What investors expect from an AI company before the next round

Investors in the next round look at evidence the company starts building in the first month after this one. They read the revenue, how long customers stay, the monthly money records and, for a software company, the code. This page sets out what investors at Andreessen Horowitz (a16z) have written they look for, what Carta's data says about the later rounds in 2026, and the checks lawyers list for the code in a company sale. It is part of the guide on building your AI product after you raise.

The revenue bar for AI companies has risen

In June 2025, Olivia Moore and Marc Andrusko of a16z wrote: "Pre-AI, a common benchmark for best-in-class enterprise startups was $1 million in ARR in its first 12 months" [1]. ARR, annual recurring revenue, is the revenue a company earns each year from subscriptions and other repeating contracts. An enterprise startup sells to businesses.

They then report what they see in AI companies: "The median enterprise company in our sample set reached more than $2 million in ARR in its first year, raising a Series A just nine months post-monetization" [1]. Post-monetization means after the company started charging. For companies selling to consumers, the figures were higher: "Median consumer companies performed even better, reaching $4.2 million in ARR and raising an A round within eight months" [1].

These figures describe a16z's own sample, which it calls "hundreds of companies" seen over 18 months, with no exact count [1]. They are one firm's view of the companies it saw.

a16z figure (June 2025)	Value
Common benchmark before AI, best enterprise startups	$1 million ARR in the first 12 months
Median enterprise AI company in a16z's sample	More than $2 million ARR in year one
Time from first charge to Series A, enterprise	Nine months
Median consumer AI company in a16z's sample	$4.2 million ARR
Time to Series A, consumer	Within eight months

What if the company has no revenue yet? The same article says investors look for evidence elsewhere: "If not yet in live commercial traction, then certainly in shipping speed and product iteration" [1]. That means how fast the company releases new versions and how much each version improves. Keep a dated record of every release and what it changed, because that record is the evidence.

Retention and churn

a16z also looks at whether customers stay. The article says: "When evaluating companies at the Series A stage, we often have no more than 12 months of usage and retention data" [1]. Retention is the share of customers or users who keep using the product over time. The same article warns that "rapid top-line growth won't be enough to compensate for low engagement or high churn" [1]. Churn is the share of customers who stop paying.

For an AI product, measure retention from the first paying customer. Show it by cohort, which is a group of users who started in the same month, so the investor can see whether each new group stays longer than the one before.

What a16z asks for in a data room

A data room is the shared folder of documents an investor reads before investing. In a 2022 guide, Justine Moore of a16z listed "the top five things we recommend including" [2]:

Item on a16z's list	What it is	What to add for an AI product
1. Pitch deck	The presentation used to raise	Claims about the AI that match what the product does
2. Cap table	The list of who owns what share of the company	Every option grant and SAFE, signed
3. Historical P&L and burn	Profit and loss and cash spent, month by month	Model fees and hosting shown as their own lines
4. Usage data	How customers use the product	Usage by cohort, with no group left out
5. LTV / CAC and payback period	What a customer is worth over time against the cost of winning them, and how many months it takes to earn that cost back	Cost to serve each customer, including model fees

The first two columns are a16z's list. The third column is this guide's suggestion. A SAFE (simple agreement for future equity) is a contract for early investment that turns into shares later.

The same guide says a16z wants to hear "the key milestones you're looking to hit in the coming 12-18 months (and what you'll need to get there), but we don't expect a fully-baked model" [2]. A plan that names each milestone and the money it needs answers that request.

The warning signs a16z lists

The 2022 guide also lists what makes investors doubt a data room [2]:

• "Numbers that aren't consistent with what's in the deck."
• "Numbers that aren't consistent across tabs or spreadsheets."
• "Limited historical financials."
• "Selectively presented metrics."

On retention, it adds: "When you present retention or engagement data, don't cherry-pick your best cohorts of users" [2]. Show every group of users, including the weakest.

One practice addresses all four: one set of numbers, kept every month from the start and used for the deck and the spreadsheets alike, with every group of users shown. A company that builds its records this way from the first month after the raise has little to fix when the next round starts.

Investors check records, and an SEC case shows why. In June 2024 the SEC charged Ilit Raz, founder of the AI recruiting startup Joonko. The SEC alleges that "When an investor grew suspicious of Raz's claims, Raz allegedly provided the investor with falsified bank statements and forged contracts in an effort to conceal the fraud" [6]. Those are allegations in a complaint. The page describing your AI to investors and customers covers the case and others like it.

The later rounds in 2026, from Carta's data

Carta's report on the first quarter of 2026 gives the state of the rounds a seed or Series A company will face next [3]:

Carta figure, Q1 2026	Value
Down-round rate	11.4%, "the lowest figure in nearly four years"
Series B primary pre-money valuations, year over year	Up 17.2%
Series C primary pre-money valuations, year over year	Up 12.5%
Primary rounds with a liquidation preference above 1x	2.7%
Series B round counts	"stable for two years"
Hardest stage to raise	Series C

A down round is a round priced lower than the one before it. A pre-money valuation is the price of the company before the new money comes in. A liquidation preference sets how much investors get back first if the company is sold, and "above 1x" means more than the amount they invested. Carta writes that "Series C remains the hardest stage to raise in the current market" [3].

David Cahn of Sequoia Capital went further in December 2025. He wrote that startups in 2025 "coined the idea of a '$0 to $100M' club of rapidly scaling AI companies; in 2026, we'll begin to talk about the '$0 to $1B' club" [4]. That is a prediction about the fastest-growing companies. A seed company should plan against the a16z and Carta figures above.

The code is checked too

Investors and buyers check the codebase, and the open-source code inside it is part of what they check. Open-source software is code published under a licence that lets others use it, with conditions. Lawyers at Morgan Lewis, writing in June 2026 about due diligence for a company sale, list these checks [5]:

• "Buyers frequently assess whether a company maintains a written OSS policy governing the use of OSS." OSS is open-source software.
• "One aspect of due diligence typically conducted requires a scan of the company's software."
• Diligence includes "a review of the company's practices relating to the use of software licensed under copyleft licenses, such as the GNU General Public License (GPL), Affero General Public License (AGPL), Lesser GPL (LGPL), and similar licenses." A copyleft licence can require a company that distributes the code to publish its own changes under the same licence.
• "Companies may maintain inventories of OSS components to assist with identifying applicable licenses, assessing compliance obligations, and understanding software dependencies."

They add that the absence of a policy could be a warning sign and, at the least, "requires further investigation and discussion with the company" [5].

Those checks were written for the sale of a company, and the same records answer a technical review in a fundraise. Write a one-page open-source policy now, keep a list of the open-source parts the product uses with their licences, and run a licence scan before the data room opens. The guide on preparing for technical due diligence and its page on the engineering data room cover the full list. For the investor's side, see AI startup due diligence.

Reveneau's view

Reveneau's view is that the evidence for the next round should be produced by the way the company works every week: one set of monthly numbers, a dated release record, retention by cohort, and code whose ownership and licences are written down. Built that way, the data room is a copy of records the company already keeps. For an AI product, add the eval suite and its dated results, which show how the product's answers scored on each test run. The page who owns code a partner or an AI writes covers the ownership records.

Reveneau is an AI software development consultancy with offices in San Francisco and Vancouver. Reveneau says AI writes all of its code and every change must pass an eval suite written from the specification, which is graded by Jev, TypeSafe AI's decision model. It says it takes responsibility for the whole project through production and after release. To plan a build around your next milestone, see AI development at Reveneau or contact Reveneau.