I counted every AI startup that raised money in Q1 2021. The numbers are strange.

I have a hobby that normal people would find deeply boring. Every quarter, I go through Crunchbase, CB Insights, and TechCrunch and manually catalog every AI startup that raised a funding round. I record the company name, round size, category, and a one-line description of what they do.

For Q1 2021, I logged 127 AI startups that raised funding between January 1 and March 31. Then I spent two weekends categorizing them.

The results are... not what the AI Twitter hype would have you believe.

The big picture

Total funding across all 127 companies: approximately $8.4 billion.

But that number is wildly misleading. The distribution is absurd.

| Metric | Value | |--------|-------| | Total companies | 127 | | Total funding | ~$8.4B | | Median round size | $18M | | Mean round size | $66M | | Largest single round | $1.3B (SenseTime) | | Smallest round tracked | $1.5M | | Top 10 companies share of total | 62% |

The top 10 companies raised 62% of all money. The bottom 50 companies combined raised less than 6%. This isn't a market. It's a power law.

Where the money is actually going

Here's where it gets interesting. I categorized all 127 companies by what they actually build:

| Category | Companies | Total raised | Avg round | |----------|-----------|-------------|-----------| | Enterprise MLOps / data platform | 31 | $2.1B | $68M | | Computer vision / surveillance | 22 | $2.3B | $105M | | Autonomous vehicles | 14 | $1.6B | $114M | | Healthcare / biotech AI | 16 | $890M | $56M | | NLP / conversational AI | 12 | $540M | $45M | | Robotics | 9 | $380M | $42M | | Cybersecurity AI | 8 | $290M | $36M | | Fintech / risk AI | 7 | $180M | $26M | | AI chip / hardware | 4 | $120M | $30M | | Generative AI | 3 | $42M | $14M | | Other | 1 | $8M | $8M |

Read that last category. Generative AI: 3 companies, $42 million total. In a quarter where $8.4 billion went to AI startups, generative AI got 0.5% of the money.

The three generative AI companies were all small seed or Series A rounds. None of them were household names. (One was working on music generation, one on marketing copy, one on code.)

In Q1 2021, "generative AI" as a VC category barely exists. The money is in MLOps platforms, surveillance cameras, and self-driving cars.

Enterprise MLOps dominance

The biggest category by company count (31 out of 127) is what I'm calling "Enterprise MLOps / data platforms." These are companies building tools for other companies to deploy machine learning: model monitoring, feature stores, data labeling, experiment tracking.

Some names from the list: Dataiku, Weights & Biases, Scale AI, Snorkel AI, Tecton, Labelbox.

These aren't sexy. Nobody's making viral tweets about feature stores. But this is where the money is flowing because this is where the actual pain point is for enterprises trying to use ML. Most companies can't even get a model into production reliably, let alone worry about generative AI.

The data says the AI industry in Q1 2021 is focused on plumbing. Not poetry.

The computer vision money

Computer vision and surveillance pulled in $2.3 billion across just 22 companies, giving it the highest average round size at $105M. The big raises are almost all from Chinese companies: SenseTime ($1.3B), Megvii, Yitu.

This is a geopolitical story masquerading as a tech story. The US and China are pouring money into competing vision AI ecosystems. PitchBook data shows that Chinese AI companies received approximately 38% of global AI funding in Q1, almost entirely concentrated in computer vision.

Stage breakdown

| Stage | Companies | Total raised | Median round | |-------|-----------|-------------|--------------| | Seed / Pre-seed | 23 | $89M | $3.5M | | Series A | 34 | $612M | $16M | | Series B | 28 | $1.2B | $42M | | Series C+ | 26 | $4.8B | $140M | | Growth / Late | 16 | $1.7B | $95M |

Late-stage companies dominate the total dollars (no surprise), but the Series A count is interesting. 34 companies raised Series A rounds for AI products, which suggests a healthy pipeline of companies that have found some product-market fit and are scaling up.

The seed stage is thin, though. Only 23 new AI startups raised seed rounds in Q1. Compared to SaaS (where hundreds of companies raise seed rounds every quarter), the AI startup pipeline is narrow. High infrastructure costs make it expensive to even start an AI company, which filters out a lot of potential founders.

What I expected vs what I found

I expected more NLP companies. GPT-3 has been dominating the conversation on Twitter and Hacker News for months. But NLP / conversational AI raised only $540M across 12 companies. That's 6.4% of total funding.

I expected fewer autonomous vehicle companies. I thought the AV hype had cooled after some high-profile setbacks. Nope. 14 companies, $1.6 billion. The AV industry continues to attract enormous capital despite the product being perpetually "2 years away."

I expected generative AI to be bigger than $42 million. This is probably the biggest gap between Twitter discourse and actual capital allocation that I've ever measured. On Twitter, AI is GPT-3 and DALL-E. In the funding data, AI is enterprise data platforms and Chinese surveillance cameras.

A note on methodology

I used Crunchbase as my primary data source and cross-referenced with CB Insights reports and TechCrunch articles. I only counted companies where AI/ML is the core product (not just a feature). Rounds under $1.5M were excluded to reduce noise from tiny angel rounds.

Some edge cases were hard to categorize. A company using NLP for cybersecurity could go in either category. I defaulted to the company's self-description. This introduces some subjectivity, which is why I'm showing you the raw categories rather than just the totals.

What I'll be watching

Q2 data collection starts now. My specific predictions:

Enterprise MLOps will stay the #1 category by company count
Generative AI funding will increase, but stay under 2% of total
Total AI funding will exceed Q1 (bull market momentum)

I'll check these in three months. If generative AI breaks 5% of total funding before the end of 2021, I'll buy myself a nice bottle of sake to celebrate being wrong. (I don't think I'll be buying sake.)

If you found this interesting, you might also like:

-- dataku