Atlas XIV And Reflections On The VC Math Problem
Today we announced the closing of Atlas Venture Fund XIV, a $450M investment vehicle focused exclusively on early stage biotech investing.
With the closing of Fund XIV, we’re privileged to be able to continue backing innovative new biotech companies for another fund cycle, reinforcing our science-first and seed-led venture creation model for novel therapeutics – and delivering on our mission of doing well by doing good.
As noted in the press release, this fund builds on the momentum we’ve seen across both platforms and asset plays, with multiple portfolio companies being acquired (Nimbus Tyk2, Versanis, Aiolos, Mariana) and several entering the public markets (Disc, Korro, Q32, and Third Harmonic). Just in the last two years, we’ve launched 16 new companies and recruited scores of talented executives to lead startups in our portfolio.
While the seed-led venture creation model is staying the same, there is one bittersweet change associated with Fund XIV: my long-time mentor, partner, and friend JF Formela is transitioning his role in this fund. He’ll continue to be actively involved as a partner in the firm and his existing legacy portfolio, but won’t be making new investments out of Fund XIV. He’s been a leading light in the biotech venture field, investing in essentially every major innovation wave of the last few decades (e.g., combinatorial chemistry, human genetics, functional genomics, mRNA, I/O, CRISPR, RNA editing, and many others). Thankfully he’s not actually going anywhere anytime soon, and will continue contributing to the firm in countless ways. Much of the firm’s disciplined fundraising and consistent returns in recent years can be attributed to JF’s leadership and experience.
It’s fair to say this fundraising was the smoothest, most efficient, and most oversubscribed fundraise we’ve had in the twenty years and ten funds since I joined the firm. Given the level of demand, we could have raised several multiples of what we set as our hard cap ($450M), which was kept the same size as Fund XIII in Feb 2022.
Many successful VCs continue building their AUM with increasing fund sizes. Why didn’t we?
Fundamentally, we wanted to stay disciplined and focused on the model that we continue to try to perfect: seed-led venture creation.
We also have historic scar tissue as a firm, and that institutional memory hasn’t gone away. In 1999-2001, Atlas raised three funds in quick sequence, aggregating over $2B of “early stage” multi-sector venture capital and aimed to scale the business to global proportions; we opened nine offices on several continents, had dozens of partners (and layers of partners), and invested in every area under the sun. When I joined in 2005, the “patient” (Atlas) was undergoing surgery and we came out of the decade in 2010 with a small partnership, around one table, in one office in Cambridge. In 2014, we simplified further by focusing Atlas only on biotech. The takeaway of all this history is that trying to scale venture often has negative and painful organizational consequences, and leads to firms taking their eye off the ball. We’ve resisted the urge in part because of this history, and our desire to focus on the simplicity of the model.
But, more importantly, we also conducted a deep strategic fund analysis last year as part of our continual aspiration to refine and improve our model. Being dynamic and adaptive is the key to success, so we explored what in our model needs to be refined.
Based on that strategic review, we continue to have strong conviction that our approach to venture creation aligns best with a modest fund size. We want to remain company builders focused on disciplined capital allocation and driving superlative returns, and resist the temptation to become asset aggregators in order to absorb higher fees.
The short version of the conclusion around our optimal fund size is simply the “Venture Capital Math Problem” as I’ve called it here over a decade ago. Others have also written on it extensively (here). VCs underwrite to a certain expectation of returns, and those returns are a function of average company exit value and ownership.
The longer version requires exploring some context around those two variables, as well as the implied portfolio assumptions for fund returns.
Company exit value is, of course, a function of the market cycle: in hot markets, crazy valuations and exits can happen; in cooler markets, more modest returns. Over the past decade (2015-2024), looking only at meaningful exits above $100M values:
- M&As of private biotechs have had a median and average of $270M and $600M, respectively. For R&D-stage public biotech companies, it’s been $700M and $1.3B, respectively. The averages tend to skew up to the top quartile of exits. (I limit it to R&D-stage since it takes longer than a fund cycle to bring a new drug from discovery to commercial).
- IPOs of currently-traded companies that went public between 2014-2024 have a median and average market cap today of $260M and $1.5B, where the top quartile is close to that average again as well. Top decile is $3B or so. And this only includes the IPOs still actively traded today, not the 100+ IPOs that have closed shop over the decade, so is a favorably higher distribution of outcomes.
With that as context, if you assume your deals will all be top decile, or that you’ll time the sale at a stock’s all-time high, that’s likely delusional thinking. And while there are multiple investment cycles embedded in the medians and averages listed above, the reality is the market also changes over time – so one has to take into account where structural changes are occurring around the type, value, and flavor of exits. Thinking through expected fund returns requires thinking through all of these valuation parameters. This was an essential part of our analysis.
The second key variable is your average ownership at exit. Our initial investments as seeds of co-created startups can often provide Atlas with 50-70% ownership. This goes down as more capital comes in – and the business of biotech is very capital intensive. An early M&A deal where we have 30-40% ownership can be quite attractive, but often lower exit values (e.g., Padlock or IFM). To go “long” towards billion-dollar outcomes and deeper into development obviously requires taking significant dilution. Owning 10-20% at exit can still be a very material outcome on a large exit value where one has gone “longer” (e.g., Nimbus or Akero). We work hard to maximize our ownership and to instill an equity capital efficiency in our companies to protect against unnecessary dilution – but some dilution is always “necessary” in the model when you are scaling biotechs by bringing drugs into the clinic.
Combining those two key variables gives you the picture of your “winners”: owning 10-20% of a $1-2B billion-dollar outcome for an early stage venture firm is generally seen as a great outcome today.
But how do these great deal outcomes roll up into fund size?
This is where fund math comes in as it relates to portfolio construction. Here are a few simple observations that drive home the challenge, using only general industry data not anything specific to Atlas.
A typical early stage biotech venture fund has a portfolio of 20-25 deals. Some work, some don’t. The rough Pareto or power law of venture capital is that 20% of the deals (“big winners”) drive 80% of the fund’s returns. While every vintage is different and depends on the investment cycle at work, this power law holds in the long run for venture. Biotech venture has less power law skew than tech venture, but it is still a dominating force in portfolio construction. Beyond the winners, roughly a third of the dollars invested will go into money-losing <1x deals (i.e. deal- and dollar-weighted loss ratios are 45% and 35% in the biopharma VC industry, respectively, over the past decade). The middle of the portfolio is largely breakeven or a push (1-2x). So if you want aspire to underwrite to a 3x gross return at the fund-level, ~5 big winner deals need to drive ~2.5x of your funds’ overall size (80%). Essentially, and as a general rule of thumb for driving stellar fund returns, each of these winners need to return half the fund’s size. The remaining middling winners will fill in the final 0.5x to get to a 3x fund return overall.
Therefore, for a $450M fund, five big winners need to collectively generate north of ~$1.1B in proceeds, where each one needs to return ~$200M on average back to the fund (equivalent to an ~8x on $25M invested). These exits come in the form of M&A or public positions that are gradually sold over time (and with further dilution). If the average ownership at time of exit is 15-20%, then one fund needs to have backed five winning companies worth $5-7B in aggregate market value at the time of exit. With five really good winners via M&A and IPO paths, and acknowledging the exit value ranges noted above, this value threshold is certainly doable under normal capital market cycles but is a high bar.
If we now consider a $2B fund, the VC math problem becomes much harder. The five big winners need to create ~$5B of equity value for the fund, or $1B each. At an average ownership of 15-20%, that’s $25-35B of value that needs to get created by roughly five companies in a single venture capital vintage (and during the timeframe of a 10 year VC fund).
Other than Flagship’s fund with the monster Moderna position in 2020-2022, I don’t think that’s ever been done before in the history of the biotech industry. At the very least, it would be an exceedingly rare black swan level event in biotech venture to generate that kind of aggregate value in each vintage: essentially all the big winners in one 2-3 year time period need to be in one firm’s fund. Seems unlikely.
Alnylam, Horizon, Actelion and other large Atlas legacy investments have grown into $20B+ valued companies, but they got there a decade or so after we exited those positions. The large aggregate market values described above need to happen while a venture fund still has a meaningful and active equity position.
Importantly, these larger venture funds may have different ambitions for their expected return profiles, perhaps focusing more on IRR than multiple, and those could make the math more viable. They often focus on very different part of the private investing spectrum (later stage, growth funds). But that’s not the right model for Atlas.
At the end of the day, biotech venture is a game of probabilities. Failure is a constant – of drug programs and companies alike. It’s like gravity, a constant pull downward. This is why when I started this blog the tag line was “a biotech optimist fighting gravity.”
The layered probabilities are stacked against larger early stage VC funds being able to deliver 3x+ portfolio multiples, given the expected range of exit value distributions and ownerships. This is the venture capital math problem and is the essence of why we’ve reinforced our conviction that the fund size we have is the right one for our early stage model.
So we’re going to be disciplined, respect the math problem, and stick to the strategy that works for us, our portfolio, and the patient needs we are trying to address. Looking forward to what exciting innovative new medicines Fund XIV will help advance!
