How I Made $20M with AI in One Year

Summary

Ben Cera describes building Polsia, an AI product intended to help run a business, from a solo-founder project into a company reporting a $20 million annualized run rate in less than a year. The account centers on launching quickly and acquiring customers, then reducing AI costs and stabilizing the product before scaling further.

Timeline and business decisions

• November–December — Build and launch

  • After developing the idea while traveling in Japan, Cera worked alone from Paris on Polsia’s first version.
  • To avoid extending development indefinitely, he set a launch date despite having an unfinished feature list.
  • After launch, he emailed users of his earlier products, posted actively on X/Twitter and LinkedIn, joined AI-agent discussions, and tested Meta ads. These efforts brought in the first 100 paying customers.

• San Francisco — Generate attention and raise capital

  • Cera moved to San Francisco to meet people and increase the company’s visibility.
  • He configured Polsia to respond to investor emails and made the provocative claim that the AI would raise a funding round itself. The stunt attracted investor interest and helped produce a $30 million fundraise at a $250 million valuation.
  • The funding brought a new challenge: the product needed to become more reliable before handling greater demand.

• Infrastructure and unit economics

  • The company worked with an AI-infrastructure partner and brought in engineers to address bugs and harden the platform.
  • AI model costs reached about $1.2 million in one month, with most spending going to a commercial provider. Cera said this was not sustainable.
  • The team began shifting workloads to open-source models running on rented GPUs. Cera described a goal of reducing the monthly bill from roughly $1 million to about $50,000, while monitoring whether the lower-cost models were capable enough. The subtitles do not confirm that the $50,000 target was reached.

• May — Orchestrated launch

  • At roughly an $8 million run rate, the company coordinated a launch video around the AI fundraising story.
  • The team organized amplification from contacts and reached out to the press. The video received millions of views on its first day; reactions ranged from enthusiasm to criticism of the product’s bugs and the video’s quality.
  • Cera judged the launch successful for its main purpose: generating visibility.

• June — Lower price to broaden adoption

  • After GPU and model-cost changes improved the cost structure, Polsia’s price was cut from $50 to $20 to reach more users.
  • Cera said the reported run rate fell from $11 million to $8 million in one month, but customer acquisition and product usage improved. He viewed wider access and increased engagement as valuable, even at the cost of near-term revenue.
  • The product was designed to consume compute when users engaged with Polsia, including by replying to its daily email.

• July — Stabilize before scaling

  • More customers exposed reliability problems. Cera and engineering collaborators spent much of the month fixing bugs and improving affordability and platform stability.
  • Revenue stopped falling and began climbing again after this work.

• August–September — Growth accelerates

  • Cera took a break in Europe but continued working as the product grew.
  • He said the business went from zero to a $10 million run rate in about five months, with growth accelerating from roughly $1 million to $10 million between March and June.
  • By the end of September, he reported reaching a $20 million run rate.

• Next phase — Rework the product

  • After reaching that milestone, Cera returned to Japan, where the original idea began, to develop a more ambitious next version of Polsia.

Frameworks, processes, and actionable lessons

• Set a launch deadline to avoid overbuilding: Ship a first version once it is usable rather than waiting for an unlimited feature list to be complete.
• Use founder-led, low-cost customer acquisition: Contact existing users, participate in relevant online conversations, and test paid ads.
• Treat publicity as a coordinated campaign: Pair a memorable story with a launch video, personal-network amplification, and press outreach; evaluate success against visibility goals.
• Manage AI economics as a core operating constraint: Track model costs, test lower-cost alternatives, and weigh model quality against sustainable margins.
• Use pricing as a growth lever: A price reduction can lower short-term revenue while improving adoption and engagement; monitor both outcomes.
• Stabilize before adding more growth: Fix reliability and cost issues exposed by new customer volume before pushing the product harder.
• Use engagement to guide product direction: Cera described using user interests and activity to suggest new projects or ideas when someone loses interest.

Key metrics reported

• Fundraising: $30 million raised at a $250 million valuation.
• Run rate: Approximately $8 million by May, about $11 million before the June price cut, then $8 million after it; $20 million by the end of September.
• Growth: From roughly $1 million to $10 million in run rate from March to June; zero to $10 million in about five months, according to Cera.
• AI costs: Approximately $1.2 million in one month; a target of about $50,000 after moving more workloads to open-source models and rented GPUs.
• Customer acquisition: First 100 paying customers after launch.
• Pricing: Reduced from $50 to $20.
• Launch reach: Millions of video views on day one.

All financial figures and timelines are as stated in the subtitles; the subtitles do not consistently distinguish run rate from other financial measures.

Presenter/source: Ben Cera, founder of Polsia. An AI-infrastructure partner is named inconsistently in the subtitles as Sapium/Sapion.