For the 2010-2014 vintage, the median VC fund still reports year-ten NAV that exceeds total committed capital, meaning a significant share of fund value sits unrealized when the vehicle should be winding down: The ten-year horizon no longer corresponds to anything in the underlying economics of venture. Portfolio companies are simply staying private longer and getting much larger. The paper finds that this fund lengthening is not because modern funds convert NAV to cash more slowly; the speed of distributions after a liquidity event has not changed significantly.

In 2025, 33% of all US venture dollars flowed to the top 1% of companies by valuation (up from 12% in 2022), and mega-deal rounds over $500M captured a larger share than at the 2021 peak: The venture market has split into two separate industries operating out of one allocation bucket. At one end are massive growth rounds dominated by mega-funds; at the other, a shrinking cohort of disciplined early-stage investors. Fund extensions are simply a fact of life for LPs, with top-quartile funds routinely taking 16-20 years to fully return capital.

A 2025 paper in the European Journal of Finance finds that the optimal payoff for VC funds peaks between years 8 and 10, with the curve flattening or declining after year 10: The essay argues that mega-funds should follow Sequoia's path toward permanent capital (matching their reality of indefinite holding periods), while small funds should head in the opposite direction: using the ten-year horizon as a competitive feature, concentrating on early stages where pricing edges are real, and treating Series C/D as the default hand-off point via secondary sales.

The investment that returns a fund is one where the founder is highly unusual, the market is unproven, the comparable does not exist, and at least two of your partners think it is a mistake: Conviction is metabolically expensive: it costs political capital inside the fund and the comfort of being able to say "we all agreed" when a deal goes to zero. The investor championing a polarizing founder is making a personal bet that their judgment is worth more than the room's average judgment, and most of the time that is a losing trade.

A "legibility tax" operates at every layer of the institutional stack: the LP wants to know why you backed it, the partner wants to know how to defend it at review, the associate wants to know how to write the memo: At each translation layer, a little of the original conviction leaks out. Founders who pattern-match cleanly survive translation. Founders who do not die in the translation layer. Eventually, the investor who originally believed stops bringing those founders into the process because the cost of championing them exceeds the cost of staying quiet.

Every firm that talks about being conviction-led and then runs a 10-partner unanimous-vote process is producing consensus portfolios that will generate benchmark returns: Once a fund's brand becomes a constraint (the cost of looking foolish rises, the willingness to look foolish falls), the firm transitions from being an underwriter of bets to a curator of bets. Curators do not produce outsize returns because they are rewarded for taste that is broadly recognized.

About half of all acquired startups have raised less than $10M on the day they get snapped up, and fewer than 10% have raised $100M or more: Raising a lot of venture money narrows the potential pool of acquirers while simultaneously raising the bar for what constitutes a "good" exit. At some point the choice becomes IPO or bust.

You cannot realistically IPO at a billion-dollar valuation today. It is probably a $3B minimum, more like $5B in practice: This creates a widening valley between acquisition exits and IPO exits. Startups that have raised too much to make an acquisition attractive but too little to reach IPO scale can wander this valley for years, trapped by their own capital structure.

Walker acknowledges that US startups could technically go public at lower valuations (or in non-US markets) but doubts this will happen in meaningful numbers: The practical effect is that founders underestimate the impact of overcapitalization on exit strategy. Every additional round of funding narrows your options, and the narrowing is not linear but accelerating.

A software company that grew revenue 65% in 2024 (the median) only grew 28% in 2025, and the average company ran 5 core GTM channels plus another 5.5 channel experiments simultaneously: The volume of GTM work keeps piling up to achieve the same or worse results. The winners are shifting from one-off tactics to recurring campaigns that compound within their best channels.

The highest-ROI channels shift dramatically by stage: pre-$1M ARR founders should lead with LinkedIn, warm outbound, and founder brand; $1-10M ARR companies should prioritize warm outbound, LinkedIn, and intent-based outbound; $10M+ scaleups benefit most from large conferences, SEO, and paid ads: At early stages, all three top channels compound into each other: share a point of view on LinkedIn, generate pipeline from followers, and repurpose the best-performing content off-platform.

The data comes from a joint report with Maja Voje that includes real company examples of founders who "bent the growth curve back up": The practical insight is that growth endurance (how much of last year's growth rate you retain this year) is the metric that separates winners from the rest. Top-quartile companies maintain 80%+ growth endurance vs. the 43% median.

HBS leads with 88 MBA unicorn founders, followed by Stanford GSB (58) and Wharton (39), but the MBA ranking does not mirror the undergraduate one: Stanford leads in undergraduate alumni founders; Harvard leads in MBA alumni. Whatever HBS is doing at the graduate level, it is producing a different kind of pipeline. Columbia (25), MIT Sloan (19), Chicago Booth (18), UCLA Anderson (17), and Berkeley Haas (16) round out the top tier.

The ranking measures post-MBA ecosystems as much as the programs themselves: Most founders build companies after graduation, within the networks, cities, and job opportunities their MBA gave them access to. The question is whether these programs teach founders how to build companies, or whether they provide the credential, network, and capital access that enables building.

School size is a major confound: Harvard's MBA class is more than twice Stanford's, making a per-capita ranking essential for fair comparison: Ilya VC notes this follow-up analysis is coming soon, which will likely shift the rankings significantly. On a per-capita basis, Stanford GSB is almost certainly the most productive MBA program for unicorn founders.

The prevailing prediction six months ago was that Claude Code and Codex would let companies deliver their roadmaps with a fraction of the engineers. The opposite is happening: R&D headcount share has climbed 3.4 percentage points in just over two years. Companies are hiring more engineers, not fewer, as AI expands the scope of what is buildable.

The trend likely undercounts the actual shift: engineers and data scientists are increasingly hired into HR, Sales, and Ops to bring AI into those functions: The "Rest of Business" category is quietly becoming more technical. Traditional non-technical departments are absorbing technical talent as AI integration becomes a cross-functional priority, blurring the line between R&D and the broader organization.

This mirrors the Redpoint finding that software engineer job postings have recovered to near-baseline (97 on an indexed basis): The convergent signal from Pave's headcount data and Redpoint's job posting data confirms that AI is expanding the demand for engineering talent, not contracting it. The 10x productivity narrative has not translated into headcount reduction at the macro level.