The ceiling moved. The odds are still low.
How often hypergrowth actually happens, measured three ways: the record holders, the cohort this report uses, and the population everyone else is in.
From the summary: “The ceiling moved dramatically, and selected venture cohorts are moving faster too. But in the broadest dataset here, fewer than 1% reached $10M within a year.” Back to the summary
The ceiling
Slack reached $100M ARR in roughly 30 months and held the record for the fastest business app in history. Wiz reset it at 18 months in 2022, and Deel ran $1M to $100M in 20 months at the same time. Both looked like once-in-a-decade events when they happened.
They weren’t. At least seven later companies have reported reaching $100M faster than Wiz’s 18 months, and the fastest AI-native cases now cluster near eight months from launch to a first reported $100M.
Months from launch to first reported $100M ARR
Approximate; most sources give month and year only. Mercor is gross marketplace revenue.
- AI-native, 2023 onward
- Enterprise sales-led AI
- Pre-AI controls
The AI-native cases occupy the top of the chart, but several of them start the clock at a relaunch or an agent release rather than at founding. And the pre-AI controls are not slow companies. Wiz at 18 months in 2022 would still be near the top of this chart today.
The band
This report uses $1M to $20M within 18 months because that is the question that started the research, and because 31 companies in the dataset have dated milestones bracketing the interval. It is not a universal definition of hypergrowth. Three other benchmarks give it context:
- $10M within 12 months measures how rare a very fast start is across the whole population.
- $100M within 18 months describes the Supernova ceiling.
- $100M in three to five years describes the Shooting Star and strong Compounder path.
Estimated months from $1M to $20M ARR, 31 companies
The selected median describes the winners studied here. It does not estimate the probability that a company with similar characteristics will produce the same result.
- Eighteen months or less
- Longer than eighteen months
The median $1M-to-$20M interval is eight months. That figure describes the winners studied here, because these companies were selected after their outcomes were already known.
The population
Cohort studies measure everyone, not only the companies that made it. They are the denominator to keep in mind when someone quotes Lovable at you.
| Source | Metric | Figure | Context |
|---|---|---|---|
| Stripe, 2025 | Median time to $1M annualized revenue, top 100 AI companies | 11.5 months | About four months faster than the fastest SaaS cohort of the subscription era. Still a top-performer cohort. |
| a16z, 2025 | Median ARR at month 12, enterprise companies in a16z’s sample | Just over $2M | Hundreds of companies a16z had spent significant time with; a venture-selected sample, not a population estimate. The pre-AI $1M comparison was a best-in-class rule of thumb, not a median. |
| ChartMogul, 2025 | Share of AI-native startups reaching $10M ARR within 12 months | Under 1% | Eight times likelier than non-AI peers, from a very small base. |
| Bessemer, 2025 | Average time to $100M ARR, AI-native companies in the Cloud 100 | 5.7 years | Versus 7.5 years for the Cloud 100 overall. Both describe a cohort already selected for success. |
Read together: the best companies reach large ARR numbers much faster than they used to, and selected venture cohorts are moving faster too. The broadest sample here still puts the chance of reaching $10M in the first year below 1%. A target set from the ceiling is a bet that your company is a category winner. A target grounded in the relevant cohort is a plan.
What “ARR” means
Not every headline number describes the same thing, and the differences matter more at speed.
- Subscription ARR
- The annualized value of active, recurring subscriptions. The traditional definition, and the one board models usually assume.
- Run-rate or annualized revenue
- The most recent month or week multiplied by 12 or 52. Legitimate as a current-pace indicator, but it captures spikes, and a single strong month can move it materially.
- Contracted revenue
- Signed value, which may include contracts still inside a trial period, a pilot, or a break clause. It can be reported before any of it is collectable.
- Consumption or credit revenue
- Charged against usage. It can grow without a conversation and shrink the same way. Durable when usage is embedded in a workflow, fragile when it is driven by experimentation.
- Gross marketplace revenue
- The full transaction value before payouts to the supply side. Third-party estimates put Mercor’s net revenue at roughly one-third of its reported $2B gross run-rate.
- Outcome-based annualized revenue
- A run-rate built from completed outcomes, such as resolved support cases. It may be durable, but it varies with the customer’s activity volume.
These can all describe real demand and real growth while overstating durability. When you evaluate a growth story, or a target you are being handed, ask for four numbers together: the basis of the ARR figure, gross margin, cash conversion, and cohort retention.
Method and limits
I found reports and raw data, organized my writing, and formatted the results into this report and HTML with the help of Anthropic's Opus 5, checked the content with OpenAI's GPT-6 Astra, and edited it the old fashioned way - too many hours staring at a screen.
- Selection on the outcome
- The companies here were chosen because their growth was already known. That makes the conditions descriptive of winners, not predictive of winning.
- Imperfect controls
- Most of the pre-AI comparisons are successful companies from an earlier era rather than matched companies that failed in the same conditions. A stronger design would follow matched companies chosen before outcomes were known, scored on the same definitions.
- Self-reported figures
- Many milestones come from company announcements and founder posts. Each row in the dataset carries a credibility rating and a basis note for this reason.
- Approximate dating
- Most sources give month and year. Intervals should be read as ranges, not precise measurements.
- Disputed starting points
- Some companies have conflicting public figures for the same moment. Replit is the clearest case: its CEO put end-of-2024 ARR at about $2.8M, Sacra estimated $16M, and Growth Unhinged reports $10M. The $10M-to-$100M interval used on the summary is the best-sourced of the three, and the dataset includes all of them.
The dataset, with every source link and credibility rating →