By Ellie McDonald
Everywhere you look, venture headlines imply that seed rounds have meaningfully changed shape.
Yann LeCun raised $1 billion for a company that didn’t exist a week earlier. Project Prometheus launched with $6.2 billion out the gate. Unconventional AI hit $475 million two months after founding.
It’s easy to read those headlines and conclude the venture model has been rewritten, that AI is a once-in-a-generation opportunity requiring once-in-a-generation capital.
We disagree. And so does the data.
The biotech parallel
At Bison Ventures, we’ve built deep domain expertise in biotech, the sector with the longest history of mega first rounds in venture.
Biotech mega-seeds are common because the science requires it, you can’t run a Phase 1 trial on $3 million, but the return profile is often humbling. Large first rounds in biotech have produced a handful of strong outcomes for first-check investors … and a very long tail of modest ones. Our experience with this trend in biotech motivated us to compile a dataset and pressure-test our intuition more broadly.
We pulled every publicly available $100 million-plus first round we could find over the last 15 years (roughly 200 deals) and found that only 20% had recorded exits. Of those, only a few delivered what we’d call a venture-like return: 10x MOIC or better for the first-round investor. In other words, approximately 1% of companies that publicly raised $100 million or more in their first financing round generated returns that justify the asset class. Capital intensity, as it turns out, actually worked against venture outcomes.
That distribution will improve with a few well-placed AI outcomes this year. OpenAI and Anthropic alone will essentially double the number of outlier returns in this data set when they exit. But even there, the return math is nuanced for first round investors. According to reports, first-round investors are looking at 30-40x returns at OpenAI’s projected IPO valuations.
That’s a fantastic outcome, but it’s also a fraction of what early institutional investors made on the generational outcomes of prior eras.
Sequoia Capital and Kleiner Perkins each turned roughly $12.5 million of their Google checks into around $4 billion, driving reported returns somewhere north of 300x. First Round Capital reportedly turned a roughly $500,000 investment in Uber into $2.5 billion — nearly 5,000x.
These are exponentially larger outcomes. Why? The difference wasn’t a byproduct of company quality but of entry price. Those historical investors got in at a price that left room for the upside to actually compound.
The mega round is real, but not replacing the market
The number of $50 million-plus seed rounds has exploded since 2018. But traditionally sized first rounds are also growing. The headline-grabbing rounds are a small fraction of what’s actually getting funded, and an even smaller fraction of what will return venture-scale capital.
Moreover, the companies people now hold up as AI winners started small, only further reinforcing this point.
Cursor‘s first round was less than $10 million. ElevenLabs‘ was $2 million. Legora‘s was $11 million. Sierra‘s was $25 million. Even at the frontier-model layer, Cohere‘s first round was $5 million. Today, every one of those companies is valued north of $5 billion and generating hundreds of millions in revenue.
Cursor at less than $10 million is the more representative data point. Project Prometheus at $6.2 billion is the exception.
Capital intensity is not a moat
Raising a massive first round doesn’t inherently make a company more likely to generate venture size returns for its investors. Sometimes it’s a necessary cost of doing business, but the venture math is unforgiving.
High entry prices leave less room for the upside to accrue, regardless of the underlying opportunity. The playbook that has worked across every prior technology wave is to buy meaningful ownership in capital-efficient companies at prices that leave room for the upside.
That playbook doesn’t make for dramatic headlines in 2025. But it’s what the historical data, from Google to Uber to Cursor, consistently vindicates.
A few of today’s mega-seeded AI companies will absolutely deliver 10x-plus MOICs, just as a few winners have in every era. But the data’s been consistent for 15 years, and building a portfolio around the exceptions, rather than the pattern, is a bet with a long losing track record.
Ellie McDonald is a principal at Bison Ventures, where she draws on a decade of infrastructure and technology investing experience as well as a systems engineering background to support exceptional entrepreneurs building the next generation of frontier technology companies. Prior, McDonald was an investor at G2 Venture Partners, where she focused on growth-stage climate tech companies. She began her career in Barclays‘ power and utilities group and then at Morgan Stanley Infrastructure Partners, where she developed deep expertise across energy, infrastructure and project finance.
Related Crunchbase query:
Related reading:
- The Largest Recent Seed Rounds Are All For AI Companies
- In Charts: Seed Deals Keep Getting Bigger As Odds Of Reaching Series A Fall Dramatically
- Data: The Seed Funding Boom Is Concentrating Capital In The San Francisco Bay Area
- Seed Funding Is Bigger Than Ever — And Harder To Get
Illustration: Dom Guzman
Stay up to date with recent funding rounds, acquisitions, and more with the Crunchbase Daily.
67.1K Followers
Facts Only
* Yann LeCun raised $1 billion for a company.
* Project Prometheus launched with $6.2 billion.
* Unconventional AI hit $475 million two months after founding.
* Biotech mega-seeds are common due to scientific requirements, but returns often yield modest outcomes.
* Of publicly available $100 million-plus first rounds over the last 15 years, only 20% had recorded exits.
* Few of these exiting companies delivered venture-like returns (10x MOIC or better for the first-round investor).
* Historical investors achieved returns significantly larger than current projections based on early entry prices ($300\text{x}$ for Google, $5,000\text{x}$ for Uber).
* Companies like Cursor raised less than $10 million, contrasting with the mega-rounds.
* The author notes that in historical contexts, large initial investments provided greater upside potential.
Executive Summary
Venture funding headlines suggest a significant shift in the venture model, particularly concerning seed rounds for AI companies. Recent examples include Yann LeCun raising $1 billion and Project Prometheus launching with $6.2 billion. The article contrasts this narrative with data from the biotech sector, where large first-round investments often result in modest returns, with only a small percentage generating venture-like outcomes. The author suggests that high capital intensity in initial funding may work against venture performance, as historical benchmarks show that massive early exits (like those seen with Google or Uber) involved entry prices that left greater room for compounding upside.
The narrative shifts when examining the distribution of funding sizes versus company outcomes. While headline-grabbing rounds exist, the majority of funding activity involves smaller companies, such as Cursor or ElevenLabs, which are now valued in the billions. This suggests that while mega-rounds are occurring, they represent an exception rather than the general rule for venture success across all technology waves.
Full Take
The central tension in the argument lies between observed headline capital events and historical performance data regarding venture returns. The observation that high entry prices diminish the compounding effect of upside—as demonstrated by analyzing historical outcomes from Google or Uber—challenges the assumption that an explosive funding event automatically guarantees superior venture returns for the initial investors. The pattern suggests that while exceptional outlier results are possible, relying on this exception as a guide for future investment strategy is risky.
The implication for building portfolios is a pivot toward identifying capital-efficient opportunities rather than chasing sheer funding size. The focus should shift from the *scale* of the first round to the *entry price relative to potential*. If today's mega-rounds are indeed anomalous, the historical data suggests that systemic risk remains in prioritizing volume over ownership economics. This requires a critical re-evaluation of what constitutes a "venture-like return" across evolving technological paradigms and recognizing that current high valuations may be an artifact of different capital dynamics than those that generated generational wealth in previous eras.
Bridge questions: If the historical pattern favors lower entry prices for compounding, how should investors quantify the risk differential between investing in demonstrably exceptional outlier companies versus following the established rules of capital efficiency? What data is needed to determine if the current AI momentum represents a true paradigm shift or merely a temporary statistical anomaly influenced by unique market conditions?
Sentinel — Human
The text presents a reasoned, experience-based argument about venture capital distribution and capital intensity, blending specific examples with broader economic principles rather than simply synthesizing external reports.