What are AI startups? They are young companies whose main product depends on artificial intelligence, from apps built on someone else's model to labs that train their own. If you want to see real ones as they launch, claudegpt.si, a free AI news site, writes up their announcements with the sources linked.
A plain definition
A startup is a new company trying to grow fast, usually with money from investors rather than from years of sales. An AI startup is one where AI is the product, not a feature added later. Take the AI away and there is nothing left to sell.
That covers a wide range. A two-person team selling a writing app and a lab training large models are both AI startups, even though their costs and risks are very different.
The main kinds of AI startup
Most fall into one of four groups. Each example below links to a news story with its sources.
1. Model makers. These train their own AI models and sell access or give the weights away. Recursive released Flint, an open coding model you can run on your own computer.
2. Tool and app builders. These build products on top of models, often for one job. Entire built a Git forge for AI agents and developers, a place to store and review code that agents help write.
3. Infrastructure companies. These sell the computing that runs AI. General Compute raised money for an inference cloud that avoids NVIDIA chips.
4. AI for one industry. These apply AI to a specific field. Pathos AI works on AI agents for drug development, with backing led by Bristol Myers Squibb.
| Kind | What they sell | Example |
|---|---|---|
| Model maker | Access to a model, or open weights | Recursive |
| Tool or app | A product built on models | Entire |
| Infrastructure | Computing power for AI | General Compute |
| Industry AI | AI for one field | Pathos AI |
How AI startups make news
Three kinds of event put an AI startup in the headlines:
- A launch, when it releases a product or model.
- A funding round, when investors put in money. Rounds have letters: a Series A is usually the first big one, and later rounds such as a Series E come once the company is larger.
- A deal, when a bigger company buys it. Intel, for example, agreed to acquire the quantum software company Classiq.
How to tell a real one from hype
Many AI products are a thin layer over a model from a big lab, and some vanish within months. A few quick checks help:
- Does the company explain what it built on its own site, in plain words?
- Can you try the product, or read independent coverage of it?
- Does the news story you read link to where its facts came from?
claudegpt.si lists the sources at the end of each article and shows when it was published and last updated, which makes the third check easy. You can follow it on the site or through its RSS feed; there is no email newsletter.
If you want a full directory of startups by sector, a database or Y Combinator's public list of the companies it has backed will show more names. To understand what a new AI company actually does, a sourced news story is the better read.
FAQ
What is an example of an AI startup? Recursive, which makes open coding models, and Entire, which builds developer tools for AI agents, are two recent examples.
Are all AI apps made by startups? No. Large companies such as Google, Microsoft and Apple also build AI products. Startups are the young, fast-growing companies among them.
Where can I read about new AI startups? claudegpt.si covers AI startup launches, funding rounds and deals for free, with sources listed in every article.