How Bolt went from zero to $40M ARR in five months

Executive overview

StackBlitz spent seven years building browser-based dev infrastructure without finding a monetisation path. A last-chance experiment — an AI app-builder called Bolt — went viral from a single tweet and reached $20M ARR in two months.

The key was timing: they had tried Bolt a year earlier, shelved it when AI models weren't good enough, then relaunched within 90 days of getting early access to Claude Sonnet 3.5.

Waiting for the right model, then moving fast with existing technology, beat every competitor who bolted on AI prematurely.

From nearly shutting down to product-market fit

• Seven years building a browser-based dev environment; strong technology, weak monetisation
• Tried multiple experiments over two years — technically impressive but none moved revenue
• Board conversation about winding down; decided to run every remaining idea before closing
• Bolt launched with a single tweet — no press release, no blog post
• $60K ARR added on day one; $20M ARR within two months; $40M ARR by month five
• Nearly four million registered users and still growing

The AI timing decision

• First Bolt prototype built in early 2024; shelved because frontier model output quality was too low
• Got early access to Claude Sonnet 3.5 in May 2024 — recognised it as a step-change for coding
• Relaunched Bolt within 90 days using Anthropic APIs
• Competitors added AI reactively after ChatGPT; many have since shut down or stalled
• Holding off until there was genuine conviction meant they moved once, and moved all-in

How Bolt works

• Users type a natural-language description of an app or website
• Bolt generates a working web application in ~30 seconds
• Built on StackBlitz's core technology: a full dev environment running inside the browser, no server, no local install

Scaling user feedback

• Hired team members whose sole job is talking to users and unblocking them
• Invested heavily in community (Discord, Reddit, Twitter) — community leaders surface recurring issues
• Prioritise only what users have explicitly asked for or what has been directly validated
• Signal-to-noise is a real challenge at thousands of tickets per day; community structure is the solution

Open source as a growth lever

• Open-sourced the core Bolt codebase at launch — still the most popular AI product by users and revenue with an open-source version
• Users fork, improve, and experiment; many convert to paid
• StackBlitz is a major backer of Vite and employs full-time open-source contributors
• Open-source relationships ensure the broader web dev toolchain works well inside their environment

Knowing when to pivot vs. iterate

• No formula exists; gut instinct develops only through running many ideas over years
• Distinguish between "not quitting because people say you shouldn't" and "not quitting because you have real conviction"
• When you do commit to going deep, exhaust every option — don't look back wondering if you tried everything
• When driving through hard times, don't stop halfway

AI's impact on software development

• AI will be the highest-leverage application of AI overall, measured by revenue and impact
• Most of the world's software will be written primarily by AI within single-digit years
• Developers move to higher-value, intellectually challenging work; the monotonous parts get automated
• Bolt's five-year vision: Fortune 100 companies and ten-year-olds building production software with the same tool