Builder.ai's $1.5B Collapse Is the AI-Washing Reckoning No-Code Needed

Builder.ai was a $1.5 billion unicorn backed by Microsoft. Its AI, "Natasha," supposedly made building an app "as easy as ordering a pizza." Last week, it collapsed into insolvency. Behind the curtain: roughly 700 human engineers in India writing code by hand, revenues inflated by 300%, and a federal criminal investigation.

The no-code industry should be studying this corpse closely. Not because Builder.ai was one of us, but because it pretended to be, and the distinction between real platforms and AI vapour has never mattered more.

TL;DR: Builder.ai, once valued at $1.5B, collapsed after revelations that its "AI" was actually hundreds of human developers in India, and its revenues were inflated via circular invoicing with an Indian partner. It owes $115M to Amazon and Microsoft. The lesson for no-code: if a platform can't show you the model, it's a services business wearing an AI costume. We've built a framework for telling the difference.

What actually happened at Builder.ai?

Let's get the facts straight before we talk about what they mean.

Builder.ai, founded in 2016 by Sachin Dev Duggal (who styled himself "Chief Wizard"), raised over $500 million. Its pitch was irresistible: an AI called Natasha that could autonomously build custom applications. Microsoft bought in. The Qatar Investment Authority led a $250 million Series D at a $1.5 billion valuation in May 2023.

None of it was real.

When the new CEO, Manpreet Ratia, commissioned an internal audit in early 2025, the numbers unravelled fast. Revenue for 2023 was actually $45 million, not the $180 million claimed. The 2024 projection of $220 million? Actual: roughly $55 million. That's a 300% overstatement.

The mechanism was classic "round-tripping." Builder.ai and VerSe Innovation, an Indian social media firm, allegedly exchanged invoices for similar amounts, nearly $60 million total, for services that may never have been delivered. Both companies shared major investors and both looked great on paper.

Then there's the technology lie. The Wall Street Journal first reported in 2019 that Builder.ai's AI was largely human-powered. By 2025, the scale became clear: an estimated 700 engineers in India were doing the actual coding. They were reportedly instructed to mimic UK working hours and avoid Indian-English phrasing in client communications. Natasha was a chat interface that passed requirements to humans.

Creditors seized $37-40 million from the company's accounts in May 2025. The U.S. Attorney's Office for the Southern District of New York issued a subpoena. Chapter 7 bankruptcy followed, with debts of $85 million to Amazon and $30 million to Microsoft.

This isn't a startup that "didn't make it." This is Theranos with a SaaS login page.

Why should the no-code industry care?

Builder.ai marketed itself squarely in the no-code and low-code space. "Build apps without writing code." "AI-powered development." "As easy as ordering a pizza." These are the same promises our industry makes.

The difference, and it's a chasm, is that real no-code platforms don't need to fake it.

Bubble gives you a visual programming environment. You build logic, design interfaces, and deploy. The platform executes your work. There is no hidden call centre of developers. Webflow generates actual HTML, CSS, and JS from a visual canvas. Softr turns Airtable bases into functional apps through configuration, not deception. Stacker does the same for enterprise data.

Builder.ai had none of this. It had a marketing website and an outsourced dev shop in Gurgaon. The "AI" was window dressing on a labour arbitrage business. And labour arbitrage doesn't scale like software, which is precisely why the company had to invent revenues to keep the valuation narrative alive.

This matters because customers are going to get spooked. When a company that called itself an "AI-powered app builder" turns out to be a fraud, the stain spreads. Every no-code platform now faces a harder sell: "How do I know you're not another Builder.ai?"

The answer is simple. And it leads directly to the framework we need.

So how do you tell a real platform from vapour?

After watching this collapse unfold, I've put together four questions every buyer, investor, and journalist should ask when evaluating any no-code or AI development tool. If a vendor can't answer these cleanly, walk away.

1. Can you see the output?

Real platforms produce artefacts you can inspect. Bubble exports code. Webflow generates sites you can host anywhere. Softr and Stacker produce working applications with visible, queryable data layers.

Builder.ai's clients reportedly received code that was "unreadable" and "glitchy." One executive described a six-month project that became a 28-month ordeal with no working MVP. If the platform can't show you clean, reviewable output that you own, it's not a platform. It's a consultancy.

2. Where is the model?

This is the simplest test and the one almost nobody asks. If a company claims AI builds your app, ask: what model? Fine-tuned on what? Running on whose infrastructure?

Builder.ai never answered this because there was no model. There were humans. Any real AI tool, whether it's GitHub Copilot, Cursor, Bolt.new, or Lovable, has a specific model architecture, training methodology, and inference pipeline. They might not share proprietary details, but the existence of the technology is verifiable.

If you get a hand-wavy answer about "proprietary AI" with no technical specificity, you're looking at a services company with a domain name.

3. Does pricing scale with usage or with headcount?

This is the tell that would have caught Builder.ai years ago. Software platforms charge per seat, per compute, or per usage metric. Services businesses charge per project and scale by adding people.

Builder.ai's pricing was per-project, and the projects got more expensive and took longer than promised. The model was human-dependent by design. More customers meant hiring more engineers, which is the opposite of software economics.

No-code platforms that charge per editor seat, per app, or per API call have the cost structure of a technology company. Those that charge per "project" with long timelines and opaque delivery have the cost structure of an agency.

4. Can you build something yourself right now?

Go to bubble.io. Sign up. Build a button that does something. It takes five minutes.

Go to webflow.com. Drag a div. Style it. You're building.

Go to softr.io. Connect Airtable. You have a working app in ten minutes.

Builder.ai required a sales call. Then a scoping call. Then a proposal. Then a contract. Then months of waiting. If you can't build something functional in the first session without talking to a human, the "platform" is a lead-gen frontend for a services pipeline.

The takeaway

Builder.ai wasn't a no-code company that failed. It was a fraud that borrowed no-code's language because that language was credible and well-funded. The real damage isn't to Builder.ai's investors, who should have asked harder questions. It's to every legitimate platform that now has to prove it isn't a scam.

But here's the thing: proving it is easy. Real platforms have visible outputs, specifiable technology, scalable pricing, and instant self-serve. If you're building on Bubble, Webflow, Stacker, or Softr, you already know the difference because you've felt it. You built something.

The Builder.ai collapse isn't a crisis for no-code. It's a purification. The fakes can't hide behind a slick pitch deck and a .ai domain anymore. And that's good for everyone who builds things that actually work.