Everett Randle on Venture Fund Size, AI Margins and Founder Partnerships

Harry Stebbings interviews Everett Randle, a newly appointed Benchmark general partner, about venture investing and AI economics. Randle previously worked at Founders Fund, Bond and Kleiner Perkins. He discusses his current and former firms' investments and competing strategies, so his assessments carry financial and professional interests.

Asked what he learned from Mary Meeker, Peter Thiel and Mamoon Hamid, Randle says successful investors can use very different methods. He describes Meeker as using financial models to visualize how a company and its customers may evolve over many years, rather than treating numbers as an end in themselves.

He says Thiel designed Founders Fund to test conviction, including a program allowing employees to invest personally alongside the firm. Stebbings points out that younger employees may have little spare cash. Randle acknowledges the concern and says some junior staff used unsecured credit lines; his own positive outcome is not a recommendation to borrow for investment.

Randle describes intense disagreements at Founders Fund investment meetings as, in his experience, part of a culture of seeking the best answer. His description is a former employee's perspective and does not establish how every employee experienced that culture.

From Mamoon Hamid, he says he learned the value of seeing excellent founders and management teams up close early in a career. Randle associates Hamid's investing style with products that combine enterprise use with strong user engagement, citing Figma, Glean and Rippling as examples.

Stebbings asks about a time when expertise or fixed assumptions obscured an opportunity. Randle recalls passing on an OpenAI round at a reported $32 billion valuation while at Kleiner Perkins because he worried about corporate structure and dilution. He says the risks were real but believes he underestimated the product and its growth.

Randle predicts OpenAI could raise at a trillion-dollar valuation and compares it with Anthropic. He favors OpenAI in the host's hypothetical comparison because of ChatGPT's consumer reach, while noting Anthropic's enterprise and coding strength. These are speculative valuations and competitive judgments; Kleiner Perkins invested in Anthropic.

On Cursor, he argues that a declining share of a rapidly expanding coding market can still support growth. The market-size, share and annual recurring revenue figures he offers are rough estimates without disclosed sources. Benchmark is an investor in Cursor, a relevant interest in his assessment.

Stebbings asks whether AI makes every category a large new market. Randle says it remains to be seen: some verticals may be too small, while others can expand if software performs work previously purchased as labor. His comments are a framework for evaluating markets, not proof of their future size.

He describes an AI receptionist company in which Kleiner Perkins invested and recounts customers comparing its cost with their broader ServiceTitan spending. He says it could support calls outside staffed hours and reduce some staffing needs. This is an investor's anecdote, not independent evidence of typical customer savings or net employment effects.

Randle wants a different way to assess AI application businesses. Traditional SaaS benchmarks often emphasize high gross-margin percentages, retention and low capital spending; AI inference can raise cost of goods while a product potentially captures a larger customer budget.

His proposed comparison is absolute gross profit per customer as well as the percentage margin. A larger contract at a lower margin can produce more gross profit dollars, but it may also have different customer-acquisition costs, capital needs and risks. The figures in his comparison are illustrative, not audited company results.

Randle says investors should consider potential long-term margins rather than insist that early AI companies match mature SaaS immediately. He expects usage to produce inference expense, while acknowledging that the margin profile in five to seven years is uncertain. High or low margins alone do not prove product usage or future profitability.

He uses AWS as an example of a large infrastructure bill that can generate substantial dollars even at a lower percentage margin than software. He explicitly does not know its exact gross margin and supplies estimates; his suggested standalone value is speculative.

Randle says he underestimated AI cloud providers because he thought they were reselling a commodity. Strong demand changed his mind. The market capitalizations and demand he cites are his time-specific observations, not evidence that the business model will sustain its returns indefinitely.

Stebbings asks how to distinguish durable growth from momentum. Randle recalls Jasper's rapid rise and subsequent contraction as an example of revenue that can prove less sticky, while saying he believes the company later improved. His account is not an independent review of its financial history.

He says application companies must offer enough value beyond a model provider's own product to justify their price. In his view, deeper workflow integration and customer relationships can provide differentiation. His description of Jasper's competitive response is an investor's assessment.

Randle disagrees that distribution has replaced technology as the main moat. Building a high-quality AI application requires careful integration, workflow design and skilled people, he argues. Distribution still matters, but he sees talent and product execution as important sources of differentiation.

Stebbings asks whether Benchmark can participate in multibillion-dollar lab rounds given its smaller fund. Randle says a firm's fund size and team structure shape which investments it can pursue. Benchmark may forego some large financings while seeking higher multiples from smaller positions in other companies.

He says five investments in Benchmark's latest fund were marked at roughly 60 times the entry price for one investment, 30 times for two and 20 times for two. These are last-round paper marks he reports, not distributed cash or audited net fund returns; future values can rise or fall.

The host notes that ownership dilution changes the realized multiple on a highly valued model company. Randle agrees that Benchmark's goal is strong cash-on-cash returns, but comparisons based on hypothetical dilution and recent private marks are not evidence of final investor outcomes.

Does Benchmark need a lab investment to stay relevant to founders? Randle says the firm must monitor its network and access, while pointing to relationships with founders at Sierra and Mercor. His confidence in those relationships is a Benchmark partner's perspective, and both are investments discussed by the firm.

Asked about ownership discipline, Randle says Benchmark aims to be a meaningful founder partner and generate high multiples for limited partners rather than insist on a fixed stake in every round. He says Mercor involved less than the traditional high-teens or 20% holding but expects it to be valuable; the final return is unknown.

Stebbings asks whether a firm known for replacing founders can also claim to be their best partner. Randle disputes a simplified characterization and says individual board decisions have context. He favors founder leadership but argues board members must act when laws or ethical boundaries are crossed.

Stebbings adds that boards can also fail shareholders if they avoid difficult conversations merely to preserve founder approval. Both favor candid challenge in the boardroom. Their comments are general governance views, not findings about a named company's board or allegations against a specific founder.

Randle says Benchmark's partners have different preferred stages, even though the firm shares goals of founder partnership and strong returns. With a background concentrated in growth investing, he expects to do more A- and B-stage work initially than seed deals, while leaving room to invest elsewhere.

He admits feeling insecure about moving from growth investing to Benchmark's earlier-stage work. A colleague reminded him that investors can come from varied backgrounds. Randle says his goal is to build judgment across stages and that he does not yet claim a proven record as an exceptional investor.

On price, he recalls assessing a SpaceX investment at a $150 billion valuation while at Kleiner Perkins. He says looking at potential market size, position and downside rather than the size of the headline number helped his decision. The investment and his hypothetical upside are not a guarantee of return.

He uses a simple expected trajectory as a yardstick: what outcome do current investors appear to be underwriting, and does his view of the company differ? He cautions that exceptional successes can exceed models. The method still depends on assumptions about growth, valuation and risk.

Asked to rank people, product and market, Randle chooses that order. He sees people as shaping product and says young companies can change markets through pivots. This is his investing preference, not a demonstrated rule that market size or timing are unimportant.

Randle recalls writing about a spectrum from high-velocity investing to a smaller, high-touch model. He argues that several large firms now emphasize deploying capital quickly, while Benchmark deliberately stays smaller. That characterization is a competitor's interpretation, not a verified description of each firm's internal incentives.

Stebbings disputes applying the same label to all large firms, noting that their strategies differ. Randle concedes his description of Thrive was too blunt, but argues that very large checks can become the central economic focus of a firm. He does not present internal data from those organizations.

They distinguish absolute dollars earned from net multiples delivered to limited partners. Randle doubts that several large fund groups can generate five times net across their funds, while agreeing they may earn substantial absolute dollars. His claim is a forecast and the interview provides no complete audited comparison of those firms' returns.

Both think Tiger Global may ultimately fare better than some critics expected. Randle points to private positions and potential protection from preferred securities, but his scenarios depend on future valuations and exits. A liquidation preference does not guarantee full recovery in every outcome.

Randle says investment professionals in large teams can have narrow coverage and pressure to deploy capital, which he finds different from his ideal of the venture craft. He acknowledges that the work remains privileged; his account does not establish the day-to-day culture of every large fund.

On making his first Benchmark investment, he admits feeling the weight of the firm's history. A partner told him that an early failure could make the next decision less intimidating; he takes this as permission to recover from mistakes, not a plan to seek a bad deal.

He praises former colleagues' investing records and recalls the exuberance of a Miami event as public-market valuations began to fall around 2021–22. He says the memory made the gap between celebratory culture and changing markets clear to him. It is an anecdote, not a dated market analysis.

Randle sees parallels with previous technology cycles: many companies can decline sharply or fail even if a few create enduring value. He says Benchmark wants fund sizes and investment choices that can withstand a downturn. His expectation of exceptional future winners and a crash are predictions, not certainties.

In the quick-fire round, he revisits his skepticism of AI cloud providers and says demand changed his view. He also warns that well-funded coding companies without widely used products may struggle, while disclosing that Benchmark invests in Cursor. These are competitive views, not established outcomes.

He says Founders Fund's ability to incubate companies may give it a distinctive route to equity and returns. He admires Benchmark's history and says the chance to work closely with its partners drove his decision to join. These are personal views informed by his past employment and present partnership.

His most painful miss remains the earlier OpenAI round. Randle says concerns about structure and dilution distracted him from the wider opportunity, though he had acknowledged those risks were real. The company's eventual value and the returns from any individual round remain uncertain.

Asked about Benchmark's biggest threat, he names stasis: the firm must adapt as venture changes without losing its goals of strong founder partnerships and investment returns. He says continued access to exceptional founders matters more than preserving a rigid process.

He closes with an optimistic view that AI may support economic growth, while criticizing social-media incentives that compete for attention. His hopes for broad prosperity and social cohesion are predictions, not evidence that the gains from AI will be evenly distributed.