What investors expect at each stage

“Is $1M of ARR enough for a Series A?” Founders ask this question frequently. The honest answer is that no single number exists. CRV wrote in March 2026 that no “single minimum” ARR for a Series A “applies across sectors.”

But investors look at the same few numbers, and published data shows the ranges. This lesson gives those ranges with their dates. It also teaches the two efficiency ratios that investors use most, and why AI companies get more questions.

Medians are not thresholds

Most benchmarks are medians. A median is the middle value when you put all the values in order. Half of the companies have more, and half have less. Thus half of the companies that raised a round had less than the median.

A median is also different from an average. Five seed rounds are $2M, $2.5M, $3M, $3.5M, and $100M:

Median  = The middle value = $3M
Average = $111M ÷ 5        = $22.2M

One very large round moves the average a lot, but the median almost does not move. For founders, the median is the more useful benchmark.

Predict first

SVB measured companies that raised a Series A in 2025. What was their median revenue over the last 12 months before the round?

Type a guess before you continue. A guess first, even a wrong one, makes the answer easier to remember.

Your guessSave my guess

What each stage must show

Each round pays for the proof that the next investors want to see. Investors at each stage look for a different kind of proof.

Pre-seed and seed: proof that customers want the product

At pre-seed and seed, investors look for signs of strong demand: usage, early revenue, fast weekly growth, and customers who stay. For the cohorts of users, a16z wants “Stabilization of retention in each cohort after a period such as 6 or 12 months.” It also wants “Newer cohorts performing progressively better than older cohorts.”

A cohort is a group of users who started in the same month. In January, 1,000 users sign up. 45% are active in month 1, 30% in month 3, 27% in month 6, and 26% in month 12. The curve becomes flat at approximately 26%. A flat curve is a sign of product-market fit, because a part of each group keeps the product.

Series A: proof that growth repeats

At Series A, investors look for revenue that grows month after month, from sales that the company can do again. They examine ARR, growth, NRR, gross margin, and burn. The lesson The funding ladder gives the round sizes and valuations at each stage.

The burn multiple

Growth is easy to buy with sufficient cash. The burn multiple, from David Sacks in April 2020, measures how much cash a company burns to get its growth:

Burn multiple
  = Net burn ÷ Net new ARR
  = $2M ÷ $1M
  = 2x

Net new ARR is the ARR at the end of the period minus the ARR at the start. Thus churn subtracts from new sales. A lower burn multiple is better. Sacks called 2x “reasonable for an early-stage startup.” If the company burns $5M for the same $1M of new ARR, the burn multiple is 5x, which Sacks called “terrible.”

Burn multiple	Sacks’s level
Less than 1x	Amazing
1x to 1.5x	Great
1.5x to 2x	Good
2x to 3x	Suspect
More than 3x	Bad

Sacks also wrote that a seed company “might have a Burn Multiple of 3,” and that after the Series A, “it might drop to 2.” The levels are for companies with ARR, and early companies have higher numbers.

Calculator

Burn multiple: cash burned for each new dollar of ARR

Set the net burn for one quarter and the ARR at the start and at the end of the same quarter.

Net burn in the quarter

$2M

ARR at the start of the quarter

$4M

ARR at the end of the quarter

$5M

Net new ARR

$1M

End − start

Burn multiple

2x

$2M ÷ $1M

Level

Good

David Sacks's levels (2020)

1. AmazingLess than 1x
2. Great1x to 1.5x
3. Good1.5x to 2x
4. Suspect2x to 3x
5. BadMore than 3x

The company burned $2M to add $1M of net new ARR: $2.00 of cash for each new dollar of ARR. Sacks calls this level good.

To get to 1.5x at this burn, the company must add $1.33M of net new ARR in the quarter. At this growth, the burn must be $1.5M or less.

Move the ARR at the end of the quarter down to the ARR at the start. The burn multiple has no value when ARR does not grow, because the burn bought no growth.

Work it out

Last quarter, a company burned $3M and its ARR went from $8M to $9.5M. What is its burn multiple?

Your answerCheck

The Rule of 40

The Rule of 40 says that revenue growth % plus profit margin % must be 40 or more. Brad Feld made the rule popular in February 2015. He got it from a late-stage investor, and he wrote that it applies to companies with “at least $50 million in revenue.” Here are two companies, with the growth first and the profit margin second:

Passes:        30% + 15%    = 45
Does not pass: 60% + (−25%) = 35

Feld used EBITDA margin. Many investors now use free-cash-flow margin: the cash that the business makes or loses, as a percentage of revenue. Under the rule, more growth can balance less profit. A company that grows 50% can lose 10% and pass.

Work it out

A company grows revenue 70% a year with a free-cash-flow margin of −40%. What is its Rule of 40 score?

Your answerCheck

The magic number is a third ratio, for sales efficiency. It is (revenue this quarter − revenue last quarter) × 4 ÷ sales and marketing costs last quarter. Secondary sources give the credit to Lars Leckie (2008). His rule: at more than 0.75, spend more on sales. Here is an example: revenue grew from $2.0M to $2.3M in one quarter, after $1.5M of sales and marketing costs. The magic number is 0.3 × 4 ÷ 1.5 = 0.8.

Why AI companies get measured differently

Many AI startups grew very fast in 2025 and 2026. Investors ask them the same questions as other startups. But for AI companies, three questions get more attention.

Is the gross margin like software?

Traditional software keeps 80% to 90% of each revenue dollar (a16z, 2015). AI products pay for compute each time a customer uses them, so their margins are lower:

• ICONIQ reported in July 2026 that the average gross margin of AI products was 45% in 2025. It projects 53% for 2026.
• Bessemer reported in August 2025 that the fastest AI startups, which it calls “Supernovas”, had an average gross margin of approximately 25%. Bessemer adds that in very competitive markets, margins “are often stretched close to zero or even negative.”

CRV writes that investors “scrutinize compute economics, usage depth, workflow entrenchment and the path to sustainable, software-like gross margins.” The simulator shows why usage is the key word. Each user pays the same flat price, but heavy users cost much more.

Simulator

Why AI products have lower gross margins

Each user pays one flat price, but each user costs the company a different amount of AI. Change the usage.

Price for one user each month

$30

AI cost for a typical user each month

$9

Other direct costs

10%

Hosting, support, and payment fees, as a share of the price.

Heavy users, as a share of all users

20%

A heavy user uses this much more AI

4x

Gross margin

Typical user

60%

Heavy user

−30%

All users

42%

Where each dollar of the price goes, all users

• AI cost
• Other direct costs
• Gross profit

A typical user pays $30 and costs $9 of AI, so the margin is 60%. A heavy user costs $39 a month in total, so the company loses $9 on each heavy user.

Across all users, the margin is 42%. To earn the same gross profit as software with an 80% margin, this product needs 1.9x the revenue.

Move the share of heavy users from 20% to 40%. At a flat price, the margin falls with each heavy user that you add. One answer is to charge by usage. ICONIQ reported that the share of AI builders with usage-based pricing increased from 35% to 42% between Q4 2025 and Q2 2026.

Do customers stay?

Fast growth means little if customers leave fast. In the ChartMogul data, the median GRR of AI-native products was 40% in September 2025, up from 27% in January 2025. For AI-native products at less than $50 a month, GRR was 23%. Bessemer writes that “early hypergrowth alone means less now than ever before.”

Is the ARR real?

The lesson ARR, run rate, and other revenue words shows how founders stretch “ARR” with contracts, pilots, and usage. Investors in 2026 ask how much of the ARR is live, paid, and under a contract that the customer cannot leave in three months.

Quiz · question 1 of 6

Test what you learned

Last quarter, you burned $3M and added $1.5M of net new ARR. What is your burn multiple, and how does Sacks rate it?

• A4.5x, bad
• B0.5x, amazing
• C2.0x, between good and suspect
• DYou cannot calculate it without the gross margin.