Fireside Chat, Interview
The Untold Story of Higgsfield | Burning $4M a Month on AI Models | CEO, Alex Mashrabov
- Hicksfield has achieved $1 billion in annualized revenue within 18 months of reaching $1 million, surpassing the 24-month timeline of Coursera to reach this milestone.
- The company operates with a team of approximately 400 people, with the majority based in Kazakhstan and Central Asia, while roughly 50 are located in California and 50 are remote.
- Internal monthly model usage costs exceed $4 million, averaging over $10,000 per team member, with top "10X" creatives occasionally spending over $30,000 in a single week on tools like Astra.
- Revenue composition is split with business revenue slightly over 50%, while consumer revenue accounts for the remainder, of which only about 10% comes from pure mobile consumer use cases.
- Net Revenue Retention (NRR) at the 12-month mark exceeds 300%, a figure the founder describes as unprecedented in the B2B SaaS sector despite high initial churn.
- The company generates over 80% margins on open-source or post-trained open weights models, compared to roughly 20-30% margins on closed-source proprietary models.
- Hicksfield pivoted from an initial focus on hype-driven narratives to a product-led growth strategy after burning over $10 million of seed capital in a failed year-long search for product-market fit.
- The company abandoned the development of proprietary foundational video models after realizing that customer needs for specific workflows were better served by model routing and leveraging open weights models.
- Current model strategy involves "tokenomics" and model routing, allowing the company to select the most efficient model for specific tasks, with the company itself choosing the model in over 40% of cases.
- Internal AI usage is expanding rapidly, with the legal and customer support teams heavily integrated with AI agents to manage high product velocity, though the company has not eliminated these headcounts yet.
- Founder Alex's philosophy rejects traditional management theory, advocating instead for hiring the best talent and empowering them to do high-impact work, mirroring the "no bullshit" culture of leaders like Frank Slootman and Chad Peetz.
- The company has moved its entire engineering and creative team to CodeC (previously Cursor) to support "vibe coding," enabling non-engineers to build production-ready assets, a shift that increased internal costs significantly.
- Higgsfield's revenue growth is largely driven by Asian trends, specifically the direct-to-consumer e-commerce shift and the $10 billion+ industry of AI-generated short-form dramas.
- Founder Alex acknowledges that his early success in competitive programming and the acquisition of his previous company, AI Factory, for $166 million by Snap, provided the capital and experience to launch Hicksfield, though the acquisition involved severe dilution.
- The company invests heavily in community building and open-sourcing projects, growing from 10 open-source projects to over 10,000 in eight weeks to create network effects.
- Founder Alex predicts that Hicksfield could reach $10 billion in revenue within the next 12 months, significantly outpacing the financial team's conservative projection of $4.5 billion.
- The company aims to become the infrastructure for AI-native distribution for direct-to-consumer businesses, positioning itself to potentially scale larger than Apple, Alibaba, and Shopify.
- Alex cites a deep-seated drive from his immigrant upbringing in Kazakhstan, where technology is viewed as a meritocratic path to success, and a personal philosophy that there are no shortcuts to hard work.
- The company is actively hiring from Kazakhstan, leveraging a top-tier talent pool in physics and math, offering 15% personal income tax rates to attract top engineering talent from Europe and Asia.
- Hicksfield has faced challenges with influencer marketing distribution, leading to a strategic shift toward owning distribution channels and creating AI-generated digital replicas of creators to solve content fatigue.
- The founder is critical of the current obsession with AI benchmarks, arguing they do not reflect real-world workflows where prompts often exceed 3,000 words and require extensive visual references.
- Alex notes that while he works 80-90 hours a week, he prioritizes spending at least three hours a week with his wife and five hours with his son, though he admits this creates tension regarding work-life balance.
- The company is building a "system of records" for media assets that leverages semantic understanding to search and organize content, a capability that traditional file storage systems like Dropbox cannot provide.
- Alex believes that the moat for future AI companies will be either delivering specific business outcomes (e.g., increased sales via AI ads) or network effects through open-source communities, rather than proprietary models alone.