Does the biotech IPO market misprice innovation?

Public-market biotech investors and pharma companies overpay for innovation. Investors who buy at the IPO offer price generate returns by harvesting that overpayment: their return comes primarily from the IPO pop and from M&A deals where acquirers overpay.

At the IPO day-1 close, the market paid $1.56 for every $1 of intrinsic value the companies went on to deliver (90% CI $1.12–$2.33, p=0.007), discounted at each IPO year's cost of capital (ranging from 5.1-11.7%) ^{1, 2}. At the offer price it paid $1.22 (90% CI $0.89–$1.75, p=0.29). For the companies that were acquired, acquirers paid $1.38 for every $1 of product value their acquisitions went on to deliver, at their own cost of capital.

Those figures represent the enterprise value at IPO, rather than the return earned by specific shareholders. We also calculate the returns earned by investors in each IPO, assuming they held each stake through exit or the terminal date, and decompose that return into four components: IPO pop, M&A premium, and the residual intrinsic return.

An investor who took every offer-price allocation from 2012 to 2021 and held each stake to the company’s exit earned a pooled money-weighted 7.2% (90% [2.9%, 11.3%]), against a capital-weighted cost of capital of 8.4% (P(> 8.4%) = 0.31). Take away the M&A premium, valuing each exit at intrinsic value instead of the price acquirers paid, and the same allocation earned 0.9% ([−3.2%, 4.4%]); at the day-1 close without the premium it lost 3.4% a year ([−7.6%, 0.2%]), below the 1.9% risk-free rate (P(> rf) = 0.01). These are returns on the diluted stake that IPO investors held, and incorporate financing conditions, not just enterprise cash flows. The breakeven discount rates (r_be) in the Results section are analogous measurements of growth in the enterprise value.

IPO investors who are allocated at the offer price across all deals clear the risk-free rate only because of the M&A premium, and even with it does not reach the cost of capital, although the cost of capital is within the 90% confidence interval of the return, and the risk-free rate is within the 90% CI of the offer-price return. (P(> 8.4%) = 0.31 for the 7.2% return). Most of the M&A premium was paid above what the acquired products delivered (60% by dollars on the 47 measured deals). The day-1 buyer earned 2.0% ([−2.2%, 5.8%]) with the premium and lost money without it.

Summary of analysis

We analyzed 465 biotech companies that went public on NASDAQ or NYSE from 2009 through 2021. Of those, 431 were included in the final analysis (411 for time-series analyses, excluding the small 2009-2011 cohorts).

For each company, we took the quarterly financial statements from IPO to exit (or to Q1 2026 for companies that still trade), computed the net present value of the cash flows discounted to the IPO date, and calculated the intrinsic value of the IPO investors’ stake. For still-active companies, we use the enterprise value on March 31, 2026 as the terminal value. We compare that intrinsic value to the market value of the IPO investment, at the offer price and at the day-1 close.

For acquired companies, we recorded the revenue of the acquired products as disclosed in the acquirer’s financial statements, projected revenue and contribution margin through estimated loss of exclusivity, and discounted back to the IPO date. We take this value to be the intrinsic value of the acquired company. We compare it to the price paid to acquire the company.

We account for follow-on equity issuances that are priced above or below intrinsic value by adding a 'transfer leg'. A company that raises money at a fair price does not make its existing shareholders worse off: the new investors contribute cash equal to the value they receive, so the incumbents' ownership percentage falls by exactly the proportion by which the company's equity value rises, and their dollar claim is unchanged. However, in reality, follow-on raises are often priced above or below fair value. If an equity raise is priced above fair value, that represents a transfer from new investors to existing investors, and vice versa.

We measure this with a 'transfer leg'. For each quarter an equity raise occurs, we compute the value of the IPO block's claim, then the value it would have had if the same amount of money had been raised at the model's estimate of fair value; the gap is the transfer leg. It is zero for a raise at fair value, positive when a company sold stock above intrinsic value — existing holders gain at the expense of the new money — and negative when it sold below. The three terms, (1) the opening value of IPO investors' claim on the business, plus (2) the growth of the intrinsic value over time, plus (3) the transfers in or out, sum to the value of IPO investors' claim on the business at exit. However, where a company's pre-money value is already negative, there is no fair price to benchmark against, so any raise at a positive price is almost pure transfer; those quarters are counted and reported separately.

Analysis population

The finding that these companies are overvalued on an intrinsic-value basis is robust to the discount rate: we valued every company at fixed rates from 8% to 25%, and at time-varying rates built from the risk-free rate, the equity risk premium and the biotech industry beta from Aswath Damodaran’s data (5.1%–11.7% by IPO year).

Because biopharma products generally exhibit negative perpetuity growth (when patent and regulatory exclusivity end), analyzing the biopharma sector lets us isolate discount-rate effects from growth expectations: there is no perpetuity growth embedded in these companies. For actively trading companies, we use the enterprise value as of 3/31/2026 as the terminal value of the DCF. These values embed the market's assumptions about pipeline and platform value. For acquired companies, we model revenue only from products that were under development by the acquired company at the time of the acquisition. Despite claims that biotech platforms generate perpetual value, we find that (1) most acquisitions are of mature product companies (though this is somewhat less true after our window ends in 2021), and (2) the acquired companies that did have platforms ultimately generated nearly all their value from one or two products that were mature at the time of acquisition.

We are interested in the question of how investors price innovation, so we focus on early-stage, R&D heavy companies. We focus on IPO investments because they are the closest public-market analogue to venture funding for which quarterly financial data exist: IPOs are a common path to liquidity for venture investors, in biopharma they usually come within 3–5 years of founding, and they are primary offerings whose proceeds fund R&D. Focusing on IPO investments also gives us a discrete universe where we can analyze the entire population. The window ends with the 2021 cohort so that every company has five years of actual operations; for later cohorts, the intrinsic value would be dominated by the current market valuation.

We calculate aggregate statistics across cohorts and the full analysis population using capital-weighted and equal-weighted measures, though capital-weighting is the primary method (weighted by amount raised in IPO).

We compute r_be as a pooled, capital-weighted IRR on the enterprise, scaled to the IPO block’s entry ownership (r_be is invariant to the IPO investor’s position and is a measure of enterprise value, not a return to IPO investors), solved on the summed cash flows and TV of the companies in the analysis population. Intervals are determined with a 2,000-draw company bootstrap.

In addition to r_be, which represents an enterprise-level metric that is agnostic to how a company's burn is funded, we also compute an XIRR to IPO investors' stake, at both the offer and day-1 close price. This metric includes information about subsequent financings and dilution, and represents the return to IPO holders as a class, rather than the enterprise as a whole.

This method cannot say whether a single company was mispriced. Most biotech IPOs are development-stage companies whose drugs are not yet approved; the probability of approval is on the order of 10% at Phase 1 and rises stepwise through development. Our method uses ex-post cash flows, where we know whether a drug was approved or not (in the case, especially for more recent vintages, when drugs developed by IPO-cohort companies are not yet approved, the terminal value does implicitly embed the market's pTS adjustment). Ex ante, the probability of success is unknown, so IPO investors rationally include a pTS discount in their valuation. A company that ultimately generated $500M of NPV could have been fairly valued at $200M ex ante with a rational probability-of-success adjustment. By analyzing the universe of biotech IPOs, the ex post analysis aggregates the winners and losers and embeds the pTS of the analysis population.

Walking through an example deal: Reata Pharmaceuticals

Reata went public in May 2016 at $11 and was acquired by Biogen for $7.3B ($172.50 a share) in a deal announced in July 2023, a few months after the FDA approved Skyclarys, the first drug for Friedreich’s ataxia. Skyclarys (omaveloxolone) is a small molecule with an Orange Book patent estate (last listed US patent covering the compound expiring in 2033; US regulatory exclusivity expiring 2030), so we model cash flows through 2033, then model typical generic erosion rates for small molecules.

We project revenue based on a kNN model trained on the launch curves of drugs with disclosed quarterly revenue that have reached a revealed peak — 149 drugs, 3,050 launch-quarter observations. Eligible drugs had at least eight quarters of disclosed revenue, a peak run-rate of at least $40M a year, and trailing-twelve-month growth below 10% at the last observed quarter. The model projects peak revenue of $1.0B for Skyclarys. The model applies a 52.93% after-tax margin (8% COGS, 25% SG&A, 21% tax rate).

Our analysis shows that IPO investors bought $74M of intrinsic value for their $60.5M IPO investment. Investors who bought in the open market after the IPO paid $72M for the $74M of value. At face value, this suggests that the IPO was fairly priced. However, these are ex post valuations that do not include a pTS discount -- this valuation knows Skyclarys was approved. At IPO, omaveloxolone was in Phase 2 (no results had been released from the Phase 2 programs at the time of IPO), and Reata's other lead program, bardoxalone methyl, was in Phase 2, with a Phase 3 planned in the near-term (bardoxalone methyl was later abandoned). Thus, at IPO, investors would have applied significant pTS adjustments to Reata's programs -- likelihood of approval at Phase 2 is on the order of 15% (BIO 'Clinical Development Success Rates and Contributing Factors, 2011-2020; Citeline 'Why are Clinical Development Success Rates Falling, 2024). To compare an ex ante price for a Phase 2 company with our ex post value that has no probability of success adjustment, we must 'unadjust' the ex ante valuation, which would result in a valuation that is much higher than the intrinsic value our ex post analysis calculates (note that, for this reason, our ex post valuation cannot say whether a single company is misvalued, but when aggregated, it can assess whether a cohort is appropriately valued).

At the time of the deal's announcement, analysts estimated peak sales in the $1-1.5B range (recall our model projects $1B peak sales), and many analysts thought Biogen had overpaid ³.

All indications are that Biogen overpaid. Using an 8% discount rate and the actual and projected cash flows generated by Skyclarys post-acquisition, through exclusivity, gives an enterprise value of $2.7B, compared to the $7B that Biogen paid. Biogen did acquire other products than Skyclarys; however, none were in late stages of development.

While the IPO was overvalued on an intrinsic value basis, IPO investors' generated high returns from 1) later investors funding the company at prices above fair value and 2) Biogen significantly overpaying for the asset, providing a windfall to investors. Reata funded its R&D by issuing stock in 13 of the 29 quarters it was public, and it sold those shares consistently above what our DCF says they were worth: of the $271M by which the IPO investors' claim on intrinsic value grew between its first reporting quarter and the acquisition, $119M was due to growth of the business (the $129M roll-forward of the stake's cash flows and terminal value — specifically, less discounting of the terminal value) and $152M was transferred to IPO investors from later investors who bought follow-on offerings at prices exceeding fair value.

Walking through the decomposition of returns

Every company’s total return is split into multiplicative legs. The below figure lets you step through the identity for Reata (or Juno): the price return, the legs it splits into, and the same legs as a dollar waterfall for the IPO stake. Hover a leg for its definition.

Flat 12% discount rate, intrinsic value measured at the IPO date. Exit price per share is as of the last trading date, rather than the announced deal price ($172.36 vs. $172.50 for Reata; $86.96 vs. $87.00 for Juno). The waterfall’s grey bar is the chain residual (Reata −$10.1M, Juno +$0.3M: the dollar legs are undiscounted while the endpoints are present values); the residual is included in the business-leg factor.

Results

The primary analysis metric is intrinsic value at IPO / market value at IPO - 1. When aggregated across companies, this represents the overvaluation or undervaluation of the sample. We report this metric at the offer price (an underwriter-determined price) and the day-1 close (an open-market cost). All aggregate figures are capital-weighted by amount raised in IPO.

We find that, using a 12% practitioners' discount rate (the median midpoint of the discount-rate ranges disclosed in fairness opinions of a sample of 38 acquired companies in our analysis population, applied in the fairness opinions to probability adjusted cash flows), IPO investors overpay at the offer price (P(≥1) = 0.04) and day-1 closing price (P(≥1) = 0.00) — P(≥1) is the share of 2,000 company-bootstrap draws in which value delivered per $1 paid is at or above 1 (one-sided).

At time-varying CAPM-based discount rates (using RFR, ERP and biotech industry beta from Aswath Damodaran's website, ranging from 5.1-11.7% from 2012-2021), IPO investors overpay significantly (P(≥1) = 0.02) at the day-1 closing price, but only nominally at the IPO offer price ($0.82 delivered per $1 paid, 90% [0.57, 1.12], P(≥1) = 0.14; t-test p = 0.29).

These statistics use a March 31, 2026 terminal date for the terminal value of companies that are still trading. Companies that are actively trading comprise 59% of the value of the sample, and the above findings are sensitive to the terminal date. Marking companies at March 31, 2025 (one year earlier) moves the time-varying CAPM discount rate, offer-price finding from investors getting $0.82 per $1 invested, to getting $0.69 per $1 (90% [0.47, 0.97]). Marking the terminal values of actively trading names at an August 2026 mark moves the time-varying CAPM, offer-price finding to $0.91 delivered per $1 invested (90% [0.62, 1.24], P(≥1) = 0.29), and also flips the 12% discount rate, offer-price finding from significance to non-significance (0.74, 90% [0.49, 1.04], p=0.07).

We compute a breakeven discount rate (r_be), the discount rate that sets market value of IPO shares equal to the intrinsic value. We find that r_be over the pooled cohorts is 2.5% at the IPO day-1 closing price, and 6.0% at the IPO offer price. The risk-free rate (the 10-year treasury yield at IPO date, calculated across all cohorts on an IPO-capital-weighted basis), is 1.9%. The time-varying discount rate for the period (using 10-year treasury yields, equity risk premium, and biotech sector beta from Aswath Damodaran's website) is 8.4%. Nominally, neither IPO investors at offer price or day-1 close price return r_be greater than the discount rate. At the day-1 close price, r_be cannot be distinguished from the risk-free rate (2.5%, 90% [−2.2%, +7.0%]; P(r_be > rf) = 0.53).

Below summarizes the breakeven discount rates computed at IPO day-1 close and offer price. We compute it with two bases: 'intrinsic', which is the sum of the discounted cash flows and terminal value, with terminal value being equal to 1) market value for still-trading companies, 2) present value of actual+projected revenues * margin from acquired drugs for acquired companies with disclosed product revenue, 3) market-value pre-deal announcement for acquired companies with no disclosed cash flows, 4) market value at last day of trading for delisted, bankrupt, or reverse-merger companies. The 'price acquirers paid' basis uses the last market value prior to deal close (which includes the acquisition premium) for all acquired names. Intervals are a 2,000-draw company bootstrap. rf = 10-year Treasury at each IPO year, capital-weighted (1.9%); cost of capital = each IPO year’s CAPM rate, capital-weighted (8.4%). 2012–21, n = 411.

The offer-price row differs from the day-1 row only by the IPO pop. The strategic rows differ from the intrinsic rows only by what acquirers paid over intrinsic value (delivered product value for 47 acquired names, pre-acquisition announcement market value for 72).

Are IPOs mispriced?

While r_be is lower than the discount rate across the window, this does not necessarily imply the IPOs were mispriced. Realized returns can differ from ex-ante expected returns in either direction over any given window without implying mispricing, as stock returns exhibit high volatility and estimates of returns have large standard errors. The precision of an expected-return estimate depends on the calendar span, not on the number of observations inside it: sampling more finely, or adding companies, within the same time horizon, does not increase precision about the mean of the expected return (Merton 1980). Ten cohorts share one ten-year systematic path, so n = 411 has a much lower effective n. The 90% intervals we show are a company bootstrap: they measure how much each figure depends on which names are in the sample, not how far a realized rate can sit from an ex-ante required return. The latter represents additional uncertainty beyond the intervals shown, and is not quantified here.

We compare the IPO returns to the XBI benchmark to assess relative performance (this controls for the sector's performance over the time period, but does not control for other factors, such as size, quality, etc.). An investor who put the same capital into the XBI on the same IPO dates earned 5.7% holding to March 2026 and 4.4% holding each stake to that company’s own exit date on the day-1 weighting ($63.1B); 6.0% and 4.8% on the offer weighting ($49.4B). Since each IPO stake is valued at its own terminal date, the matched-exit column is the like-for-like comparator.

We compare the XBI performance to the IPO allocation holder’s realized return (not r_be):

The excess is measured as a paired quantity, not by subtracting the two columns: the same dollars go into XBI on the same IPO dates and come out on the same exit dates, and the 2,000-draw company bootstrap resamples both legs together, so the interval is on the difference. XBI weighted by the day-1 block for the day-1 rows and by capital deployed for the offer rows.

On fundamentals alone, the IPO allocation underperformed the sector it was drawn from by 4.0 points at the offer price and 7.9 at the day-1 close, and in both cases the bootstrap excludes zero (P = 0.03 and P = 0.00). It beat the sector only through the M&A channel and only at the offer price, although that cell is not distinguishable from zero (P = 0.81, interval spanning −1.8 to +6.2pp). The excess over the sector was paid by acquirers and by later buyers.

We perform a similar paired excess return analysis on each cohort. Below, each cohort’s IPO allocation holders are measured against the same capital as the IPOs, invested into the XBI on the same dates as the IPOs, and withdrawn on the same exit dates as the IPO:

Excess in percentage points of money-weighted IRR. Capital in parentheses is the offer-price column’s; the day-1 column deploys $63.1B.

The pooled excess is +2.4pp (offer × price acquirers paid), −2.5pp (day-1 × price paid), −4.0pp (offer × intrinsic) and −7.9pp (day-1 × intrinsic). Only the first is positive, and its bootstrap interval spans zero (90% [−1.8, +6.2], P(excess > 0) = 0.81); the two intrinsic cells do not (P = 0.03 and P = 0.00). So the IPO cohort did not match the sector: it underperformed it on fundamentals, and reached it only through the M&A channel and only from the offer price.

Actively trading companies are valued using a terminal value equal to the enterprise value at the terminal date. We use March 31, 2026 as the terminal date. The findings are sensitive to the terminal date, as 59% of the value of the analyzed companies is valued using market terminal value. Thus, the analysis is sensitive to the volatility of the biotech market. Moving the terminal date back one year to March 31, 2025 moves the time-varying discount rate, offer-priced value delivered per $1 paid from $0.82 to $0.69 (90% [0.47, 0.97]) and the day-1 cell from $0.64 to $0.54. Using the latest market prices (as of August 2026, when the XBI was at or near all-time highs moves the time-varying discount rate, offer-priced finding to $0.91 (90% [0.62, 1.24]) and the day-1 priced cell to $0.71 (90% [0.49, 0.98]).

Below shows the values and confidence intervals of the $ delivered per $1 invested result at both discount rate conventions and across all three terminal dates.

$ of intrinsic value per $1 paid, 2012–21 IPOs (n = 411), capital-weighted, IPO-date convention; brackets are the bootstrap 90% interval and P is the share of draws at or above 1 (one-sided; bold = below 1 at the 5% level).

We also calculate the pooled IPO investor returns at the same three terminal dates:

We also sensitize the intrinsic value used for acquired names. For acquired companies where the acquirer discloses product revenue, we calculate intrinsic value using the actual and projected cash flows generated by the acquired products through LOE (the 'product-revenue terminal value basis'). For acquired companies where the products fail clinical trials, are discontinued by the acquirer, or are withdrawn from the market, we value $0 terminal value. For acquired companies with no disclosed revenue data, or no approved products, but that still have ongoing development programs (or, if the acquired company has at least one product with disclosed revenue and at least one major unapproved product), we use the enterprise value 42 trading days before acquisition announcement as the intrinsic value.

For acquired products valued using the product-revenue terminal value basis, we aggregate the terminal values, and divide by the aggregated price paid to acquire the companies, to get a realized value / price paid ratio.

How much of the realized-value ratio is our projection?

85% of the product value in the ratio is projected beyond the quarters the acquirer has disclosed, so the ratio is sensitive to the projection model.

We also investigated whether the IPO valuations were rational ex ante, with probability of success adjustments. Based on each company's lead-asset stage at IPO and industry approval benchmarks, the typical company's lead asset's likelihood of approval was ~20%. Day-1 prices are not consistent with any probability of that order. Among the 47 IPOs that went on to deliver a marketed product, the value those winners ultimately delivered was 0.75× the day-1 price, capital-weighted at 12%: even conditioning on eventual commercial success, day-1 buyers paid more than the winners delivered, so no success probability can justify the price. Only at an 8% discount rate do the winners cover their price (1.36×), and there the price implies a success probability of 0.74 — still roughly 4× the base rate.

Mispricing by vintage

We observe meaningful variation in mispricing by IPO year (each cohort is aggregated by weighting by IPO proceeds). Only two IPO cohorts, 2016 and 2017, delivered more intrinsic value than the price paid for IPO shares. Within any one year the median company delivered close to nothing, so the cohort ratio is set by how many dollars its two or three winners returned against how many dollars the rest burned.

Capital-weighted = Σ intrinsic value of the IPO stakes ÷ Σ day-1 market value of the same stakes, at 12%. “Median company” is the median of (intrinsic ÷ price − 1). “Without it” removes the largest winner from both sums. 2009–10: 20 names, 14 of them with non-positive delivered value; the pooled realized value is negative, so the ratio is reported as <0.

The above table shows that some cohorts delivered just pennies on the dollar, even though the cohorts contained large winners. This is because, using our ex post methodology, many companies have negative intrinsic value -- they go public, burn hundreds of millions of dollars, and then shut down. Ex ante, these companies had some probability of success, and a higher intrinsic value. But because our ex post analysis has perfect hindsight and knows which companies win and lose, the ex post analysis does not have probability of success adjustments. Therefore, many companies have negative enterprise value.

The example of Atara can help illustrate this. Atara's IPO stake cost $55M and is worth about $1M today. The company burned $1.97B over eleven years and is worth $82M of enterprise value. A DCF using those numbers clearly shows a negative enterprise value. An equity investor can never lose more money than they invest. However, in our calculations it is possible for the intrinsic value of an IPO investors' stake to be less than (in absolute value) the amount raised in IPO. In Atara's case, the intrinsic value claim of IPO investors is −$237M. Our intrinsic value measure values the business, less net debt, multiplied by the ownership of IPO investors. For ATRA, this includes the IPO investors' 25% share of everything the enterprise went on to generate, including the burn and the terminal value. Under the assumption that later companies are funded at fair value, future equity raises are value-neutral to existing equity investors. In practice, financings are often done at premiums / discounts to fair value. We account for this with the 'transfer leg'.

The transfer leg reconciles the −$237M intrinsic value of IPO investors' stake to the $1M outcome. Later investors paid $318M above fair value for shares in a company worth less than nothing (31 of Atara's 35 raising quarters were at a negative pre-money). Mispricing at entry excludes that transfer — a mispricing measure should not credit a future mispricing — while the intrinsic-growth leg absorbs it, and across the panel it nets to −$5.2B on $49.4B of offer capital (+$29.5B to incumbents, −$34.8B away).

Pooled, follow-on pricing was not uniformly generous but two-sided and outcome-sorted: buyers post-IPO overpaid into the names that went on to fail and underpaid into the eventual winners, netting to a modest −$5.2B against IPO-block incumbents on $49.4B of offer capital. Follow-on buyers moved +$10.5B to incumbents in the names that returned under 0.25× and −$15.4B away from incumbents in the names that returned over 3× (argenx alone −$3.7B).

The two positive years are driven by four companies

The biggest winner in the entire pool was argenx, which went public in 2017. Its IPO stake cost $135M and is worth $5.8B today, a 43x multiple on IPO investment, and ARGX alone is 80% of every dollar the cohort's winners delivered in excess of the price paid. Excluding ARGX, the 2017 cohort delivered $0.39 per dollar of day-1 value — against a break-even of $1.00, and in line with every other year in 2012–21; removing KRYS as well, the second biggest winner, takes it to +$0.05.

2016 is a $1.5B cohort, the second-smallest in the window after 2012; BeiGene and Protagonist together account for 63% of its positive dollars, and without the two of them the cohort delivered $0.71 per dollar. But BGNE and PTGX were typical winners rather than outlier winners like ARGX: 2016 produced seven winners returning $2.31B above cost, which is comparable to the number and value produced by winners in other years (2014 $2.67B, 2018 $2.26B, 2021 $2.44B, and a largest name of $759M against $0.7–1.5B in 2013–21 outside 2017). The 2016 cohort was undervalued because this 'typical' level of positive value was against a low denominator of total IPO dollars invested, and because the winners were priced relatively cheaply (BGNE at $187M of day-1 value, PTGX at $86M). Further, only 57% of capital in the 2016 vintage went to losers, which is the lowest rate in the window (2012: 60%); the 2013–21 cohorts otherwise put 71–90% of capital into losers. Neither 2016 nor 2017 was priced accurately on an intrinsic value basis; the median company in each was still valued at roughly twice what it delivered.

The largest winners

Ranked by dollars delivered above the day-1 price (intrinsic − market). “Delivered” is the present value at the IPO date, 12%, of everything the IPO stake receives: the burn it funded, the dilution it suffered, and its share of the terminal value. These twelve are 52% of all positive dollars in the panel (top ten: 47%). Prometheus carries a market terminal value because Merck has not disclosed post-deal revenue for a product that was still pre-approval at the close.

Most large winners still trade

Of the ten largest winners by dollars delivered, six are companies that have not been acquired (argenx, Revolution Medicines, Krystal, BridgeBio, Axsome, BeiGene); four were bought (Auspex, Endocyte, Prometheus, Acceleron). For a still-active company the terminal value is its market enterprise value at March 31, 2026, so its “intrinsic” value includes whatever the market currently believes about its pipeline. This analysis attempts to separate market sentiment from intrinsic value, but for these material public names, much of the value does embed the market's assumptions. Just under two-thirds of all positive dollars delivered in the panel (63% of $30.6B, the gains over cost of the 77 companies that delivered more than they cost, 2009–21) come from still-trading companies, against 27% from acquired companies valued on disclosed product revenue. We address this limitation by sensitizing the terminal date (as discussed above).

How much of this rests on a few surviving winners

argenx holds 17.8% of the cohort’s measured fundamental value on 0.2% of the day-1 capital; the six largest still-trading names hold 37.8% on 3.1%. Removing argenx alone moves the day-1 mispricing from −48.1% to −57.2%; removing those six takes it to −66.7%; removing all 178 companies still trading takes it to −70.4%. The effective sample size behind the dollar-weighted average is about 77 of 411 companies, and argenx is absent from 37% of bootstrap resamples — which is most of why the intervals around these figures are wide.

The offer price $ delivered vs. $1 invested at time-varying CAPM discount rates, which has a point estimate of $0.82 received for $1 invested and is not statistically significant (compared to the result using the other conventions for this analysis, including offer price at 12% fixed discount rate and the day-1 closing price at time-varying discount rates, all of which are significant) $0.69 with p = 0.01 once argenx is removed. The fact that the exclusion of one company flips the significance of this result further supports that the insignificance of the $0.82 result is not evidence that the IPO price is fair; rather, it is a failure to reject the IPO price being fair.

¹ Measured at IPO-day closing price.

² Using risk-free rate, equity risk premium, and biotech industry beta from Aswath Damodaran's website.Rates by IPO year run 5.1% (2021) to 11.7% (2019); each cohort is valued at its own year’s rate. Both cells in this row are the IPO-date (C2p) convention, like the rest of the table; the earlier day-1 value (0.60) was the first-reporting-date convention.

³ FiercePharma, July 28, 2023: https://www.fiercepharma.com/pharma/biogen-ponies-73b-rare-disease-specialist-reata-and-potential-blockbuster-skyclarys

Data: Tiingo fundamentals and prices; SEC XBRL; acquirer product disclosures; FDA Orange Book. Universe: 472 US biotech IPOs 2009–2021; 411 with valid day-1 legs 2012–21. Discount rate 12% flat with an 8–25% grid and per-year CAPM alternative. M&A premium anchored 42 trading days before announcement (Schwert 1996).