Rocket.new, a startup only 16 weeks old, has just completed a $15 million seed round led by Salesforce Ventures, directly targeting rising competitors such as Lovable, Cursor, and Bolt.

The company's ambitions go far beyond that. Unlike tools focused on rapid prototyping, Rocket.new is dedicated to generating complete production-grade applications through natural language prompts, making it stand out in the competitive AI programming landscape with this differentiated approach.

Accel and Together Fund also participated in this equity round, which took place only three months after Rocket.new launched its beta platform in June. In such a short time frame, the company has achieved impressive growth metrics: the platform has surpassed 400,000 users, including over 10,000 paid subscribers, covering 180 countries globally, and achieving an annual recurring revenue of $4.5 million.

In an exclusive interview with TechCrunch, co-founder and CEO Vishal Virani revealed more ambitious growth plans: the company aims to increase its annual recurring revenue to $20-25 million by the end of the year, and to a staggering $60-70 million by June next year.

Rocket.new was born in Surat, known for its diamond and textile industries, far from India's traditional tech hubs. This geographical context itself carries symbolic meaning. Virani, along with Rahul Shingala and Deepak Dhanak, founded the company, marking their successful transition from their previous project, DhiWise, which focused on developer workflows.

Virani explained the company's core philosophy to TechCrunch: We are building the first ambient solution platform, focusing not on solving day one problems but on addressing day two challenges. This forward-thinking approach is what sets Rocket.new apart from its competitors.

This 16-week-old startup is boldly aiming to become a comprehensive agent system. This system can not only use AI to build applications and websites, but also conduct competitive research and product development, and even has the potential to completely replace the traditional product manager role.

Virani confidently stated: Our entire agent system will help organizations build various features around products, not only generating source code, but also providing facilities to expand products, all achievable with just natural language prompts.

The current 0.3 version of the model has successfully built 500,000 applications, attracting a diverse user base, including product managers, individual entrepreneurs, and front-end developers. More impressively, employees from well-known companies such as Meta, PayPal, KPMG, PwC, and Times Internet are also using the platform for personal project development.

Data shows that approximately 80% of Rocket.new users build what Virani calls serious applications, rather than simple landing pages or brand websites. About 12% of users have created e-commerce platforms in niche areas such as groceries and clothing, 10% have developed financial technology applications, 5-6% have created B2B tools, and 4-5% have launched mental health applications.

In terms of application types, about 45% of users choose to build mobile applications, while 55% focus on website development. Virani revealed that many users first build websites on Lovable or Replit, then switch to Rocket.new to generate native mobile applications by integrating existing Supabase backends.

Rocket.new's technical architecture is unique, combining large language models from Anthropic, OpenAI, and Google Gemini with its own deep learning systems. These proprietary systems are trained on DhiWise's unique dataset.

Virani emphasized: Our underlying architecture is completely different from Lovable, Bolt, and all other competitors. This differentiated technology approach is the confidence that allows Rocket.new to challenge industry giants.

In terms of performance, Rocket.new takes about 25 minutes to generate the first application, which is much slower than most ambient coding tools' 3 minutes. However, according to early test results, Rocket.new offers a more comprehensive user experience, including all necessary module functions.

In terms of business model design, Rocket.new provides a free trial limited to 1 million tokens. After exceeding this limit, users need to pay a monthly subscription fee, starting with a $25 package for 5 million tokens.

This pricing strategy effectively filters out casual hobbyists, while bringing Rocket.new a healthy gross margin of 50-55%, with the company planning to increase this figure to 60-70% in the coming months.

In terms of market distribution, the United States is Rocket.new's largest revenue source, contributing 26% of revenue, followed closely by Europe at 15-20%, and the Indian domestic market at 10%. To better serve American users, the startup is establishing its U.S. headquarters in Palo Alto.

Virani pointed out that Rocket.new's early growth was entirely driven by word-of-mouth and viral social media. With new seed funding support, the company will focus on refining its market entry strategy, deepening its influence in key markets, and accelerating the development of its proprietary models and research work.

Kartik Gupta, an investor from Salesforce Ventures, said via email to TechCrunch: We saw a clear gap between the magic of AI code generation and actually putting code into production. Rocket.new was specifically designed to solve enterprise-level iteration, maintenance, and deployment issues.

Rocket.new currently has 58 employees, mainly based in Surat. The company plans to double the size of its engineering and product teams in India within the next 12 months to support the rapid growth of its business.