Avlok Kohli on Startup Funding, Fund Liquidity and AngelList’s AI Workflows

Harry Stebbings asks AngelList CEO Avlok Kohli how the startup funding market has changed. Kohli draws on the funds and startup investments administered by AngelList. This conversation was published in May 2023; its figures and forecasts describe that period, not current market conditions.

Kohli says AngelList supports thousands of funds and syndicates and receives investment documents on their behalf. That gives the company a broad view of its own portfolio, although its data is not necessarily representative of every venture investment.

He describes a quarterly chart with two measures: the share of startups receiving an investment and the share of those financings that are up rounds. A high reading on both measures indicates an active market with rising valuations; low readings suggest fewer deals and fewer up rounds.

On AngelList's measures, he says the market moved rapidly from the 2021 and early-2022 boom toward a much weaker period. He calls the distance and speed of the shift unusually large within the data the company tracks.

Harry asks why pre-seed and seed prices still feel high to investors. Kohli distinguishes those early rounds from Series A, B and later financing, where he reports sharper declines in deal volume and valuations at the time.

He suggests that some investors who previously invested later moved earlier, bringing more capital into pre-seed and seed. His expectation was that those prices would remain comparatively resilient, perhaps with some compression. That was his forecast, not a current valuation guide.

Harry adds his own explanation: junior investors writing early checks for multistage funds may be less sensitive to price as their partners manage existing portfolio problems. This is his interpretation of investor behavior, not a conclusion demonstrated by AngelList's data.

Harry asks whether available Series A deals suffer adverse selection because the best companies may raise from existing investors before reaching the wider market. Kohli says some companies cannot raise at all, but AngelList's data does not establish a quality difference between companies led by insiders and those with outside leads.

They define a party round for the discussion as one in which no single investor supplies more than 40% of the checks. Kohli says the share of such rounds relative to rounds with a lead looked similar in AngelList data, even as total volume contracted.

What had disappeared, he says, were some companies' repeated small extensions that never reached the original target. Investors could still assemble a round without a formal lead, particularly at smaller pre-seed or seed sizes, but a larger round was harder to complete that way.

Kohli reports that time to raise and time between valuation markups were increasing. AngelList compared a recent sample with its historical data back to 2015, rather than only comparing with the unusually busy 2021 period.

In that comparison, he says roughly one-third fewer companies than expected had raised by the observed point. He interprets it as evidence of a financing freeze and repricing, while noting that founders were still adjusting to a different valuation environment.

Asked whether repricing had already appeared or whether companies still had cash, Kohli says both were true but emphasizes the latter. He reports more flat rounds, down rounds and pay-to-play structures in early 2023, while saying many companies had extended runway through layoffs or other changes.

His expectation was that pressure would become more visible in later quarters as cash-out dates approached. He contrasts a startup's finite runway with a fund manager's ability to ask limited partners for more time to deploy capital. This is a forecast made in the interview.

Harry asks whether the downturn will create a wave of cheap acquisitions. Kohli is skeptical: public companies face pressure over cash generation and stock compensation, and private companies may disagree about the valuation used in an acquisition.

Even acquiring a startup for little or no purchase price is not costless. Harry and Kohli point to the expense of retaining people, integrating products or intellectual property, and taking management attention away from the acquirer's own work.

Kohli says AngelList's observed merger-and-acquisition activity had declined and predicts it may remain subdued for the next six months. He sees the longer horizon as uncertain and dependent on interest rates, inflation and wider market conditions.

He explains that higher interest rates make venture capital compete with other uses of capital, including government debt. His comparison is a description of the allocation choices he saw in 2023, not a recommendation to buy any asset.

The discussion turns to fund secondaries and limited partners seeking liquidity. Kohli says AngelList had seen more interest in selling positions, including positions in well-known companies, for reasons ranging from portfolio allocation to immediate cash needs.

He says transactions often involve one limited partner purchasing another's fund position. In AngelList's observed examples, discounts to the last marked value ranged from about 20% to more commonly 40–50%. These were selected transactions and historical marks, not a universal price for a fund position.

Harry notes that other secondary data showed even larger discounts, then asks about microfunds whose previous individual backers may no longer be writing checks.

Kohli separates individual and institutional commitments in AngelList's data and distinguishes rolling funds from traditional venture funds. He reports a smaller decline for rolling-fund commitments, while individual commitments to traditional venture funds had fallen much more from early 2022.

One surprise in his data was that institutional commitments to venture funds fell and then recently recovered. Kohli and Harry discuss greater macroeconomic certainty and interest in the AI technology cycle as possible explanations; Kohli calls these hypotheses.

The number of newly formed funds raising capital was down, Kohli says, while some more experienced managers were raising larger amounts. He also describes the interval between a fund's first dollar and its first close increasing from around three months in the boom to more than six months.

He says new and smaller managers were also encountering smaller fund sizes than they might have raised previously. The figures are AngelList's observations during the interview period, not a promise about any particular fundraising process.

Finally, Harry asks how AngelList uses large language models internally. Kohli describes fund administration work spanning legal documents, setup, filings, portfolio management and distributions, much of which traditionally required people to classify and route information.

As an example, he says AngelList receives large volumes of time-sensitive email for funds. An automated workflow can inspect email and attached documents, identify items such as a share purchase agreement or capitalization table, and detect missing documents.

If a required capitalization table is missing, the workflow can request it. Kohli says years of operational data helped the team develop this process. His example describes an internal workflow, not evidence that all legal or investment judgment can be automated.