As my long-time readers know, Rebel Fund builds and maintains the most comprehensive database that exists of Y Combinator startups and founders (perhaps outside of YC itself) with hundreds of thousands of new data points added each year.
I’ve now published dozens of data-driven posts on YC startups based on the trends we can see in this data, including analyses by cohort On the Y Combinator Summer 2022 batch and more recently On the Y Combinator Winter 2023 batch (written with GPT-4).
Now the tradition continues with this analysis of the YC Summer 2023 (S23) batch, once again with the help of GPT-4's AI-powered data analysis features. I’m taking a slightly different approach this time though, by 1) leaning on GPT-4 only for charting and not drafting (turns out AI’s writing style can be quite dry!) and 2) looking at trends across recent batches and not just a snapshot of the batch.
So, without further ado, let’s see what GPT-4 and I found most interesting. We’ll start with the two trends I’m asked about most often, YC batch sizes and valuations:
The YC S23 batch was the smallest one in recent memory with ~220 startups and had the lowest application acceptance rate in YC history at <1%. It felt like a very high-quality batch from an investor standpoint, though it will probably take a few years to see that in the outcome statistics.
Valuations have remained fairly consistent over the past few batches at around a ~$20M post-money SAFE cap on average, though they dipped slightly since last summer. Valuation caps got as high as ~$30M in early 2022, but were closer to ~$15M pre-covid, so current valuations are much more like a return to normalcy than a crash (late-stage valuations have fallen more sharply).
Next let’s look at some geographic trends, which are quite telling:
YC has been noticeably refocusing on American startups, with US (and a few Canadian) companies representing nearly 90% of the S23 batch, and the balance being almost entirely from Europe.
This is a sharp deviation from YC’s previous trend towards emerging market startups, mostly from India, LATAM, and SE Asia, which had been approaching ~40% of YC batches just a couple of years ago.
I believe this could be for several reasons:
- US startups just tend to grow faster — we see this in Rebel’s data and YC surely sees it in theirs
- US startups have better access to capital — this is especially important in the current market environment
- YC batches went back to 100% in-person — this creates more friction for non-US startups, especially from emerging markets
- YC is under new leadership — its new president, Garry Tan, was a YC partner in the early years and seems to have a “back to the basics” approach to running the program
Now let’s look at some sector trends across recent batches:
This is another indication of YC getting “back to the basics.” The accelerator has always had a preference for B2B startups (see On the shifting landscape of YC batches (2023 update)) but now more than ever, with >70% of YC startups building something for SMB or enterprise customers. Why? Because that’s where the money is!
The biggest loser industry-wise across recent batches is fintech. I have always believed — and still believe — this is a great sector, but a lot of the low-hanging fruit has been picked now, it’s been negatively impacted by high interest rates and recent regulatory changes, and it doesn’t lend itself to new AI-powered applications as much as other sectors (more on that below).
A cool way to look at recent YC sector trends is “word clouds” showing the most commonly used startup tags by YC batch:
Here we can easily see the persistence of B2B, the fall of fintech, and most notably, the rapid rise of AI.
Over 60% of the S23 batch could be classified as AI startups, up from 40% of the W23 batch and 15% of the S22 batch. I’ve been to every YC Demo Day since my own in 2013, and invested in every YC batch since 2017, and I’ve never seen a single new technology become so dominant so quickly.
It’s worth noting that AI isn’t really a sector as much as a platform shift, and AI startups are building new applications across many different industries. Amongst S23 batch startups, the vast majority are building AI applications (predictably) in the B2B space, followed by healthcare, consumer, and fintech — but I don’t think any technology sector or startup won’t be at least touched by AI in the months and years ahead.
There are dozens of other dimensions that we assess and score YC startups on at Rebel via our propietary Rebel Theorem 2.0 algorithm, and next I’ll highlight a few that I think are especially illlustrative of recent YC batch trends.
The chart above shows the average team size and co-founder count of recent YC startups by batch at the time of their Demo Day.
The most noticeable trend is that average team sizes have gotten much smaller, down from ~6 team members in S22 to just ~3 in S23, which shows just how much earlier in their journey YC is admitting companies. I think this is another example of Garry realigning YC with its historical roots.
Despite team sizes getting smaller, the number of co-founders at YC startups has stayed quite consistent at ~2.25 on average. YC has long had a bias against solo founders, and as you’ll see below, the vast majority of S23 startups have 2 or 3 co-founders.
To better understand what these co-founders look like, let’s look at recent trends in their years of work experience and co-founder experience:
There’s a clear downward trend in average years of work experience, from 10 years in the S22 batch down to 8 years in S23. This is yet another example of YC getting back to its roots, as the prototypical YC founder has historically been young and technical. Assuming most YC founders started working around 22 years old, this means the average founder age has decreased from ~32 to ~30 years.
Interestingly, despite YC founder work experience trending downwards, their years of co-founder experience has trended slightly upwards to ~3.5 years on average. YC has likely seen that same thing in their data that Rebel has in ours — years of co-founder experience is positively correlated with startup success.
Last but not least, let’s look a key indicator of startup potential, what percentage of YC companies have a co-founder who has taken a previous startup all the way to acquisition:
8% of S22 batch startups had acquired co-founders on their team, but only 3% of S23 startups. This is yet another sign that YC is trying to get back to what it’s always done best — helping to turn relatively unproven founders into successful company builders.
While this may surprise you, it’s not necessarily a good sign if a YC startup has exited co-founders on their team. What we’ve seen in our data at Rebel is founders with an acquisition in the past are generally more likely to be successful than their non-exited peers, however, there may be adverse selection at play when an exited founder goes through YC (i.e., the best exited founders may not be willing to give up 7% of their equity to YC when they already have a strong Silicon Valley reputation and access to capital).
It’s a complicated story though, and you really have to look at these on a case-by-base basis, because there are plenty of successful exited founders who choose to go through YC a first or even second time (one example from my YC batch in 2013 was Parker Conrad, who famously founded YC unicorns Zenefits and then Rippling).
There are dozens of other YC startup trends we could look at from Rebel’s data, but I think this is good place to stop for now. Hope you enjoyed!