AI startups alone raised more in July than every startup did a year ago

The money is concentrated in very large rounds. July set a record for billion-dollar rounds; the biggest included Safe Superintelligence at $5 billion, Moonshot AI at $3.5 billion and Kling AI at $2.8 billion, alongside the defence companies Helsing and Quantum Systems. It follows a record first half in which startups raised about $515 billion worldwide.

The number that matters to you

For founders who aren't training models, the useful figure is the other $20 billion. That's the pool that SaaS, fintech, developer tools and consumer startups share, and on its own it's well below the roughly $32.5 billion all startups raised in July 2025 (a FER calculation from Crunchbase's figures).

Put simply: the headline says venture is booming, but for most companies the room is tighter than it was a year ago. A handful of very large AI rounds are doing the lifting.

That cuts two ways, depending on what you run.

If you're raising

Expect to be asked about your AI story, even when the honest answer is "we use it, we don't sell it". Investors aren't asking because they want a chatbot bolted on. They're asking whether AI makes you cheaper to run, faster to build or easier to replace. Have an answer to each.

A bolted-on AI feature added to satisfy the question is worse than none. It tells an investor you're chasing the room rather than running the business.

Here's a version that works. Fill in the brackets:

• "We use AI to [a specific job, such as drafting first-line support replies], which means [a specific result, such as one person covering what took three]."
• "We don't sell AI as the product, because our customers pay for [the outcome they actually buy]."
• "If models got ten times better tomorrow, it would help us by [X], and it wouldn't replace us because [what a model can't easily copy: your data, relationships, distribution or place in the workflow]."

If you can't fill in that last line, better to find out now than in a pitch meeting.

If you sell to startups

Agencies, consultancies and freelancers who sell to startups should read the same numbers differently. The companies with fresh money right now are disproportionately AI companies, and they need help with the things they aren't built to do: brand, content, demand generation, hiring, finance, legal.

That's a reason to look at your pipeline, not just your positioning:

1. Pull the funding announcements in your niche from the past two months. Crunchbase News publishes a monthly round-up, and your local tech press will have the rest.
2. Pick the five companies that look most like your best existing clients in size and stage, not just sector.
3. Look at who they've hired in the last quarter. A new head of marketing or a first growth hire usually means budget, plus a gap they'll want filled in their first 90 days.
4. Write one specific note to that person about the problem you'd expect them to have in month one. Not a capabilities deck.
5. Track replies for a month before you decide whether AI startups are a niche worth chasing.

Where it goes wrong

Two traps. Founders see the headline, assume money is easy, and spend months on a raise that was never going to happen at the valuation they wanted. And services firms rebrand as "AI agencies" overnight, which makes them sound exactly like everyone else on the shortlist.

The Monday-morning version: if you're raising, write your three AI sentences and test them on someone who'll be honest with you. If you sell to startups, pull last month's rounds in your niche and pick five names.