$442 Billion Poured Into AI, But 67% Went to Just 10 Companies
By Fast AI Startup Jobs
Venture capital loves a narrative, and the current one is that AI is eating the world. Money is pouring in, every pitch deck mentions foundation models, and the headline totals look enormous.
But that headline number hides a distribution so lopsided it should change how you think about where to work.
We pulled funding data on the 1,044 AI startups we track at Fast AI Startup Jobs. The shape of it is not what the headlines suggest.
Where the money actually went
Total funding across the dataset comes to $442 billion. That sounds like an industry swimming in cash, until you see where it landed:
- OpenAI alone: $168 billion, 38% of all AI startup funding
- Top 3 companies: 57.8% of all funding
- Top 10 companies: 66.6%
- Top 35 companies (everyone above $1B): 75.9% of all capital
Thirty-five companies, about 3.4% of the dataset, hold more than three-quarters of the money.
The Gini coefficient for this funding is 0.896. Global income inequality sits around 0.7. So capital among AI startups is spread more unequally than income across the entire planet.
Mean vs. median
The mean AI startup in the dataset has raised $444 million. The median has raised $43 million, a 10.3x gap. A few mega-rounds from OpenAI, Anthropic, and xAI pull the average way up while the typical company runs on a small fraction of it. When someone quotes you an "average" funding figure for AI, this is why it's close to useless.
The full percentile spread:
| Percentile | Funding |
|---|---|
| P10 | $9.3M |
| P25 | $20M |
| P50 (Median) | $43M |
| P75 | $138M |
| P90 | $450M |
| P99 | $3B |
The company at the median has raised about $43 million. Enough to build a real product and a small team, nowhere near the billions that make the headlines.
The $10M to $50M band
The biggest single cluster sits in the $10M to $50M band: 436 companies, or 41.8% of the dataset.
These companies have raised enough to ship a product and hire a core team, but not enough to be safe. They're still proving the business works.
If you're weighing an offer from a company in this band, take it seriously but not on faith. It's real, it has funding and customers, and it can still fail. Look hard at the fundamentals before you sign.
Funding gets bigger every year
The median raise for AI startups has been climbing:
- 2022: $24M median
- 2024: ~$40M median
- 2026 YTD: $75M median
Two things follow from this. It's more expensive than ever to build an AI company, between compute costs, the talent war, and data. And investors are still willing to pay the premium, at least for now.
That rise isn't spread evenly. The companies raising big rounds are raising bigger ones, and the distance between them and everyone else keeps growing.
126 companies in the danger zone
We found 126 companies that match a worrying pattern: more than two years since their last funding round, and fewer than 50 employees.
These may be running low on runway. Some couldn't raise a follow-on, some chose not to, and some are quietly winding down. Not all are doomed; a few are probably profitable or in deep stealth. But stale funding plus a small team is a red flag worth taking seriously.
Before you accept an offer from any AI startup, check when they last raised. If it's been more than two years and the team is small, ask the runway question directly instead of assuming it away.
Capital per employee
Funding divided by headcount is one of the more revealing numbers you can compute. It hints at burn rate and how a company spends:
| Company | Capital per Employee |
|---|---|
| OpenAI | $47.6M |
| Safe Superintelligence | $30M |
| Typical AI Startup | ~$1-5M |
| Quora | $250K |
OpenAI's $47.6M per head is not efficiency by any normal standard. It's a moonshot, burning cash to stay at the frontier. Safe Superintelligence looks similar at $30M: a small team sitting on a war chest, optimizing for long-term research over revenue. Quora is the opposite at $250K per head, a mature operation lean enough to run without constant fundraising.
As a sanity check before you sign: a very high number usually means "spend now, figure it out later," and a very low one means bootstrapped or barely funded. Neither is automatically good or bad, but both tell you something about the culture you'd be walking into.
Funding by category
Where is the capital concentrated by industry?
| Category | Total Funding | # Companies |
|---|---|---|
| LLMs / Foundation Models | $240.3B | 16 |
| AI Infrastructure | $56.3B | 69 |
| FinTech | $15.1B | 108 |
| Cybersecurity | ~$12B | 65 |
| Healthcare AI | ~$10B | 80 |
LLMs and foundation models dominate: $240.3 billion across just 16 companies, roughly $15 billion each. This is the OpenAI/Anthropic/xAI tier, the people actually building the models. AI infrastructure is second at $56.3 billion across 69 companies, the picks and shovels everyone else builds on. FinTech has far more companies, 108 of them, but only $15.1 billion between them, split into smaller and more specialized bets.
Be skeptical of category labels. "AI" isn't one industry. A cybersecurity company that uses AI and a foundation-model lab share a buzzword and almost nothing else, including how they're funded and how they hire.
Funding and hiring barely move together
The most counterintuitive number in the set: funding and hiring barely track each other.
The Pearson correlation between total funding and number of open roles is just 0.356. Raising more money does not reliably predict more hiring.
And at any given moment, 85.7% of the companies have zero open roles. Most AI startups aren't hiring most of the time. They hire in bursts, usually right after a raise.
Open roles by funding band:
| Funding Band | Avg. Open Roles |
|---|---|
| < $10M | 0.3 |
| $10M - $50M | 0.4 |
| $50M - $100M | 1.4 |
| $100M - $500M | 6.9 |
| $500M+ | 23.9 |
The leap from the $100M to $500M band (6.9 roles) to the $500M+ tier (23.9) is the real story. The mega-funded are building big teams. Everyone else hires one careful seat at a time.
The Midjourney exception
Not every successful AI company plays the venture game. Midjourney is the standout: bootstrapped, 51 to 200 employees, zero VC funding.
They built one of the most popular AI products in the world without raising a single venture dollar. They're profitable, they're growing, and they answer to no one but their users.
It's rare, but it's a reminder that venture money isn't the only path. Don't write off a bootstrapped company. It might be the steadiest employer on your list.
How to use this before an interview
Most of this data is public, which means you can run the same checks on any company before you walk into a room with them.
Pull their total funding and find the band they're in: the $10M to $50M crowd is real but unproven, the $500M+ crowd is flush but often burning hard. Check the date of their last round, because more than two years quiet with a team under 50 is a question you should ask out loud. Divide funding by headcount, since north of $20M per person means aggressive spend and under $500K means lean, and either one shapes the job. And ignore the category label on the homepage. What matters is the actual product, the actual customer, and whether anyone is paying.
Put it together and the picture is narrower than $442 billion suggests. The median company has $43 million, real but not safe. A few dozen giants are doing most of the hiring, and 85% of the field isn't hiring at all on any given day.
Chasing the biggest names or the fattest rounds is the obvious move, and usually the wrong one. The companies worth your time are the ones whose specific numbers hold up when you actually look: runway, burn, customers. You can look. The data is mostly public.
Data sourced from funding records of 1,044 AI startups tracked on Fast AI Startup Jobs. Analysis date: March 2026.