AI Is Lifting Venture Valuations. Is It Fixing Venture Capital?

Venture capital can look healthier in a fundraising announcement than it feels in an investor's bank account.

A large financing gives a startup more runway. A higher valuation can lift the reported value of a fund. A successful exit can eventually return capital to the people financing that fund. These are different events, and the distance between them is central to understanding the AI investment boom.

My view is that AI can produce extraordinary companies while leaving important weaknesses in venture capital unresolved. Concentrated investment, expensive entry prices, slow distributions and incentives tied to fundraising can coexist with genuine technological progress. The industry should be judged by how reliably it converts that progress into returns after costs and over a reasonable period.

TheMarker's October 3 headline and public summary raise the question of whether AI is reviving venture capital or concealing its problems. This article develops an independent analysis using the public sources listed below; it does not reproduce or purport to summarise the subscription article's full text. [1]

A record total does not describe the typical fund

The public PitchBook-NVCA overview reports that U.S. startups raised more than $400 billion in the first half of 2026, exceeding every previous full-year total. It also describes concentration in AI, large financings and established fund managers, alongside improving but uneven exit conditions. These are first-half and Q2 observations, not final Q3 figures. [2]

More recent public commentary from PitchBook analyst Emily M. Zheng puts AI at nearly 80% of U.S. venture deal value over the trailing twelve months, compared with 43.2% of deal count. The period is trailing twelve months, not one quarter. [3]

The gap between value and count matters. It suggests that the industry's largest dollar figures are influenced heavily by the size of AI transactions. It does not establish that most founders can raise easily, that most funds hold the winning companies, or that investors have realised the reported gains.

An allocator therefore needs two views of the market: the aggregate and the distribution beneath it. How much activity is concentrated in the largest transactions? How many managers are receiving commitments? Are exits spread across portfolios or dependent on a handful of exceptional assets? A rising total can be real while the experience of the median participant remains difficult.

Read the fund's whole scorecard

A limited partner, or LP, supplies capital to a fund; the general partner, or GP, manages it. Three measures help separate what has been distributed from what remains invested. The definitions below follow ILPA's glossary, with residual value expressed relative to paid-in capital. Use consistent valuation dates and accounting conventions when comparing them. [5]

Distributions can include securities as well as cash, and some may be recallable. A DPI number therefore needs a breakdown before it can be treated as spendable, permanently returned cash. Comparisons also need consistent treatment of fees, carried interest and fund borrowing. [5]

Consider an illustrative fund with $100 million of paid-in capital, $40 million distributed entirely in non-recallable cash, and $220 million of remaining net asset value. Its DPI is 0.4x, residual-value multiple is 2.2x and TVPI is 2.6x. The LP has received $40 million, even though total reported value is $260 million.

If remaining value falls to $180 million without any new cash flows, TVPI becomes 2.2x while DPI stays 0.4x. A valuation change alters the estimate of wealth; it does not reverse or create the cash already distributed. This simplified example excludes additional contributions and assumes consistent net values throughout.

The evidence points to patience being stretched

Carta's public Q2 2026 summary reports median DPI of 0.37x for its 2017 fund vintage, 0.15x for 2018 and 0.04x for 2019. These are observations from Carta's covered funds, not a census of the entire venture market. Low DPI measures limited distributions to date; it does not by itself establish the final outcome. [4]

The same release shows the 2017 vintage's 90th-percentile TVPI rising from 3.31x to 4.14x over two years, while its 90th-percentile IRR fell from 28.7% to 25.7%. Those are separate percentile statistics, not necessarily the performance of the same fund, and they should not be combined with the median DPI to invent a representative portfolio. [4]

Time deserves its own place in the discussion. In a deliberately simplified investment with one payment in and one payment out, tripling money over five years produces an annualised return of about 24.6%. Tripling it over ten years produces about 11.6%. Both are 3x outcomes, before any separately imposed fees or taxes, but they reward the investor very differently.

Young venture funds should not be judged as though they were mature income portfolios. They need time to build companies. Older funds, however, should explain why the expected realisation date has moved, what additional capital may be needed and which milestones would justify waiting longer. Patience becomes an investment decision when its price is visible.

A financing round is not an exit

New money raised by a company normally finances that company's activities. Unless a transaction includes a secondary sale, it does not pay existing shareholders for their holdings. Even when some holders sell, the transaction may provide only partial liquidity.

The last financing price is useful evidence, but it cannot answer every valuation question. Share classes can have different rights, new investors may receive protections, and a small transaction need not establish the price at which a much larger holding could be sold. A credible valuation process should explain those differences.

Similarly, an initial public offering does not instantly convert an entire private holding into distributed cash. The fund must consider the shares it still owns, any sale restrictions, market depth, subsequent prices and the timing of distributions. An acquisition headline may also include consideration that is deferred or conditional.

The practical question is a chain of events: what can be sold, to whom, on what terms, when can proceeds be collected, and how much reaches the LP after the fund's obligations? Celebrating the first step while leaving the rest unspecified makes an investment story difficult to evaluate.

Concentration can be rational and still be dangerous

Venture investing is designed to tolerate failures in pursuit of unusually large successes. Concentration in a promising technology is therefore not inherently evidence of poor judgment. A specialist can understand a market better than a broadly diversified manager, and a few successful investments can matter enormously.

The risk emerges when nominally different investments depend on the same economic assumptions. Several AI companies may rely on the same infrastructure providers, enterprise budgets or access to subsequent financing. An LP can hold multiple funds and still be exposed repeatedly to the same late-stage companies.

Look-through analysis should therefore go beyond the manager's name. Which companies appear across vehicles? How much value depends on a common business model? Could a change in compute pricing, customer adoption or financing conditions affect many holdings together?

Entry price also changes the result. A business can become more valuable while a late investor earns an inadequate return after dilution and the passage of time. Access to a celebrated company is commercially useful only if the ownership, terms and price support a credible investment outcome.

Calling every AI investment a bubble would avoid the hard work of distinguishing good companies from good investments. Treating every large funding round as proof of quality avoids the same work from the other direction.

Who carries the cost of waiting?

The effects of delayed distributions depend on the investor. An institution with ample liquidity and a long horizon may tolerate a slow realisation schedule. Another may need cash for spending commitments, liabilities or capital calls from other managers.

If expected distributions do not arrive, that investor must adjust somewhere: hold more liquid reserves, reduce new commitments, sell assets or accept additional funding costs. Even if the venture holdings eventually perform well, the timing can affect the rest of the portfolio.

For a manager, meanwhile, the economics depend on the fund agreement. Where management fees continue while assets remain invested, the manager and LP may experience an extension differently. Carried interest, GP capital invested alongside LPs and reputational exposure can provide alignment, but their practical strength should be examined rather than presumed.

This is the accountability problem I find most important. A persuasive explanation of future value should not automatically excuse repeated changes to liquidity expectations. Investors deserve to see whether the manager is sharing the economic burden of delay and whether the terms still reward the outcome originally promised.

Liquidity engineering needs transparent labels

Secondary transactions can help an investor sell an existing interest. A continuation vehicle can provide a route for some investors to exit while others retain exposure. Both can serve legitimate needs. They also require scrutiny of price, transaction expenses and who controls the terms.

A transfer from one vehicle to another under the same manager deserves particular care because the manager may influence both sides. Investors should understand the alternatives, the independence of price discovery and whether new fees or carry change the economics of staying invested.

Borrowing against a portfolio is another distinct mechanism. A loan-funded distribution can move cash to LPs earlier, while leaving debt and financing costs inside the investment structure. Cash received through financing should be identified separately from proceeds generated by selling assets.

ILPA's 2024 NAV-facility guidance, developed for private equity strategies, recommends LP advisory committee approval when these facilities fund distributions, alongside clearer disclosure. That recommendation is an industry governance position, not a claim that every venture fund uses NAV borrowing or that the guidance is a universal legal requirement. [7]

The broader principle is simple: explain where the distribution came from, what obligations remain and whether it can be recalled. Without those details, a better-looking liquidity metric can obscure the trade-off that produced it.

Accountability should change decisions and economics

It would be wrong to infer misconduct from a disappointing investment or a low distribution figure alone. Venture capital involves uncertainty, and a responsible manager can lose money. A blanket claim that there are no consequences for fund managers would also exceed what these sources establish.

The stronger argument is that consequences should be connected to controllable decisions: honest reporting, adherence to the agreed strategy, treatment of conflicts, capital discipline and the handling of extensions. Investors can demand contractual safeguards without pretending that every failed commercial judgment deserves punishment.

My preferred framework begins with an explicit explanation of how the fund expects to return capital. This should include several exit paths, assumptions about timing, follow-on funding needs and the circumstances that would lead the manager to sell earlier or stop investing.

Extensions should trigger a substantive review of economics. Depending on the agreement and circumstances, LPs could seek fee reductions, a narrower fee base, stronger reporting or independent oversight of conflicted transactions. These are proposals for negotiation, not automatic rights available in every fund.

Compensation should also be understandable across the entire fund life. Investors should know when carry can be paid, what protections address subsequent losses and how much of the manager's own capital is exposed. A long-term partnership needs incentives that survive an unfavourable period.

ILPA's updated reporting and performance templates offer a practical foundation for comparability, including cash-flow detail and gross and net measures with and without subscription facilities. Its January 2025 announcement recommended implementation beginning in Q1 2026; adoption should be verified manager by manager. [6]

The questions that would make the AI boom more credible

For a fund investor, the first question is how much reported value depends on the largest few positions, followed by how those positions were valued. The next is how the fund expects to finance companies until exit if another large round becomes unavailable.

For a founder, the relevant test is whether the capital raised supports a durable business. What proportion of growth reflects repeat customers? Do unit economics improve after serving those customers? How much additional financing is required before operations can support themselves? The right measures will differ between a model developer, an infrastructure provider and an application business.

For a fund manager raising its next vehicle, the test is whether the prior record can be separated into realised outcomes, remaining valuations and the time taken to achieve both. A compelling new opportunity does not eliminate the need to explain old commitments.

For an investment committee, the useful scenario is not only that AI adoption disappoints. It is also that AI succeeds while competition compresses margins, exit valuations decline or ownership is diluted. Technological success and investor returns are connected, but the connection depends on price and terms.

I would regard the recovery as more convincing when distributions broaden across older vintages and managers, credible transactions validate a wider set of valuations, and funding reaches viable businesses beyond the most visible names. That is a proposed scorecard for judging recovery, not a prediction of its timing.

AI may generate substantial wealth. The obligation of venture capital is to turn its share of that wealth into an attractive outcome for the investors who supplied the capital. Until that conversion is visible, higher valuations should invite better questions about liquidity, incentives and accountability.

Dor Arad (דור ארד) · Independent analysis and opinion · October 4, 2026