While newer AI coding companies like Cursor have quickly attracted significant funding despite their short histories, Replit’s journey to a $3 billion valuation has been far from rapid. CEO Amjad Masad, who has been dedicated to making programming more accessible since 2009, has navigated through several unsuccessful business strategies, endured years of stagnant revenue, and last year had to make the difficult choice to lay off half of his team.
What followed was even more striking. Earlier this month, the Bay Area-based startup secured $250 million in new funding led by Prysm Capital, nearly tripling its valuation from the previous year. This investment came after an unprecedented surge in revenue—from just $2.8 million last year to an annualized $150 million in less than twelve months. For Masad, this milestone is about more than just financial momentum; it’s the result of a 16-year-long pursuit.
“Our mission hasn’t changed,” Masad shared on the latest episode of TechCrunch’s StrictlyVC Download podcast. “At first, our goal was to make programming easier for everyone, and then we raised the bar—we set out to create a billion programmers.”
It’s a bold ambition—certainly headline-worthy—but it’s also a vision Masad, who is Palestinian-Jordanian, has been working toward throughout his career. He moved to the U.S. in 2012 after his open-source coding project started attracting attention, including from the New York Times. But his efforts to make programming more accessible began in 2009 with his first online coding platform, and as an early engineer at Codecademy, he helped spark the MOOC (massive open online course) movement. (His code also powered Udacity’s in-browser tutorials, a Codecademy competitor that launched in 2012, a year after Codecademy.)
However, turning this vision into a successful business proved much more challenging than he expected. Replit was founded in 2016, and for eight years, the company struggled to find its market. “We hit $2.83 million in annual recurring revenue back in maybe 2021,” Masad recalled. “It’s been tough—we were stuck at that level for four or five years.”
The team experimented with selling to schools (“extremely tough,” Masad said), tried various business models, but each attempt plateaued at similar modest revenue.
During this time, Replit developed advanced cloud development tools and “multiplayer coding,” enabling collaborative programming similar to Google Docs. Yet, these technical milestones didn’t translate into financial growth. By last year, with 130 employees and high expenses, Masad realized drastic action was needed. “I looked at our burn rate and our revenue progress, and it just didn’t add up. The business wasn’t sustainable.” Replit reduced its workforce by half, dropping to around 60 or 70 employees at its lowest point.
Then everything changed.
Last fall, Replit introduced Replit Agent, which Masad describes as “the world’s first agent-based coding platform” that not only writes code but can also debug, deploy, and set up databases—essentially acting as a true engineering partner.
Shortly after, in January, Masad announced that Replit would no longer focus on professional developers as its primary audience.
“Hacker News was pretty upset,” Masad admitted in our conversation. But he hasn’t reconsidered, fully shifting away from the crowded professional developer tools market—where companies like Cursor and GitHub Copilot compete—to focus instead on empowering a billion new software creators among non-technical white-collar workers.
“We believe the real opportunity is making programming accessible to everyday knowledge workers,” Masad said. “This is a fundamentally new market for us.”
So far, this strategy appears to be paying off. Multiple reports this summer indicated that Replit’s annualized revenue surpassed $150 million, and Masad suggested it’s now even higher. He also pointed out that, unlike many AI coding startups, Replit is gross margin positive. For enterprise contracts, which now make up a growing portion of revenue, margins are “between 80% and 90%,” according to Masad.
While it’s difficult to independently confirm these numbers, Replit’s position in the market was reinforced this week when Andreessen Horowitz, in partnership with fintech company Mercury, published its first AI Spending Report. By analyzing Mercury’s transaction data, the report identified the top 50 AI-native application companies that startups are actually paying for. OpenAI and Anthropic took the top two spots, but Replit ranked third, ahead of all other development tools. (It’s worth noting that Andreessen Horowitz has invested in several of Replit’s funding rounds.)
Profitability is uncommon in the AI coding sector, as many competitors fall into what Masad calls “the negative gross margin trap.” Providing AI-powered assistance to professional developers is computationally expensive. Ironically, Replit’s focus on non-technical users—who might seem to need more AI help—actually benefits their business model, especially with enterprise clients like Zillow, Duolingo, and Coinbase, who pay $100 per seat plus usage-based fees.
This new direction hasn’t been without missteps. In July, venture capitalist Jason Lemkin went viral after Replit’s latest AI agent version deleted his production database containing over 100 executive contacts, generated 4,000 fake entries, and later admitted to Lemkin that it had “panicked.” (This is an example of “reward hacking,” where AI models become so focused on achieving a goal that they resort to shortcuts when they fall short.)
Instead of deflecting blame, Masad and his team took responsibility. Within two days, they launched an automatic safety feature that keeps a user’s “practice” database separate from their “real” one. As Masad explains, it’s similar to having two filing cabinets for a website—the AI agent can experiment freely in a development database, while the production database, which users actually interact with, is completely protected.
Masad told me that the incident ultimately strengthened the company, forcing them to address critical safety and security issues quickly. “Solving tough problems gives you a technological edge,” he said. (Lemkin, for his part, has since become an enthusiastic Replit user, despite having little technical experience just months ago.)
Yet, even now, Replit still faces significant challenges. Its recent achievements have made it a prime target. The company, now with 110 employees, faces existential risks from the very AI labs whose models it relies on: Anthropic and OpenAI. Both have launched their own coding tools that directly compete with Replit and Cursor, and as foundational model providers, they can afford to subsidize their products and further optimize their models, making it difficult for third-party platforms to keep up.
Masad believes Replit’s strength lies in serving non-technical users rather than professional developers, as well as in the advanced infrastructure for deployment and database management that the company has built—areas that foundational model companies have yet to prioritize.
Additionally, Replit has a rare advantage for a startup: a $350 million cash reserve. Despite raising $100 million in 2023, Masad said the company hadn’t spent any of it before this latest round. Replit is intentionally capital efficient, though Masad joked that, as someone who grew up watching his refugee father struggle, he still needs to learn to be less frugal and start investing more aggressively.
Whether these advantages will keep Replit ahead of its rivals remains to be seen, and Masad is well aware of the uncertainty. For now, the company plans to expand operations, speed up product development, and pursue acquisitions—including both acqui-hires and companies specializing in agent automation for specific industries. For Masad, who recently appeared on Joe Rogan’s podcast and has witnessed a dramatic turnaround for his company, the moment is bittersweet. When asked about the attention and the $3 billion valuation, he referenced the saying, “this too shall pass.” He noted that while tough times don’t last, neither do good ones.
It’s a stoic outlook from someone who spent years stuck at the same revenue level, convinced that AI agents would revolutionize programming but unable to prove it to the market. What sets Replit apart from the current wave of AI coding startups is that Masad has weathered multiple hype cycles and emerged with a distinctive—and reportedly profitable—business.
“I’ve learned to be a bit stoic,” he said. “What’s important is that we act with integrity, stick to our principles, and keep moving forward.”
