The cost of Instinct’s 10% daily growth: Computing power demand doubles weekly, and the founder spends 40% of their time figuring out how to procure it.

In a Beating AI Express report, Noah Shinn, founder of personal AI agent Instinct, said the company’s user base has grown by roughly 10% to 11% daily in recent weeks, with corresponding computing power demand nearly doubling weekly. He now spends around 40% of his time deciding how much computing power to purchase in advance. The challenge is that computing power cannot be bought on-demand to match user growth. Shinn crunched the numbers: buying just twice the current computing power would be exhausted in about a week; five times the current capacity would be used up in under three weeks; while purchasing ten times the current capacity would mean tying up a large sum of capital upfront. Many resources also require booking months in advance, and emergency computing power purchases can cost 3 to 4 times more. Even if Instinct’s daily growth rate drops from 10% to 5%–8%, Shinn estimates that after 3 to 4 months of sustained growth, the company may need to prepare computing power for a 100-million-user scale. The problem is that this timeline is exactly how long it takes for new computing power to come online. Buying too little cannot support growth, while buying too much leads to inflated costs if growth slows. Shinn believes this is more difficult than scaling for regular consumer internet products. When Instagram or Facebook double their user base, they mainly handle more requests; each new user of a personal AI agent also continuously consumes large amounts of model inference resources. He added that the scaling speed Instinct is currently facing is even faster than similar challenges encountered by Claude Code and Codex in the past.