Author’s note: This is the latest in my “What’s the Bet” series. AI has the potential to change everything, and founders have to make company-defining choices in response. That is not easy. In this series, companies pay me to analyze their bet and explain it to you. I retain complete editorial independence. Today’s bet is brought to you by Lovable. If you have any feedback, just reply to this email.
Code is the raw stuff of creation. It takes problems, suffering, duress, bad vibes, and spooky juju, then applies the wizardry of pixels to make all that stuff go poof.
In my life, one of my kids is medically complex, requiring constant coordination between her medical providers, her daycare, her grandparents, and my wife and me. To track it all, we have hastily scribbled Apple Notes, scraps of paper, spreadsheets, Google Docs, and stray thoughts, all combining into a low-level buzzing sound, a deluge of data, a necessary evil that ends up taking us away from what our daughter needs the most (our love and attention).
A few weeks ago, I decided I couldn’t take it anymore. So I pulled open Claude and ranted about my problems for five minutes. I talked about the stakeholders, the stress it was causing my family, everything. From there I asked, “OK, how can we fix it?” The proposal was a simple web app to act as a central communications and data repository to make life easier. Then I just tagged Lovable in and said, “Go build this.” Claude drove Lovable through Lovable’s MCP server and away we went.
Forty-six minutes and about $10 of credits later, I had a beautiful, functional application that made my life better. That is remarkable. What made this experience magical beyond what a typical coding agent would give me is that it handled all the unfun stuff. The app had a quality design (after I gently nudged it to “not be so ass”), it set up a domain I purchased inside Lovable itself, and it even got email working. All creation, no suffering. If coding agents have solved code, the goal of Lovable appears to be to solve everything surrounding said code. I’m not going to list features and backend stuff forever because you get the idea, but having Claude sculpt the idea and Lovable handle the rest was a new way of creating software for me.
My app will never make an ounce of revenue. Neither will most of the 70 million projects built on Lovable since November 2024. That has often been held up as a critique of the company and its products. During my research for this piece, I asked myself the opposite question. What if it were its strength? What if Lovable became a company worth hundreds of billions because the software it helps create won’t ever make any money?
I think of this category of companies as competing to become the “Shopify of Software,” meaning they are trying to become the platform that non-technical people use to make software, akin to what Shopify did for ecommerce. However, Shopify became Shopify because it takes a cut of everything its merchants sell. Lovable cannot do that, because most of what gets built on Lovable will make a grand total of zilch.
So Lovable is forced to get its return from somewhere besides the revenue their apps generate. Thankfully, every app, regardless of its capitalist success, requires burning tokens to create and infrastructure to support. Lovable can build a great business if it derives revenue from everything but the software its users build. What matters is pumping up the volume of projects flowing through their platform as high as possible.
In a sign of divine favor from the business strategy gods, that playbook ends up neatly mapping onto a killer user experience. What I want to argue is that the more money Lovable makes, the happier its users will be.
Four sources of margin
If Lovable makes its money on everything but the software, the question is what “everything” is. I count four potential sources of margin.
Lovable’s competitor Wix was able to move its gross margin from “near-zero” to a target of “~60%” in the second half of 2026 simply by moving its inference to owned models. It feels fair to assume Lovable is going to run the same play. The most important source of margin for the company has to be token spread. Today, a model router guides every prompt, with Lovable’s owned models “handling a meaningful share of app-building work in production.” Allow me to translate that for you, “Open-source is good enough that the business works now! All of our founders are going to get yachts!” This is because Lovable charges on the basis of “credits” not tokens. And a credit costs the same regardless of the model used. The better Lovable gets at model routing and model training, the more money it makes.
The second is payments. Lovable shipped built-in payments in April 2026 and, so far, charges nothing on them. Users pay the same fees they would pay Stripe directly, while physical goods go to a Shopify integration that puts checkout on Shopify. I imagine that with sufficient volume, Lovable may keep users at entry-level payments pricing while pocketing the spread between what it charges and what it can bleed out of its payments partners. Still, this will never be as large a part of its business as it is for Shopify. It simply monetizes too few of its users: over a third of builders on Lovable earn revenue, which means nearly two-thirds do not. (I was surprised that the revenue generated was that high at all honestly.)
The third is the runtime. Hosting, database, auth, storage, scheduled jobs, that sorta thing. Apps built on Lovable are having 900 million visits in a month. Add to that the services it routes you to without leaving the editor, like the domain I bought. Similar to payments, this is likely small to non-existent today, but as software becomes as common as sliced bread, this will grow.
The fourth is the enterprise. Every CTO on the planet gets to justify their existence by insisting on acronym compliance. Gotta have SSO, SCIM, audit logs, publishing controls, and training-data exclusion, and insurance certification. By connecting securely to an enterprise’s databases, Lovable can be the layer that lets employees populate their apps with the data they need while staying compliant.
Combined, these four levers contributed to revenue going from $100 million in annualized revenue in July 2025 to “$500 million in annualized run rate revenue” in June 2026, and Dealroom has it near $600 million by August. This is despite the cost per application almost certainly being higher than using a Claude Code or Codex subscription to make it. However, for most people, token cost is the smaller concern. It is the cost of “somehow my sales staff broke prod” that Lovable prevents. People will pay a lot to automate away complexity and security risk.
Why users like me are happy to pay the Swedes
That last sentence is the kind of brain-dead insight accessible to anyone with a pulse. Of course people pay more to do less work. Accordingly, lots of startups are attempting this strategy. You can visualize it as three tiers of tooling.
The first is the spreadsheet. When I started as a knowledge worker, the gods among us mere mortals were the folks who used VBA to control Excel. By building a macro layer, they could manipulate spreadsheets—essentially turning them into apps. Does anyone still write VBA in Excel? I’d bet its popularity has collapsed. It’s easier just to have an AI agent do it. Still, many Lovable apps probably start as a poorly managed spreadsheet, just like my care-team app did.
The second is the coding agents and harnesses: Claude Code, Devin, and the rest. You don’t give a sixteen-year-old a Ferrari, and you don’t give an unlimited token budget to a product manager fresh out of Harvard Business School. That is too much power for most people.
The third is the digital-abundance platforms: Lovable and Replit. Today, they win on go-to-market efficacy at the prosumer level. Through the sponsorship of newsletter writers and other smart GTM things, the company tries to convince individuals to start building apps. From there, it likely runs an enterprise sales motion akin to Slack’s: bottom-up adoption, where people secretly build apps for their teams and the vendor upsells once IT notices. Today, nearly two-thirds of the Fortune 500 have employees on Lovable, and CEO Anton Osika says about “half” of accounts come from enterprises.
What power does Lovable have?
The Leverage is a publication about power. I think there are two primary sources of it for Lovable.
The first is scale economies. Because Lovable’s central source of gross margin expansion is model routing, the more data it can feed into that router, the better it gets at selecting the cheapest way to build the code. Over time, I assume the majority of its token burn will shift from frontier models to internal ones, essentially turning Lovable into a leading AI lab overnight. The margin keeps improving as long as the volume does.
The company is also slowly building up its switching costs. Frankly, they are relatively weak for individuals, because every time you think “I should go build an app,” there will always be a new tool competing for your attention. This is why the enterprise sales motion is so important for Lovable. A company does not migrate a few hundred internal apps to save on credits.
The major threat is the same threat that all applications face today: vertical integration from the labs. Coding and computer use are the premier use cases for OpenAI and Anthropic, which means Lovable is squarely in their sights. The spread Lovable earns on tokens is a spread the labs never have to pay, and they have shipped hosted app builders of their own. A vertically integrated model provider, one that has the best models and owns its own data centers, is very, very hard to compete against. To escape this trap, Lovable needs to specialize in the truly non-technical, the buyer the labs’ products are least designed for today, and to keep its routed, post-trained stack getting cheaper faster than the labs’ products get simpler.
The purest articulation of Lovable’s bet is that the world builds vastly more software once software is disposable. This is a thesis I also strongly believe. Once software becomes as cheap as water, everyone will build it. And users benefit from this thesis too! Anytime we get annoyed at work, the solution could be to just tag Lovable in and say, “Plz build an app and fix.”
I don't understand why you need to use Lovable here, if you already outlined what you want to Claude Code?