevant. The job is to see why something will work. Tisch says the loss rate Keith claims is plausible at Series A, but at seed the failure numbers are real.

David Tisch's central claim is that venture's arithmetic has moved in favor of the diversified, low-ownership seed investor, and that defenders of the concentrated model work from a flattering picture of their own win rates. His case rests on two mechanisms. The first is outcome size. A decade ago, he says, building a $10 billion company was far-fetched. Today an early-stage investor has to squint to see it as a normal outcome, and the anomalies run well past it. The second is dilution. In companies that hockey-stick quickly, Tisch says, Series B, C and D rounds now cost founders and early holders 3-8% rather than 12-20%. Across several rounds, an early check is diluted less, which he calls an economic factor offsetting a higher entry price.

The debate carries a personal edge. The host has long questioned Tisch's low-ownership, very diverse portfolios, and opens by saying that Tisch has "absolutely crushed it." When pressed on what exactly he got wrong, the host concedes he did not foresee outcome size expansion. Tisch gives a two-part diagnosis: the business math of a manager who must return outlier capital to LPs, and the dilution assumption. He does not claim a smooth path. Uber, Facebook, Airbnb and Ramp each had a down round, and the largest labs are now re-evaluating the prices at which they raise. His claim is narrower: within a three-year window, valuation momentum has been strong. He also concedes that if every entry point were a $20 million seed at $200 million pre-money, the business would fundamentally change. His answer is balance across the portfolio, not on each deal.

The sharpest part of the argument concerns the concentrated model. Every concentrated manager, Tisch says, claims to win 90% of the deals they want, which he calls total BS and puts nearer 20%. The consequence follows from simple arithmetic. A fund that wants 10% of 30 companies must, on his estimate, offer terms to 120-150 companies to land its 30. Those 30 are then the companies that accepted the terms, not necessarily the 30 the firm most wanted. Tisch's own model is built to get into more of the companies it is excited about. The logic is coherent, but the 20% figure is his estimate, and the host only agrees that stated win rates look overstated. Neither offers data, so the argument is plausible but unmeasured.

A second force is that the seed market has become more efficient and more contested by multi-stage firms. Tisch says those firms are now more concentrated, accessible and efficient at seed, and that founders run processes far better than before. Competitive rounds that once drew 3-5 term sheets now draw 10-15, or stay at 3-5 firms competing hard against each other. In that environment he does not want to be a lead-only seed fund, because the greatest firms run that strategy as part of what they do. The host adds that those firms bring bigger checkbooks and less price sensitivity. A further consequence is that waiting is no longer an option. Investors who once skipped the seed to see a company at the A or B now often never get to see it, and losing head-to-head forces a new entry point with minimal progress.

Founders hold the leverage in this market, and much of the conversation concerns how investors should respond. Tisch's stance is that the free-market dynamics are not negotiable: a firm either says yes or no. The host worries that tranched rounds, in which a rival re-prices a company at double the price, distract founders. Tisch dismisses that as a babysitting problem and calls the lean startup a total fake, arguing that anyone on the venture flywheel will end up fat and that the difference between lean and well-funded is not discipline. He dislikes tranches but separates the emotional reaction from the business decision and would still take a later tranche. His real objection is to the VC who engineered a structure that pushed him out, and he keeps a long memory.

The same long-game logic shapes his view of pro rata. He calls the idea that early-stage VCs should give it up marketing garbage, and holds that the right to invest in later rounds is earned through behavior, not paperwork. Having it in documents, he says, is more useful for the founder than for the firm. If a later investor demands 22% and 6% of the cap table already needs pro rata, the founder has leverage to reopen the math. A firm that refuses a founder's request to step back, he adds, risks both financial and reputational loss. His broader point is that relationships are the scarce asset when founders hold the leverage, and some firms are optimizing for the next deal, not the next decade.

Among stakeholders, the incentives are decoupled. Founders get the leverage, GPs can only say yes or no, and LPs want exposures that are unavailable. Tisch notes that LPs say they dislike multi-stage firms' multiple products, yet the majority of venture allocation goes to those firms for scaled capital and access to the best companies. Box Group runs a $275 million seed fund investing in around 150 companies and a second $275 million fund that writes 20-30 follow-on checks of 5-30 million dollars, with a maximum of roughly $40 million in one company. Some LPs, he says, do not believe the math. He insists it does, citing Cursor, where Box Group was in the earliest round and agreed the investment returned multiple times.

Tisch is careful not to turn Cursor into a template. Four years to a historic exit is the exception, and the other companies in that fund take a slower path. Clay was a bumpy four-year road before explosive growth, and Baseten waited for inference to become a category. This is why he says early-stage investing cannot be about immediately recognizing the best company. The signs he looks for are outlier hires and customers that are inaccessible so early. But he also says the world is changing so fast that nobody knows where it is going. On campuses, undergraduates could not see ten years ahead, and only a few saw even two. What matters is whether a founding team can navigate conditions in real time. Rogo shows the market's imperfect information: 39 of roughly 40 Series A leads passed, then Keith led the round and Sequoia and Kleiner followed.

Operationally, Tisch's philosophy becomes a rule against negativity. Box Group meets 50-100 companies a week, holds investment committee twice a week and does not vote. Someone must want the deal. People who have not met the founders may not criticize the deal, and only ethical or reputational issues about a founder count as exceptions. Tisch's logic is that any early company can be shown to fail, and with 70-90% of early-stage companies not working, a list of risks is obvious and irrelevant. The magic is seeing why something will work. This fits the diversified model, where a low per-deal threshold and a wide funnel substitute for the conviction a concentrated fund must concentrate in a few names. The tradeoff is that the model gives up the information that a skeptical partner might surface.

Tisch applies the same logic to liquidity. He does not sell early, because he thinks the best companies compound most late and that going from $2 billion to $4 billion to $8 billion is easier than reaching the first billion. He distinguishes founder secondaries, which can ease debt, stress or risk appetite, from the fund math of a VC trimming 20% of a position to make results look good. He supports employee tenders for retention. The reasoning is internally consistent, but it asks LPs to accept that the largest gains arrive last, and he gives no figures on how often that has paid off. He also acknowledges that he works for LPs while describing founders as his customers, and he calls those relationships dislocated.

The limits of the evidence are material, and the reader should keep them in view. The argument that concentrated funds win 20%, not 90%, is an estimate. Tisch's own win rate of 70-90% and his 50-75% estimate of what the firm sees are also self-reported. The returns story leans on Cursor, which Tisch himself calls a generational outcome and the biggest venture M&A deal in history. His own account of seed failure rates, 70-90% in his estimate, means the fund relies on rare winners. There is also a tension, which is an editorial observation, between the claim that Box Group has no ambition beyond early stage and the fact that it has just raised 550 million across two funds, a point the host pressed. Tisch's answer, that the style is unchanged and only scale differs, is plausible but untested at larger sizes.

What the argument finally depends on is persistence. Tisch's model works if outcomes of $10 billion and beyond keep becoming routine and if the best companies keep dilving their early investors only 3-8% in later rounds. He describes the current momentum as a three-year window, and he sees the largest labs re-pricing. If either condition weakens, the case for tolerating high entry prices weakens with it, and the 20-30 firm consolidation he predicts would favor those with ownership and access. His own model has to stay balanced across the portfolio, since he concedes that all high-priced seeds would change the business. The test of whether that balance holds will come in the next market, not this one.