The Problem AI Startups Didn't Know They Had
Creem, an Estonian fintech founded by crypto payment veterans Gabriel Ferraz and Alec Erasmus, just raised €1.8 million to solve problems most people didn't know existed. The company builds financial infrastructure specifically for AI startups that are "scaling faster than traditional financial tools can handle."Because apparently Stripe wasn't good enough for companies that think they're special because they use ChatGPT APIs.Here's the thing though
- they might actually be onto something. AI startups are scaling at speeds that break traditional finance tools. I watched one AI startup go from $50K MRR to $500K in six weeks after they launched an image generation API. Stripe literally rate-limited them during their biggest traffic spike because their volume increase triggered fraud detection. Their QuickBooks integration broke when they tried to import 10,000 transactions in a day. Their CFO was trying to forecast cash flow for a business model that didn't exist six months ago.Traditional fintech was built for predictable SaaS growth
- you know, the boring kind where you add 20% more customers per quarter and everyone pretends that's sustainable forever. AI companies don't grow like that. They explode, plateau, pivot, then explode again in completely different directions.Creem's approach is basically "what if we rebuilt financial infrastructure assuming your revenue chart looks like a heart attack monitor?" They handle payment processing, subscription management, and financial reporting for companies whose business models change faster than most people change their underwear.The founders aren't random crypto bros either. Ferraz previously built payment systems that processed millions in crypto transactions. Erasmus has a background in algorithmic trading and financial systems architecture. They understand both the technical complexity of payment infrastructure and the regulatory hellscape of financial services.Their early traction is legitimately impressive
- $1 million ARR in 10 months without a sales team suggests product-market fit for something real. That's not "we counted our founder's mom buying our product" numbers, that's actual paying customers solving actual problems.The €1.8M funding round was led by European VCs who specialize in fintech and B2B infrastructure. Not exactly the kind of money that gets wasted on buzzword bingo
- these investors write checks for boring, profitable software that solves specific problems.The timing makes sense. AI companies are graduating from "cool demo that impressed VCs" to "actual business with paying customers." That means they need real financial tools, not managing their books in a fucking Google spreadsheet.The competitive landscape is messier than Creem would like you to think. Stripe keeps adding features. Mercury is going after AI startups hard. Brex has AI company-specific products. But none of them were built from the ground up for the specific weirdness of AI business models.Look, most "AI-specific" tools are just regular tools with "AI" slapped on the marketing. But financial infrastructure for AI companies actually has unique challenges. Revenue recognition gets weird when your product is API calls that scale exponentially
- I've seen startups get flagged by their auditors because their revenue spiked 50x in a month and the accountants couldn't figure out if it was fraud. Chargeback management is different when your customers are other AI companies burning through VC cash. Tax compliance gets complicated when your revenue comes from 47 different countries and half your customers are paying in credits, not cash.The real test will be whether Creem can scale their infrastructure as fast as their customers are scaling their businesses. Building fintech is hard enough when you know what your customers need. Building it for an industry that reinvents itself every six months is genuinely difficult.If they pull it off, they could become essential infrastructure for the AI economy. If they don't, they'll join the graveyard of startups that tried to be "Stripe for X" and learned why Stripe is hard to beat.