Based on data from PitchBook’s “Quantitative Perspectives: US Market Insights,” Q3 2026
I worked my way through PitchBook’s latest quarterly report, “Quantitative Perspectives: US Market Insights,” in its Q3 2026 edition, with data as of June 30, 2026 (July 31 for the most recent series). It struck me that the best way to render its richness was to stage a fight. In the ring of American private markets, two heavyweights have always faced off: venture capital (VC), an asymmetric bet on innovation, and the buyout (LBO), a machine for steady returns built on debt. The report’s numbers make it possible to referee the duel—in an unprecedented environment where the war in Iran has rekindled energy inflation, futures are once again pricing in rate hikes, and AI is absorbing an outsized share of capital. The verdict, in five rounds.
Round 1—Performance: advantage buyout, but VC is getting up
Over ten years, the PitchBook indexes leave little doubt: middle-market buyout stands at 443.3 (indexed to 100 in 2016), ahead of PE growth at 403.5 and buyout megafunds at 358.5. Early-stage VC, at 357.1, beats the buyout megafunds but still trails the core of the LBO market; late-stage VC follows at 332.0.
The interesting part is the recent dynamic: after more than three years of stagnation, VC returns are showing signs of life and reaccelerating sharply—a rebound driven by the repricing of AI companies. PE, by contrast, is disappointing relative to its potential: PitchBook’s model, which estimates the performance PE should deliver given market conditions (equities, credit, macro), points to strong return potential—yet the returns funds have actually delivered have run several points below it since 2022. Portfolios are failing to convert a favorable environment.
But the harshest referee is the comparison with public markets: measured as annualized excess return over rolling five-year windows (PitchBook simulations), buyout generates +0.29% above its listed equivalent, while VC destroys -1.41% a year. Round to buyout.
Round 2—Risk and dispersion: two different sports
The median hides everything. Across the 2002-2019 vintages, middle-market buyout posts a median net IRR of 14.9%, with a top decile at 30.5% and a bottom decile still positive at 3.4%. VC offers a median of 11.9%—but a spread running from +31.8% (top decile) to -4.3% (bottom decile): it is the only major strategy whose bottom decile loses money in absolute terms.
The report’s Monte Carlo simulations drive the point home: across 100 simulated private portfolios, VC delivers an annualized return centered around 5.5-6%, versus roughly 7-7.5% for PE, at comparable volatility—hence a markedly lower Sharpe ratio (~0.45 versus ~0.60). In plain terms: in aggregate, VC takes more risk for less return, and only exceptional manager selection (or luck) flips the equation. Round to buyout, decisively.
Round 3—Liquidity: two groggy boxers, but VC is bleeding more
This is the round where nobody shines. The structural problem of private markets—distributions—hits both camps. On the buyout side: TTM distributions have fallen to 13.8% of NAV, nearly ten points below the historical average (23.7%); the median holding period of portfolio companies has hit a record 4.2 years, and 20.5% of the inventory has been held for more than seven years. Worse, the deals bought at the highest prices are exiting the slowest: within the 2020-2022 cohort, only 19.4% of deals in the top multiple quartile have been sold, versus 34.9% for the cheapest quartile—the entry prices of the 2021 era are now being paid for in years of holding. Dividend recaps, the liquidity crutch of 2024-2025, have just fallen back as well.
On the VC side, it is darker still: distributions are stuck at 5.4% of NAV, the lowest yield of any private asset class. The relief valve lies elsewhere: the direct secondary market, whose trading volume—on an annualized run rate—has grown from $50 billion to $107 billion between late 2024 and mid-2026. But that liquidity carries a brutal truth-telling price: companies last financed in 2019-2021 trade there at median discounts of 54% to 60% to their last round. America’s 964 unicorns carry a combined $5.4 trillion in valuation, a backlog that the IPO window, “partially open” despite SpaceX’s success in June, will not absorb any time soon. Round even, two knockdowns.
Round 4—Concentration: the same poison in both glasses
Both strategies are converging on the same flaw: megafund dominance. In VC, funds above $1 billion captured 68.2% of capital raised in 2026, an all-time record, while the number of active investors keeps shrinking (5,158 versus 10,452 at the 2021 peak) and new entrants are disappearing (543 versus 2,859). And a third of the market is a single bet: AI accounts for 36% of deals and 34% of invested value, with an off-the-charts long-term trend score.
In buyout, the same slope: megafunds are absorbing a growing share of total fundraising that has been declining since the 2023 peaks. Yet PitchBook’s EV bridge data show that the largest funds do not create more operational value—in large vehicles, up to 88% of value creation comes from revenue growth alone, with almost nothing from margin expansion. Small funds outperform, but raise less and less. In both camps, the money goes where the brand is, not where the alpha is. Round even.
Round 5—The 2026 momentum: VC wins on points
The macro context is reshuffling the deck. VC deal indicators have moved back above their long-term trend, in both count and value, and the PitchBook Dealmaking Indicator shows an early-stage market that has turned startup-friendly again (39.6, below the neutral threshold of 50)—unseen since 2021, and even more so in AI (36.6). Buyout, for its part, enjoys macro conditions that PitchBook’s model rates as supportive again, but activity remains below expectations, and the renewed rise in bond yields (futures now price Fed hikes in late 2026) is driving up the cost of leverage—the yield to maturity on B-rated loans still stands at 8.1%. Finally, PE’s expected risk premium (5.6% over 10 years) has dropped below its average since 2012 (7.1%): higher rates let allocators hit their targets without moving up the risk curve, which mechanically reduces demand for the riskiest assets. Round to VC, on momentum.
Outside the ring—the referee is doped on AI
One counterpoint seems indispensable to me before rendering the verdict: you cannot read a fight without watching the referee. And the referee—the public markets that serve as the benchmark for all these comparisons—is himself under the influence. In the second quarter of 2026 alone, the S&P 500 gained 15.2%, the Nasdaq 21.6% and the Russell 2000 21.5%: records set despite the war in Iran, 3.5% inflation and expectations of rate hikes. A performance of that magnitude, in that context, can only be explained one way: the rally is carried by a handful of AI-linked stocks, not by the market as a whole.
Market observations confirm it unambiguously. The ten largest capitalizations now represent 41% of the S&P 500—an all-time record, fourteen points above the 2000 dot-com peak: of every dollar invested in the index, nearly 50 cents goes to AI-linked stocks. The “Magnificent Seven” alone weigh roughly 34% of the index (more than $22 trillion in market value), versus 12% in 2015, with Nvidia alone above $5 trillion on quarterly revenue up 85% year over year. A simple test captures the phenomenon. The classic S&P 500 weights each company by its size: the AI giants therefore count enormously. There is an “equal-weight” version, in which each of the 500 companies weighs exactly the same (0.2%), giant or not: it measures the performance of the average American company. Over 2023-2025, the classic S&P 500 gained +68%, versus only +34% for its equal-weight version. Conclusion: the average listed company advanced only half as much as “the index”—all the rest comes from the weight of the AI giants.
To my eyes, this concentration has three direct consequences for the fight. First, it raises the benchmark: the excess returns of PE and VC are measured against indexes propelled by a single theme; beating an AI-doped Nasdaq is a higher bar than beating a diversified market—part of VC’s -1.41% is also explained by the unusual strength of its yardstick. Second, it artificially props up the VC camp: the recovery in unicorn valuations and the partial reopening of the IPO window rest on the multiples of listed AI stocks; it is the same bet, carried from one market to the other. Third, it creates a common risk: if listed AI stocks correct, VC simultaneously loses its revaluation engine and its exit door, and PE sees its benchmark come down—but its exit multiples too. The pillar holding up both boxers is the same one, and it is fragile.
The judges’ decision
On points, buyout keeps its belt: better ten-year performance, better median, contained dispersion, positive excess return over public markets where VC is structurally negative. But it is a victory without panache: record holding periods, distributions at half their historical norm, 2021 deals encysted in portfolios, and value creation at the largest funds reduced to a bet on revenue growth.
VC, staggered since 2022, is the boxer getting back up: returns recovering, deal market above trend, a secondary ecosystem finally institutionalizing its liquidity. Its fate, however, hangs on a single bet—AI—and on an IPO window that only opens halfway. For the allocator, the report’s lesson fits in one sentence: the asset class matters less than the manager (the top-to-bottom decile spread is 27 points in buyout and 36 in VC), and in 2026, fund selection has never been more truly the real fight.
Key concepts
| Concept | Explanation |
|---|---|
| PitchBook simulations (excess return) | The problem to solve: you cannot compare a private fund directly to a stock index, because the investor does not put the money in all at once—they commit an amount that the fund calls progressively (capital calls) and then returns as it sells (distributions), over 10 to 15 years. The method: PitchBook reconstructs the full experience of a typical institutional investor. Starting point: a classic 60% equities / 40% bonds portfolio, of which 20% is progressively reallocated to closed-end private funds. Each year, the virtual portfolio commits to new funds—drawn at random from the PitchBook database—at the pace needed to maintain the 20% target. All cash flows are real and historical: capital calls, distributions, fees, as the funds actually charged them. The measure: to judge the result, every dollar called by the private funds is compared with the same dollar invested on the same dates in the equivalent public index—the “public market equivalent.” The “excess return” is the annual gap between the two, smoothed over rolling five-year windows: +0.29% for buyout means the average investor earned 0.29 points more per year than in the stock market; -1.41% for VC, that they lost 1.41 points a year versus a simple index investment. The limitation: random drawing describes an investor with no selection skill; an LP with systematic access to the best funds would do far better (see the dispersion in round 2). |
| Monte Carlo simulation | The principle: replay the same investment scenario many times, letting chance vary with each draw, to obtain the full range of possible outcomes rather than a single average. In practice, four steps. (1) Define the output variable: for instance, “the portfolio’s value in 10 years.” (2) Model the uncertainty: each uncertain variable gets either a probability distribution calibrated on history, or—the more robust method, the one PitchBook uses—a direct draw from real data: existing funds are picked at random, with no mathematical distribution assumed. (3) Write the mechanics that, for a given draw, unfold a complete path: commitments, capital calls, distributions, rebalancing. (4) Run the machine 1,000 or 10,000 times and store every result—a few lines of Python or even Excel suffice for a simple case; PitchBook stopped at 100 draws because each simulated path is heavy. The reading: plot the histogram of results and look at the median (the likely outcome), the extreme percentiles (the lucky and unlucky scenarios) and the probability of falling below a threshold. Here, the verdict: even replaying history 100 times, the average VC investor finishes behind PE, barring a rare stroke of luck. |
| Sharpe ratio | Return earned per unit of risk taken: subtract the risk-free return, then divide by volatility (the size of the jolts endured along the way). The higher it is, the better risk is paid. A Sharpe of 0.60 (PE) versus 0.45 (VC) means that at comparable jolts, PE pays each unit of risk noticeably better: VC is not just less profitable on average, it is less efficient. |
| PitchBook VC Dealmaking Indicator | A composite index measuring the balance of power between investors and startups. It aggregates the supply/demand balance of capital (the number of startups seeking to raise versus capital available at funds and nontraditional investors) and the contractual terms of rounds: the frequency of investor-protective clauses (cumulative dividends, participation rights) and the time elapsed between rounds. 50 is the neutral point: above it, the market favors investors; below it, founders. Early-stage at 39.6 and AI at 36.6 mean founders are once again dictating terms—unseen since 2021. |
| Yield to maturity on B-rated loans | The all-in cost of the debt that finances LBOs. B-rated leveraged loans are the standard financing of a buyout; the yield to maturity is the lender’s total annual return if the loan is held to repayment—and thus, symmetrically, the full cost of acquisition debt for the fund. Today’s 8.1% breaks down into roughly 3.8% of base rates and 4.3% of credit spread (below its historical median of 4.8%): it is the level of base rates, not the risk premium, that is making leverage expensive. |
Coverage periods: what the numbers actually measure
| Building block | Period covered | As of |
|---|---|---|
| PitchBook simulations (excess return, round 1) | 2000 → 2025, a quarter century spanning three crises (dot-com bust, 2008, the 2022 rate shock); the rolling five-year curve is shown from 2005 to 2025 | Dec. 31, 2025 |
| Monte Carlo simulations (round 2) | Historical funds in the PitchBook database, 1995 to 2024 vintages | Dec. 31, 2025 |
| IRR dispersion (round 2) | 2002 to 2019 vintages only—younger funds have not lived long enough for their IRR to be meaningful | Dec. 31, 2025 |
| Performance indexes (round 1) | 10-year window: 2016 → Q1 2026 (preliminary) | Mar. 31, 2026 |
| Activity data (deals, exits, fundraising, secondaries, unicorns) | Current flows | Jun. 30 / Jul. 31, 2026 |
Why this matters. The 2000-2025 simulation period encompasses buyout’s golden age (2009-2021, the zero-rate era) and two VC winters (2001-2004 and 2022-2024). VC’s -1.41% therefore largely reflects the post-2021 correction: on a window ending in 2021, VC would have shown a positive excess return. This is a classic fragility of this kind of study—the chosen end point weighs heavily on the verdict—and it cuts both ways: buyout’s flattering numbers benefit, for their part, from a decade of cheap debt that will not necessarily repeat.
Sources: PitchBook, “Quantitative Perspectives: US Market Insights,” Q3 2026—figures 30-77 (data as of June 30 / July 31, 2026; index performance as of March 31, 2026, preliminary for Q1 2026). The simulations and nowcasts cited are those of the PitchBook methodology described in the report. This article is an analytical synthesis and does not constitute investment advice.
