Series A and B Investing in an AI Market: Risk, Liquidity and Timing

Harry Stebbings hosts investors Rory O’Driscoll, Jason Lemkin and Fabrice Grinda to discuss venture returns, financing risk, AI competition and the choices facing Series A and B investors.

Grinda describes the AI market in spring 2025 as a bubble and says other categories are receiving less attention. That is his investment view; the spending and funding figures he cites are estimates from the discussion.

O’Driscoll says venture investors face both high entry valuations and limited exits. He also notes that investors' complaints look very different from the perspective of founders seeking capital.

Stebbings asks whether the shortage of liquidity is cyclical or structural. O’Driscoll argues that an eventual withdrawal of private capital could change incentives to list, though he does not know when that might occur.

Lemkin recounts a portfolio company receiving a higher-priced financing offer soon after a market shock. He sees intense AI enthusiasm, but his example is an anecdote rather than a market-wide measurement.

Stebbings discloses investments in Mercor and Lovable and asks whether investors avoiding the current AI rush will miss major opportunities. Grinda says his fund still invests in AI but seeks differentiation, defensible data and viable economics.

Grinda argues rapid recurring-revenue growth can reverse and warns that model providers or well-funded competitors may enter the same categories. His examples and downside scenarios are investment judgments, not predictions about a particular company's results.

Asked how to find a major trend without taking excessive AI risk, Grinda points to the digitization of business-to-business trade. He describes opportunities in cataloging, orders, payments, financing and supply chains, while presenting market-size estimates as his own.

The panel turns to time and returns. O’Driscoll distinguishes investment selection, entry and exit valuation, and the timing of exits; he says timing is particularly hard for fund managers to control.

Lemkin cites his fund's current multiple and internal rate of return and models how a higher eventual multiple might still produce a lower annualized return if it takes longer. These are his figures and scenarios, not verified final distributions.

Grinda says his diversified strategy uses secondary sales to realize cash when valuations appear too high relative to future upside. This is his approach, and a sale also gives up the upside on the shares sold.

The guests discuss conflicts between a manager's tax position, portfolio strategy and a limited partner's desire for liquidity. Their comments describe incentives and personal choices, not tax advice.

O’Driscoll says secondary opportunities usually exist for the companies other investors want most. Selling at an unusually high price can make sense, but deciding which compounders to retain is difficult.

Lemkin prefers his biggest winners to return several times a fund, whereas Grinda explains that his diversified portfolio can build a return from many smaller outcomes. Their different desired outcomes follow from different fund construction.

O’Driscoll argues that holding strong companies can let compounding cover mistakes, while acknowledging periods when prices make selling attractive. Neither strategy can identify winners with certainty in advance.

Grinda describes his own reported mix of outcomes and secondary sales, emphasizing that its performance is specific to his portfolio. O’Driscoll says private marks lack the feedback that an exit or public price can provide.

O’Driscoll says public ownership can impose discipline and provide liquidity, but private companies often stay private while cheap capital is available. He expects incentives to change if private financing becomes scarcer.

Grinda, who has founded companies, describes public-company reporting and budgeting as costly in time and argues that founders may rationally prefer to remain private. The guests distinguish this operating burden from investors' liquidity needs.

Lemkin says smaller public companies with modest growth can struggle with analyst attention and liquidity. O’Driscoll believes a sufficiently fast-growing company should be able to list below today's effective scale threshold, but that the details matter.

They discuss leadership transitions when a startup's journey stretches over a decade. Lemkin cites his own analysis of founder CEOs at business-to-business IPOs; it is not a comprehensive study presented on the show.

O’Driscoll says replacing a founder who wants to stay is a high-bar decision, while longer timelines can create legitimate reasons for a founder to change roles. Stebbings says he looks for enduring commitment to a problem.

The guests debate whether AI can accelerate the decline of established products. Stebbings names examples as competitive concerns, while O’Driscoll treats technological obsolescence as a risk whose timing and company-level impact vary.

They discuss how startups and incumbents can respond to a platform shift. Being early can create opportunity, but a product still needs distribution, customer value and an economically sustainable model.

The panel considers which software categories may be most exposed to changing AI capabilities. The proposed divisions are investment frameworks, not established classifications or a guarantee that any company will survive.

The guests turn to ServiceNow's market performance and the question of whether established software companies can benefit from AI. Their interpretation of a share-price move reflects the recording period rather than a current quote.

They discuss whether incumbents can use customer relationships and existing data to add useful AI features. A stronger product does not automatically guarantee faster revenue growth or higher margins.

O’Driscoll discloses that his firm invested in Box and says he remains a shareholder. Lemkin and O’Driscoll view the company's AI work positively while acknowledging that its effect on growth is uncertain.

Stebbings asks why Box's market valuation does not reflect greater AI enthusiasm. The guests distinguish product value, competitive pressure and the ability to capture revenue; their projections are opinions.

They consider whether coding tools can create more value than they capture. O’Driscoll and Lemkin discuss tiered pricing as a possible way to reach many users while charging more for enterprise use; the cited prices are approximate.

Grinda argues that well-funded competitors can compete away economics even if a category becomes valuable. He prefers to wait for a clearer winner in some crowded AI markets, accepting that this approach can miss early upside.

O’Driscoll agrees capital can erode returns but is reluctant to avoid a major technological trend entirely. He describes the challenge of choosing between earlier uncertainty and paying more for later traction.

Stebbings argues his early-stage strategy focuses on exceptional founders. O’Driscoll says Series A and B investors also need evidence of product-market fit, a directionally sound market and economics that justify the price.

The panel questions whether a small amount of revenue from founders' existing networks proves product-market fit. Lemkin and O’Driscoll say investors must look closely at the quality and independence of customer demand.

O’Driscoll warns that a higher-priced round can carry risks characteristic of an earlier stage. Better selection matters more when the entry price leaves less room for error.

The discussion turns to a reported investment in Manus and the geopolitical risks of investing across jurisdictions. The guests express opinions about governance and security rather than establishing facts about any company's access to data.

Grinda describes his reluctance to invest in certain countries because of political and legal risks. O’Driscoll notes that jurisdictions can change and that investors must consider both institutions and business exposure.

The panel discusses geopolitical competition and supply chains. Their statements about possible conflicts and future policy are forecasts and perspectives, not verified predictions.

Grinda discloses investments in Anduril and Ukrainian defense startups. He argues for lower-cost defense production; the assessment of future demand and military effectiveness is his own.

The guests debate whether new defense investors have sufficient expertise or are following momentum. O’Driscoll notes that momentum-based strategies and detailed technical diligence are different approaches.

Grinda recommends that founders seek the US market, while Stebbings argues companies can build in Europe and sell to US customers. Each is giving strategic advice shaped by personal experience.

Lemkin cites founders he has worked with when describing differences in working pace. O’Driscoll challenges broad generalizations and says strong engineering teams can be located in Europe while enterprise sales often focus on the US.

The guests discuss whether AI will widen or narrow the productivity gap between engineers. They disagree, and their numerical productivity claims are estimates rather than controlled measurements.

Lemkin predicts large changes in sales and customer-success staffing and argues companies must adapt quickly. The other guests question aspects of his pace and generalizations; no workforce outcome is established.

Stebbings thanks the guests and closes the discussion.