Ask HN: Is it possible to start an AI product without venture capital?

[devanil]

I have an AI idea that I'm developing, but I don't have any money to burn. I also worry that a $20 subscription might not be enough to cover the costs of the most active users. Is it possible to create an AI product and start a business without any cash on hand to attract customers?

edit: I believe the best comparison is a product like Perplexity and the functionality of Deep Research.

[muzani]

gpt-4o-mini and gemini flash are quite cheap, almost to the point that you can sustain them with ads.

There's a lot of optimizations you can do too. Many seem to forget about options like procedural generation or things like fuzzy logic which could cut down cost and yet be more deterministic.

AI is also one of the kinds of businesses that should be self sustainable without funding. If it's not good enough that people would pay right away, it's probably not going to work with a more complex business model.

I would argue VC funding is either mandatory or unlikely though. AI startups move incredibly fast and odds are you'd get killed by another one without funding. The big AI companies like OpenAI and Anthropic are perfectly capable of releasing high quality competitors, and the mid tier ones like Voiceflow and Langchain are also capable of moving quickly.

[aqueueaqueue]

For a GPT wrapper sure. You can get $1k AWS or Azure credits pretty easy.

[quintes]

Sounds like you’d need more than 20$ for that kind of product idk

How about lo fi build and test it with people. If it has legs or not the people will let you know.

I’m bootstrapping atm same approach to show n tell

[anonzzzies]

Depends what you have in mind. Some things people are burning expensive calls for can be done much cheaper and on the other hand, some ideas require boat loads of expensive processing which make them far more expensive than what people are willing to pay. Think you should give some more info about what it is to get better feedback really.

[devanil → anonzzzies]

I believe the best comparison is a product like Perplexity and the functionality of Deep Research.

[anonzzzies → devanil]

Then I guess you know the answer... Those will make massive losses and yes, VC money required.

[scottndecker]

Without venture capital? Yes. Without $20? Probably not. $20 subscription doesn't need to cover costs of most active user. It needs to cover the cost of the average user (assuming you don't want to put any cash into it yourself...but if that's the case, maybe you don't believe in this idea enough?).

[devanil → scottndecker]

I believe in the idea, I just don't have the cash really. So I'm considering whether it could be an idea for later on when I have more cash.

[tv10]

20$ is more than enough to cover the costs. Check token prices.

[minimaxir → tv10]

Economics of LLMs are more complicated than that. It depends entirely on use cases, token use efficiency, and limits (if any) on usage, since if usage is uncapped some users will create negative revenue.

[muzani → minimaxir]

I remember early on AI Dungeon tried to be uncapped. The cost of a few power users was being redistributed to other users and then they decided to ditch the unlimited use plan.