a16z, Founders Fund and the Risks Behind Venture’s Biggest Funds

Harry Stebbings is joined by Rory O’Driscoll of Scale Venture Partners and SaaS investor Jason Lemkin to discuss AI, venture fund sizes, liquidity and investment ethics. Their figures and forecasts reflect the information and opinions presented in April 2025.

The guests begin by questioning whether success in technology or investing translates automatically into political expertise. They then turn to the changes in software investing that have made familiar metrics harder to interpret.

Lemkin recalls a period when rapid SaaS growth could attract financing almost on the strength of a spreadsheet alone. O’Driscoll distinguishes that investment style from the broader question of whether cloud software remains a viable business.

O’Driscoll says many established SaaS markets have matured while AI capabilities and customer expectations change quickly. In his experience, some young companies have gained and lost product-market fit more than once in a short period; this is an observation from his portfolio, not an industry-wide measurement.

He attributes the volatility both to model improvements and to companies still discovering how customers want to use AI. The uncertainty is not only whether a tool works today, but whether it will remain valuable when the technology or workflow changes.

Stebbings asks what investors can underwrite when revenue and product-market fit are less predictable. O’Driscoll says they are taking more risk in pursuit of a potentially larger upside, and paying too high a price can consume that upside.

Longer private holding periods add another source of uncertainty. O’Driscoll expects a wider spread of outcomes and argues that portfolio construction matters when a fund must survive more years of business and market risk.

The guests discuss SaaS companies growing at more conventional rates. O’Driscoll says he would still invest in a business with strong growth and a clear customer problem, even without an AI label, while slower-growth companies face more difficult questions about their eventual exits.

Lemkin says some investors he knows now prefer AI opportunities, based on his own conversations rather than a representative survey. The panel considers whether private equity could buy mature software companies whose growth no longer supports venture-style valuations.

Private equity and venture firms seek different outcomes. A buyer may focus on cash flow, operating improvements and acquisitions, while a venture investor typically needs some unusually large winners; a sale can therefore work differently for investors who entered at different prices.

The panel notes that a software company can still be valuable with slower growth, but its price and ability to service debt matter to a potential buyer. They caution against assuming that every late-stage venture valuation can be recovered through a private-equity sale.

Stebbings raises the reported plans for a $20 billion a16z fund and asks how a large fund can generate meaningful returns. O’Driscoll says the strategy requires enough very large outcomes or a great many smaller ones, and questions whether there will be sufficient capacity for all that capital.

He argues that companies staying private longer create more demand for capital and secondary liquidity. Limited partners may accept a lower multiple if the return and duration compare favorably with other opportunities, but that is a model and not a guaranteed outcome.

Lemkin suggests a large firm can model how much it might have invested in companies it saw early and later watched grow. Stebbings questions the premise that any firm sees every deal; O’Driscoll adds that access alone is not enough because investors must still distinguish winners from failures.

Stebbings gives an example of investors entering a model provider at a $4 billion valuation and facing heavy dilution despite a much higher later valuation. A company valuation increase is not the same as an investor’s ownership-adjusted return, and a private mark is not necessarily cash realized.

The guests praise Josh Kushner and Thrive for backing leading companies in several categories, including fintech, AI and infrastructure. O’Driscoll sees a compelling approach in the market they describe, while emphasizing that performance across a full cycle is still uncertain.

Lemkin provocatively calls seed investing a bad bargain compared with putting a large check into a later winner. His comparison depends on access, ownership, timing and the size of the exit: a lower multiple can produce more absolute dollars on a larger investment, but neither outcome is assured.

O’Driscoll points to correlated valuation risk. Companies may operate in different sectors and still fall together if investors reduce the prices they are willing to pay for growth assets; a high entry price can make the loss particularly severe.

A long holding period helps only if the company remains a winner. O’Driscoll warns against assuming all private technology leaders will keep compounding and uses older technology companies displaced by competitors as a reminder of selection risk.

He recalls a sale in which his firm received roughly 1x while earlier investors received roughly 3x. The example illustrates how the same exit can create different multiples depending on entry price; liquidation preferences do not guarantee recovery in every downside scenario.

O’Driscoll says surviving the 2000s downturn shaped his preference for a fund size and risk level that can withstand adverse markets. He prefers a durable strategy over raising much more capital merely because it might be available.

The panel cites a reported approximately $2.6 billion Wiz return against an $8.5 billion fund to illustrate scale. That reported amount is roughly 31% of the fund’s size, not the full fund, and the conversation does not establish cash realization or the complete fund return.

Fund sizes can also grow because check sizes and nominal economic values have risen. O’Driscoll says a manager must size the fund to its strategy and maintain enough positions to manage risk, rather than simply maximize the amount raised.

Stebbings questions whether a $50 million seed fund can lead enough current-size rounds after fees without becoming highly concentrated. The answer depends on check size, reserves and the manager’s objectives; the panel treats it as a portfolio-design trade-off.

The guests discuss a reported $4.6 billion Founders Fund raise and strong limited-partner demand. O’Driscoll praises the firm’s approach of concentrating in selected winners and holding for long periods, while cautioning that copying a reported result does not reproduce its decision-making.

Compounding over a long period can produce a high total multiple even when annualized returns are less spectacular. A reported fund multiple, its IRR, holding period and cash distributed to limited partners answer different questions.

Concentration can magnify successful outcomes but also magnifies the consequences of a mistake. The panel recalls a biotech investment and eventual exit as an example of taking risk, rather than proof that concentrated investments always succeed.

Stebbings asks whether more institutional capital will flow into a small number of venture firms. O’Driscoll says investors must compare the opportunity set and risk against alternatives, and that one exceptional manager’s experience cannot be generalized to every fund.

The panel discusses whether venture capital has grown proportionately with the available investment opportunities. They frame the answer as uncertain: more money can fund more ambitious companies, but competition and high entry prices may reduce returns.

One proposed check is to compare the amount invested over a period with the exits ultimately available to the industry. Even if a market is large in aggregate, individual funds still have to gain access to the right deals and convert paper value into distributions.

The conversation turns to the IPO market and the incentives for high-quality companies to remain private. A company may be able to raise capital and offer secondary liquidity without listing, but the time investors wait and the terms on which they exit still matter.

O’Driscoll compares investment returns net of fees and over time, rather than relying on a headline multiple. The discussion distinguishes a percentage annual return from a multiple of invested capital and from cash actually returned.

They discuss the treatment of liquidation preferences when a company is acquired mainly for talent. The outcome depends on contracts, bargaining and available proceeds; no investment preference is an unconditional guarantee that an investor will recover its money.

The panel debates founder secondary sales and terms in highly competitive rounds. They distinguish providing reasonable personal liquidity from arrangements that could misalign incentives or put pressure on an already limited employee equity pool.

Stebbings worries that giving a company too much capital too early can change its behavior and reduce its eventual outcome. O’Driscoll agrees this is a real consideration, though the outcome of any particular financing remains uncertain.

Discussing the dispute involving Deel and Rippling, O’Driscoll explicitly says he cannot verify the allegations and discloses that he is involved with Papaya, a company in a related market. The panel’s discussion of potential conduct and consequences is conditional, not a finding of liability.

The guests connect the controversy to due diligence and accounting discipline. O’Driscoll says annual recurring revenue can be useful as a forward-looking indicator but may rely on definitions and assumptions, while recognized GAAP revenue follows accounting rules and looks backward.

Lemkin cites a survey of SaaS respondents about misleading statements in sales. Stebbings argues there is a substantial difference between overstating a product roadmap and the more serious conduct alleged in a legal dispute; the panel does not establish facts or criminal liability.

In a quick-fire valuation exercise, the guests disagree about whether they would invest in OpenAI at a hypothetical $300 billion valuation. These are personal investment judgments at the time, not recommendations or evidence of what the company will be worth.

They consider Cursor at a hypothetical $10 billion valuation. Lemkin says strong revenue metrics could make the price attractive if revenue proves durable, but rapid switching between coding tools makes that durability an open question.

O’Driscoll concludes that exceptional companies can justify large investments, yet identifying them in advance is difficult. The discussion closes with a joke about fund sizes, returning to the underlying tension between managing more capital and keeping a focused strategy.