A letter to limited partners.
As an LP, you should be cautious when a VC says it has picked a category winner in an AI company that is six months or a year old. A business that young may be growing extraordinarily quickly on genuine demand from customers who have yet to make a renewal decision. Calling it a winner asks you to accept a conclusion about durability while the evidence needed to support that conclusion is still developing.
I understand the commercial incentives facing VCs as they deploy capital, compete for founders, and cultivate relationships with investors who can bring them into rounds led by another firm. A reputation for backing winners also helps attract LP commitments to future funds, giving a manager another reason to promote its portfolio companies as category leaders. When assessing that reputation, you should examine the evidence that the companies can sustain their success.
Brad Barber and Ayako Yasuda’s research on private equity and venture funds found that interim performance influenced both the likelihood of raising a follow-on fund and its size, with managers tending to fundraise around peaks in interim performance. That relationship helps explain the commercial value of promoting an early success while much of its eventual investment outcome remains unresolved. For an LP considering the next fund, it makes the quality of the evidence behind the performance story especially relevant.
Consider a startup that uses LP capital to develop its product, hire a sales team, offer discounts, and help customers get started. As revenue grows, another investor may finance the business at a higher valuation, giving the original manager an impressive investment to feature in its next fundraising presentation. That sequence can reflect legitimate progress and real customer value. To assess the manager’s ability to pick enduring winners, you should examine how much lasting value the company has created with the capital it received.
When assessing a fund manager’s skill, you should examine the cost and persistence of its portfolio companies’ growth alongside the price paid for each investment, giving credit where a company deploys capital unusually well. Customer renewals, the concessions required to retain accounts, and the cost of serving them help establish the quality of a rising annualized revenue run rate. A subsequent financing adds another investor’s assessment of the opportunity, with the fund’s eventual return depending on the company’s performance and the terms at which its ownership can be realized.
For AI application companies building on models available to their competitors, you should pay close attention to what happens when another well-funded VC-backed company pursues the same customers. A rival can use capital to offer lower prices, subsidize migration, and improve service, increasing the pressure on the incumbent to earn the customer’s continued business. Distribution becomes more defensible when serving customers creates something difficult to reproduce, such as trusted integration into critical workflows or a network whose usefulness grows with participation. An investor’s confidence in the customer base should reflect the strength of those relationships once buyers have meaningful alternatives.
I agree with investors who see rapid growth as an opportunity to build these advantages, because early customer access can give a capable team time to improve its product and become harder to replace. Venture investors often have to commit capital while important questions remain open, and the price paid and ownership acquired need to compensate for that uncertainty. You should ask the manager for a potential fund you want to invest in to explain how the company intends to turn its early lead into durable category leadership and what progress would support that expectation.
The dot-com era offers an example of how investors can come to accept rapid expansion as evidence that a company is securing its market. In a study of 788 internet businesses, Brent Goldfarb, David Kirsch, and David Miller found that 48% survived through 2004, a survival rate comparable to other emerging industries. Their analysis examined the widespread belief that getting big quickly would establish lasting competitive advantages, an assumption investors embraced before there was sufficient evidence that the strategy would work.
The dot-com experience reinforces why you should examine the entry price as closely as the growth story. A successful company bought at a price that assumes years of exceptional execution can deliver a disappointing investment return. I see that same speculative assumption when a non-frontier AI company’s growth rate is used to imply that its competitive position is already secure. For an LP, eventual distributions also have to justify the fees paid and the years the capital was committed.
The customer’s economic benefit is central to underwriting those expectations. In an October 24, 2026 note, Apollo chief economist Torsten Slok argued that AI’s financial payoff remained concentrated, pointing to rising revenue per worker and margins among the Magnificent Seven and flat results across the remaining S&P 500 companies. His assessment gives LPs a reason to examine where the customer’s economic benefit is showing up and how much of that value an AI vendor can retain. Several of these S&P 500 companies are paying customers of the non-frontier AI vendors that VCs are backing and promoting as category winners.
Slok’s figures describe aggregate corporate performance; evaluating an AI application also requires evidence from the specific work it improves. A published study by Erik Brynjolfsson, Danielle Li, and Lindsey Raymond found that AI assistance increased issues resolved per hour by 15% on average among 5,172 customer-support agents. To translate those productivity gains into an investment case, you should examine how much customers are willing to pay and how long they are likely to keep paying. Those revenues have to support attractive vendor economics after competition and delivery costs.
For LPs, I believe there is a place for speculative exposure when commitments and liquidity needs are planned around the range of possible outcomes. An allocation to early AI companies should allow for exceptional returns as well as substantial losses, further capital requirements, and long waits for distributions. You should also examine how much risk is shared across funds, since several managers can concentrate your exposure to similar application companies whose prospects depend on customer retention, follow-on financing, and generous exit valuations.
When evaluating a manager, you should connect their portfolio companies’ revenue growth and financing marks to the operating results beneath them, beginning with renewals under competitive pressure and the cost of retaining those customers. The same review should establish how much of the reported fund return has reached LPs as cash and how lower exit multiples and further dilution change the expected returns on unrealized positions. Those answers help establish how much confidence the operating and investment results support.
For a younger fund with limited distributions or renewal history, you should ask the manager to explain what remains unproven, the evidence that would change its conviction, and how an investment could earn an attractive return as one of several successful competitors.
I believe AI is going to create seismic change, and some of the companies making important work better will retain enough of that value to reward their investors. For the non-frontier AI companies VCs are backing today, that outcome will depend on whether customers keep paying and the economics hold up as competition develops. As an LP, you should evaluate a six-month-old company as a speculative investment and judge the manager by the quality of its underwriting and the returns it ultimately delivers.