Fundraising for AI Startups in 2025: What VCs Actually Ask
The questions that come up in every Series A meeting — and how to answer them.
The bar for AI fundraising has reset. "We built it with GPT" is no longer a pitch; it's an admission. Here's what VCs are actually asking in 2025 — and what separates the term sheets from the polite passes.
The three questions that tank pitches
- "What happens when GPT-6 launches?" If your moat is current model capability, you have no moat.
- "What's stopping your users from doing this with a prompt?" If the answer is "they're not technical enough," the threshold drops every year.
- "Why are you the team to build this?" Domain expertise > AI expertise. VCs have learned this the hard way.
What strong answers look like
Defensible AI startups in 2025 lean on one or more of:
- Proprietary data — a dataset that competitors can't replicate, ideally with a flywheel (more users → more data → better product).
- Regulated workflow — healthcare, legal, finance verticals where compliance + relationships are the moat.
- Sticky integrations — deeply embedded in the customer's stack, expensive to rip out.
- Distribution — a channel partnership, existing user base, or community that competitors can't buy.
Metrics that move conversations forward
$50K MRR with 4% weekly growth and ~5% logo churn beats $500K ARR with 40% gross margin and flat MoM. VCs are paying for the curve, not the absolute number. Also non-negotiable in 2025: gross margin above 60%, weekly active usage data, and at least one named lighthouse customer.
What's quietly dead
"AI-native" vertical SaaS without a vertical specialist on the founding team. "GPT wrapper" with no proprietary data. Anything that depends on OpenAI never improving in your category. Be honest about which of these describes you.
Bottom line
The era of getting funded for an AI prototype is over. Show defensibility, distribution, and discipline — or stay bootstrapped until you do.