ICONIQ Pacesetter Index：AI原生B2B公司的关键财务基准

ICONIQ just published The Pacesetter Index. It replaces their Enterprise Five Scorecard, and the numbers in it sit far above what most of us have used as benchmarks for the last decade.

Before any of those numbers, read this part, because it changes how you should use all of them.

Who Is Actually In This Dataset

This is not a market benchmark. It is built from the fastest-growing companies ICONIQ can see, and it excludes almost everyone by design.

The pool: the top public software companies plus ICONIQ’s own private venture and growth portfolio companies. Quarterly financial and operating data from 2024 through Q2 2026, where available.

The filter on top of that pool: only companies that qualify as a “Pacesetter,” defined as top-quartile revenue growth over the past three years AND AI-native or AI-driven.

So: a top-growth slice of one firm’s portfolio, blended with public comps. Three consequences.

• The small revenue bands are almost certainly all private ICONIQ portfolio companies. No public software company is growing 900% at sub-$10M revenue. The public comps can realistically only influence the $100M+ column and maybe the $25M-$100M one.
• Winners are the entry requirement. Top-quartile growth is the admission criterion, so this index cannot tell you anything about failure rates, survival, or what a typical AI company looks like. It describes the front of the pack.
• No sample size is disclosed. There’s no n per band anywhere on the page. Four revenue bands, filtered to top-quartile-growth AI-forward companies, drawn largely from one portfolio. Some of these cells could be a handful of companies. The 2600% top-quartile figure under $10M especially.

ICONIQ is explicit about why they built it this way. Some of today’s leading companies sit well above aggregate medians and even top-quartile figures, so identifying top performers means benchmarking against those companies rather than the broader market.

So this is a subset of outliers. Just know what you’re reading. If you don’t hit these numbers you are not failing. But the multiples being paid right now are being paid against this table, so knowing which column you’re being compared to is worth something.

The Full Index:

Source: ICONIQ Pacesetter Index, September 2026.

Eight findings.

#1. 115% Growth Is the Median at $100M+

A $100M+ ARR company growing 115% used to be a once-a-decade outlier. In this cohort it’s the median. Top quartile is 165%.

For 15 years the aspirational growth path in B2B was triple, triple, double, double, double. Most companies used it as a target they missed. Here, doubling at $100M+ is the middle of the pack. ICONIQ’s framing is that Pacesetters grow 3-5x faster than the broader market, and that some are still accelerating as they mature.

Accelerating with scale is the genuinely new behavior. Growth decay used to be reliable enough to forecast off. You grew 300%, then 150%, then 90%, then 60%. AI-native companies are landing usage and expansion revenue fast enough to bend that curve the other way for a stretch.

#2. 900% Is the Median Under $10M ARR

The sub-$10M column: 900% median, 2600% top quartile.

Some of that is small-number math. Going from $400K to $4M is 900% and it’s a handful of enterprise logos. But the distribution still says something real about early traction in 2026. The winners are compounding roughly 10x in year one at that stage, not 3x.

What changes for founders: at $2M ARR growing 200%, you had a fundable, exciting company in 2023. In this cohort you’re below median. The bar for “hot” at seed and Series A has moved further than the bar at scale.

#3. Gross Margins Start at 55%, Not 80%

Median gross margin under $10M ARR is 55%. At $10M-$25M it’s 60%.

The old rule was simple. Under 75-80% gross margins you didn’t have a software company, you had a services business with a login page. That rule is dead for AI-native products. ICONIQ’s reasoning: compute and infrastructure costs have made gross margin a metric to actively monitor, Pacesetters often run at lower margins, and the benchmark for a healthy margin is still moving because greater usage drives both more customer value and higher cost.

The curve matters more than any single number. Margins go 55% → 60% → 80% → 75%. They recover hard through the $25M-$100M band as inference costs get optimized, contracts get repriced, and the mix shifts toward higher-value workloads.

A 55% gross margin at $3M ARR is normal now. A 55% gross margin at $40M ARR is a different conversation, because Pacesetters at that stage are at 80%. The board question isn’t “why are your margins low,” it’s “what quarter do you hit 75% and what specifically gets you there.”

#4. Gross Retention Falls to 90% at $100M+

Adding gross retention to the index is the quieter, more important change.

At $100M+, median gross dollar retention is 90%. A tenth of the revenue base churns out annually, among the best-performing companies at scale.

ICONIQ’s explanation: switching tools has gotten significantly easier, sales cycles are faster, contracts are shorter, and POCs have become the default entry point, which puts existing revenue at risk in ways NDR misses.

That’s the cost of the fast-growth story. The same conditions that let AI-native companies land accounts in weeks let competitors take those accounts back in weeks.

If your NDR is 120% and your GDR is 88%, you’re growing on the backs of your best customers while the base leaks. That’s a different business from 120% NDR on 96% GDR, and diligence will price it differently.

#5. Net Revenue Retention Peaks at $25M-$100M and Then Falls

The NDR sequence is 105% → 125% → 130% → 115%.

Two surprises in there.

First, NDR under $10M ARR is only 105% median. The consumption-pricing story says AI companies expand automatically as usage grows. At the earliest stage they don’t. Early customers are running pilots, a chunk of those pilots fail, and nobody has built the expansion motion yet. At 105% NDR pre-$10M you are at the median for the best companies in the category.

Second, it drops at $100M+. Law of large numbers plus the gross retention leak above. The 130% you had at $60M is not the 130% you’ll have at $150M, and modeling it flat will break your plan.

#6. Burn Multiple Gets Worse Before It Gets Better: 1.8x at $10M-$25M

The burn multiple sequence is 1.3x → 1.8x → 0.9x → 0.3x.

The $10M-$25M band is the expensive one. Burn multiple deteriorates there, and even the top quartile only reaches 1.6x. That’s where you’re paying for GTM buildout and compute simultaneously, before either has scaled into efficiency.

Then it collapses. 0.9x at $25M-$100M, 0.3x at $100M+, with a top quartile of 0.1x. Pacesetters at scale add a dollar of new ARR for thirty cents of burn.

ICONIQ’s read: negative free cash flow is common among Pacesetters because of AI compute needs, but they convert that burn into new ARR faster than the broad market, and neither cash flow nor growth captures that alone.

At $15M ARR with a 1.7x burn multiple, you’re on benchmark. At $60M ARR with the same 1.7x, you’re roughly 2x worse than the median Pacesetter and the round gets harder than you expect.

#7. $655K in Revenue Per Employee at $100M+

The old good number was somewhere around $200K-$250K per employee. Great companies hit $300K. The median Pacesetter at $100M+ runs $655K, top quartile $890K.

Run the headcount math. A $200M ARR company at $655K per FTE has roughly 305 employees. At $250K per FTE, the same company has 800. A 500-person difference on identical revenue. That’s the clearest financial fingerprint of AI-era operating leverage in the whole index.

The trajectory doubles as an operating plan: $75K → $115K → $225K → $655K. The step change lands between $25M-$100M and $100M+, which is where tooling investments finally convert into headcount you never hire.