Nobody hits $100M ARR in eight months by being 'AI-native'

The eight-month ramp is a distribution trick, not a model trick

Lovable went from launch to $100M in annual recurring revenue in eight months — a milestone that took last-gen SaaS roughly five years — and to a $500M run rate about ten months after that. Cursor blew past a $4B run rate and, in June 2026, signed a $60B acquisition agreement with SpaceX. These are the fastest revenue ramps in the history of commercial software, and the industry has one word for all of them: AI-native.

That word explains nothing, because it is true of everyone. There are ten thousand AI-native startups and you have heard of eleven. "AI-native" is the price of a ticket; it is not why five companies lapped the field. So here is the thesis: the ramp is a distribution story, not a model story. Almost none of the demand these companies captured was new — lawyers already paid to read contracts, developers already paid to write code faster, enterprises already paid call centers to answer tickets. A foundation model made that work suddenly cheap to serve, and the winners were the ones positioned to route the cheap capability straight into a budget someone had already approved.

Below are the five distribution motions that did it, each explained with the one company that ran it best. The lesson to leave with: when you see one of these ramps, don't look at the model — find the borrowed channel.

Motion one: bottom-up self-serve, land before anyone signs

The slowest thing in enterprise software is getting into the room — years of field sales and procurement. The self-serve motion skips the room entirely: a single user pays with a credit card, no sales call, no pilot, and the company is inside the account before anyone in procurement has heard the name.

Cursor is the worked example. It sat one keystroke away from a developer's existing habit and let a $20 charge do the selling; there was never a demo to book. By the time an enterprise noticed, hundreds of its engineers were already paying out of pocket, and the company converted that ground-up footprint into contracts — today roughly 60% of Cursor's revenue comes from large corporate buyers who followed the bottom-up wedge upward.

The motion isn't a growth hack you execute; it's a posture you hold — be the thinnest thing between an individual and a capability, and let adoption precede the sale.

Motion two: product-led spread, where the product is the sales team

The second motion is a cousin of the first, but the engine is different: the product itself does the distribution, because using it produces something the user shares. Growth comes from output, not outreach.

Lovable is the worked example. People built apps with it and shipped them, and every shipped app was an advertisement that pulled in the next builder — a loop that needed no sales org at all. The tell is the headcount: Lovable reached $100M ARR with roughly 45 people and served millions of users on about 146. A company does not hit nine figures that lean by out-selling anyone; it does it when the product is the marketing channel.

That headcount is what product-led distribution looks like when the underlying capability just got cheap enough to make the output good enough to share.

Motion three: ride an incumbent's channel into a pre-approved budget

You don't have to build a channel if you can borrow one. The third motion targets a market where the budget, the buyers, and the trust relationships already exist, and slots in as a line item rather than a transformation.

Harvey is the worked example. A tiny startup does not cold-email its way into half of the largest law firms in America — that's a decade of relationships it didn't have. What it had was a product specific enough that a managing partner could green-light it against the budget already spent on billable document review, inside institutions that already buy legal software through known channels. That borrowed channel is how Harvey reached roughly 50% of the Am Law 100 without a decade of enterprise sales. The model made the product possible; the incumbent channel made it fast.

Motion four: land-and-expand on a meter that climbs by itself

The fourth motion is what happens after the land: instead of hunting new logos, revenue expands inside the accounts you already have, because usage grows on its own. The first seat is the wedge; the meter is the business.

Harvey shows this one too, in its expansion data: median seat count doubles within twelve months of a firm signing.

Land a few hundred licenses for one workflow, and the number climbs as more lawyers pull the tool into more matters — no new logo, no new sales rep. This is the motion classic SaaS always wanted and rarely got at this slope, because the AI product finds new uses inside an account faster than a human champion could roll it out. Expansion, not acquisition, is doing the work.

Motion five: outcome pricing, sold against a labor budget

The last motion changes the number you're compared to. Sell software by the seat and you're measured against a software budget. Sell an outcome — a resolved ticket, a completed task — and you're measured against a labor budget, which is far larger and already being spent on people.

Sierra is the worked example. It prices customer-service agents against resolved conversations, not seats, so the buyer compares Sierra to the cost of the call-center headcount it offsets rather than to another SaaS subscription.

That reframing opens a bigger budget and a faster yes, because the buyer isn't adding a tool — they're converting a labor line into a software line. Outcome pricing is a distribution motion disguised as a pricing choice: it routes you to the biggest pre-existing budget in the building.

"AI-native" tells you why the product could exist. The distribution motion tells you why it sold — and the ramp lives entirely in the second story.

What to actually do with this

The five motions aren't a menu of tricks; they're five ways to answer one question — whose already-approved budget am I routing a suddenly-cheap capability into? I run every ramp through the same test before I credit anything to the model, and if it doesn't resolve to one of the five, I don't understand it yet. Concretely, that means:

1. Name the motion before you credit the model. If you can't identify which of the five explains the growth, you don't understand the growth — you're just admiring a chart.
2. Find the pre-existing demand. The record ramps captured spend that already existed. If your motion requires creating demand, expect years, not months.
3. Match the budget to the pricing. Seat pricing competes for software budgets; outcome pricing competes for labor budgets. Point yourself at the bigger number on purpose.
4. Let adoption precede the sale where you can. Bottom-up and product-led motions put you inside the account before procurement wakes up — a structural head start no sales org can buy.
5. Engineer expansion, not just landing. The durable ramps grow inside accounts on a meter. Design the product so its second use sells the third, and the ARR climbs without a new logo.

The eight-month ramp is real, and the most impressive thing I've watched in twenty years of seeing software get sold. But it wasn't built by being AI-native. It was built by pointing a suddenly-cheap capability at a budget that already existed, through a channel someone else already owned. Name that channel, and the magic turns into a method you can run.