Growth Is Easy. Here’s Who’s Actually Making a Profit.

New York’s hottest AI companies are adding revenue faster than almost any businesses in history. Not one of them has disclosed a profit. Here’s how to tell the difference between a money machine and a very expensive science project.

The number everyone quotes, and the one nobody does

Every AI founder in New York can tell you their recurring revenue. Ask for their profit and the room gets quiet. That isn’t a scandal. It is how high-growth companies have always worked: spend everything on growth now, collect later. But it means that when you read “$200 million in recurring revenue” in a headline, you are reading a measure of demand, not a measure of how much money anyone takes home.

In New York’s AI Gold Rush, we profiled five New York companies with real revenue: Clay, EliseAI, Rogo, Runway and Bluefish. This article asks the harder question. Are they making money? And how would an investor, a business owner or a skeptical dinner guest tell?

What does annual recurring revenue actually mean?

Annual recurring revenue, or ARR, is what a company would collect over the next twelve months if every current customer kept paying at today’s rate. It is a snapshot multiplied by twelve. A company that signs a big customer in December can report a much higher ARR in January than the cash it actually collected that year.

ARR is useful because it shows momentum. It is limited because it says nothing about costs. A company can have $200 million in ARR and spend $250 million a year getting it. That is why the serious question for every AI company is not “how big is your ARR?” but “what does it cost you to deliver?”

Why are AI company margins lower than software margins?

Traditional software is famous for one thing: once it’s built, each new customer costs almost nothing to serve. That is why software companies typically keep 70 to 90 cents of every dollar after the direct cost of delivering the product, a figure called gross margin.

AI is different. Every time a customer asks an AI tool to draft a memo, generate a video or read a contract, the company pays for computing power to run the model. Industry data compiled by the pricing firm Solvimon puts the gross margin of AI-native software at about 52% in 2026, up from about 41% in 2024, with model hosting and running costs eating 40% to 50% of revenue. Treat those figures as directional, not exact, but the gap is real. An AI company keeps less of each dollar than an old-school software company does.

The encouraging part is the trend. Margins rose about 11 points in two years as computing got cheaper and companies got smarter about which tasks need the most expensive models. If that continues, today’s growth becomes tomorrow’s profit.

How the five New York companies stack up

Here is what each company has disclosed, and what it hasn’t. “Company” means the number came from the company itself; “estimate” means a research firm calculated it.

EliseAI has the strongest record of the group. Five straight years of doubling is not a lucky quarter; it is a business that keeps finding customers who stay. Runway has the most dramatic momentum. Clay has the most unusual signal: it has repeatedly let employees sell shares through tender offers rather than rushing to an IPO, which suggests investors are eager to buy in.

Five signs an AI company is a real business

You don’t need a finance degree to judge an AI company. These five questions will get you most of the way there.

• Is it replacing a budget that already exists? EliseAI replaces work landlords already pay leasing staff to do. Rogo replaces analyst hours banks already pay for. When the money is already in someone’s budget, the sale is easier and the customer is less likely to cancel.
• Do customers stay and spend more? Doubling for five years, as EliseAI has, is only possible if customers keep renewing.
• Is the revenue number from the company or an estimate? Company-disclosed numbers carry more weight, especially when they are tied to a funding round with outside investors checking the books.
• Who are the customers? Named enterprise customers, like Runway’s Amazon and Robinhood or Bluefish’s Adidas and American Express, are harder to fake than vague claims of “thousands of users.”
• Are margins improving? Few private companies disclose this, but it is the right question to ask any founder pitching you.

What about the big companies?

The giants have the opposite problem. They have profits but are struggling to show that AI adds to them. JPMorgan Chase CEO Jamie Dimon has said the bank’s roughly $2 billion a year in AI spending is matched by about the same in savings. He has also warned that when every bank uses the same tools, the savings tend to flow to customers rather than shareholders.

That is the paradox at the heart of this series. The startups have growth without disclosed profit. Meanwhile, the giants have profit without much AI growth. The sweet spot may belong to a third group: small, established businesses that already make money and use AI to make more of it. We look at them in Five People and a Laptop Just Beat Your 50-Person Firm.

What to watch next

Three numbers will tell us a lot over the next year. First, Runway expects to reach $350 million in recurring revenue by the end of 2026, up from $200 million in September, according to sources cited by Bloomberg. Second, Sacra has projected that Clay could reach about $238 million by year-end if its pace holds, though that is an estimate built on assumptions the firm doesn’t fully spell out. Third, margins: the same industry data that puts AI-native software at about 52% suggests a likely ceiling of 60% to 65%. Any New York AI company that announces margins near that range, or a first profitable quarter, will have answered the question this article asks.

The optimistic case

Every great software company of the last thirty years spent years losing money while it grew. The question was never whether early growth was profitable; it was whether the customers were real and the costs would fall. For New York’s AI companies, the customers are about as real as they get: banks, landlords, studios and Fortune 500 brands. The costs are falling every year.

The founders building these companies, several of whom we profile in The Under-40s Already Running New York’s AI Economy, have already done the hardest part of business. They found people willing to pay. If you know a New Yorker under 40 who has gone one step further and turned that into profit, nominate them for the Hamptons 40 Under 40.

Frequently asked questions

Sources

• Enterprise DNA: AI-native SaaS gross margins (Solvimon data)
• Digital Health News: EliseAI raises $350M at $4B
• Crypto Briefing: Runway hits $200M ARR
• Sacra: Clay crosses $150M a year
• Axios: Rogo’s growth
• Revenue Memo: Rogo funding and revenue
• Adweek: Bluefish customers
• HyperAI: Dimon on $2B AI savings (Bloomberg TV)
• Webull: Dimon on AI savings and margins
• AIbase: Runway enterprise customers
• AI Weekly: Runway targets $350M by year-end