20VC: $1BN ARR in 18 Months; The Untold Story of Higgsfield | Spending $4M Per Month on Models | Why Moats in AI are BS | Scaling a Content Team to 150 People with Alex Mashrabov

Higgsfield reached a $1bn run rate without paid ads

Higgsfield is betting that value in AI video sits in the product layer and not in owning models. It picks across open and closed models to protect margins, sells outcomes to businesses, and grows through its own content instead of paid ads. Alex Mashrabov argues this builds durable value even as other AI moats look thin.

Takeaways

• Higgsfield claims $1B run rate without paid advertising. Revenue is the last four weeks times 13, live revenue only, a method Alex says OpenAI and Anthropic also use. Business is just over 50%, mobile under 10%, the West over 70%. Growth comes from in-house content, 150-plus creatives and influencers.
• Model routing and post-training drive margins. Open-weights models with post-training give margins above 80%, against 20 to 30% for closed models. Higgsfield chooses the model itself in over 40% of cases, which Alex calls tokenomics. He expects open source share to rise, yet OpenAI and Anthropic to keep over half of dollars.
• Owning models was a bridge, not the business. Higgsfield's own image and video models took ARR from $1m to $20m in about three months, then to $100m. They now serve mainly cases customers want, such as photoshoots. Alex says lab benchmarks are gamed and video models resemble rendering engines.
• Durable value comes from outcomes and network effects. Stebbings calls moats largely bullshit and notes Lovable won as a wrapper on speed. Alex says only delivering outcomes, such as helping businesses sell through AI ads, and network effects last. Higgsfield's open source community grew from about 10 seeded projects to over 10,000 in eight weeks.
• Big labs will squeeze $20 prosumer subscriptions. Alex predicts Google and OpenAI will crush the $20 a month prosumer tier, a view Stebbings links to Canva. Higgsfield sees consumer churn fall about 30% in month one, then flatten. Net revenue retention at month 12 exceeds 300%.
• Heavy AI spending is treated as leverage. Higgsfield spends over $4m a month on internal model usage, about $10,000 per person. One employee spent over $30,000 in a week building a tool. Alex expects top engineers and creatives to reach $50,000 to $100,000 monthly, and AI handles over 60% of first-line support.

How Higgsfield found its product

1. Alex sold AI Factory to Snap for $166m, then led generative AI work there.
2. Snapchat filters showed AI media could reach huge scale, while most brands could not match social media's production pace.
3. Higgsfield spent over a year searching and burned more than $10m of its $16m seed round.
4. Conversations with eight creative directors revealed that camera control was missing from AI tools.