Epistemic status: I think you should interpret this as roughly something like “GenAI is not so powerful that it shows up in the most obvious way of analyzing the data, but maybe if someone did a more careful analysis which controlled for e.g. macroeconomic trends they would find that GenAI is indeed causing faster growth.”

1. Some people have asked me to consider going back to earning-to-give, and so I am considering founding another company. A major hesitation is that startups generally take 6 to 10 years to be acquired or go public. This is unfortunate for founders who believe that we are less than 10 years away from all human labor being automated.
2. It's possible that the advent of generative AI will make startups grow faster, which could provide a faster exit path. YCombinator CEO Garry Tan has publicly stated that recent YC batches are the fastest growing in their history because of generative AI.
3. However, I find that YCombinator companies have, if anything, been growing more slowly since the release of ChatGPT in 2022:

   1. Of the 20 companies which had the highest 2 year growth post-YC, only 1 (Tennr) was a 2023+ batch company, even though 16% of the companies I could find 2 year growth data for were 2023+.^[1] The average valuation of a YCombinator company two years after it goes through YC is only $13.3M for 2023+ batches, compared to $46.1M for <2023 batches.^[2]

   2. Of the 20 companies which had the highest 1 year growth post-YC, only 1 (Legora) was a 2023+ batch company, even though 34% of the companies I could find 1 year growth data for were 2023+. The average valuation of a YCombinator company one year after it goes through YC is only $7.9M for 2023+ batches, compared to $13.3M for <2023 batches.
4. These results are preliminary and subject to significant caveats:
   1. Public data about the valuation of privately held companies is notoriously limited. My guess is that these results aren't incredibly off because the largest companies tend to have the most publicly available information (and startup outcomes are so fat-tailed that only the largest companies matter), but I am sure that there is at least some amount of error.

   2. “YC-backed” is sometimes used as a synonym for “good startup,” but the most talked-about GenAI startups (Anthropic, Cursor, Wiz, etc.) aren’t YC-backed.^[3] It's possible that YCombinator startups have secularly become worse, and this effect overshadows any increase from generative AI.

   3. Relatedly, trends like higher interest rates may be depressing the valuation of startups in a way that makes these results confusing to interpret.
   4. Generative AI is very rapidly progressing. It seems plausible that technologies good enough to move the needle on company valuations were only developed, say, six months ago, in which case it would be too early to see any results.

All code and data can be found here.

Methodology

I scraped a list of all YC-backed companies from here. I then asked Claude Haiku 3.5 (with a web search tool) to identify the valuations of each company annually after it went through YC.

Valuation numbers were adjusted using the cpi python package to be in June 2025 dollars.

I then randomly sampled results to confirm that the extracted valuations were correct. I particularly manually verified the list of the top 20 1- and 2-year growth companies.

Identifying GenAI startups

It is hard to know for sure how much a company is using GenAI. It seems safe to assume that most YC companies were not using it much before the launch of ChatGPT (if only because the technology wasn’t available), and somewhat reasonable to assume that companies were mostly using it afterward (because YC pivoted very hard into GenAI startups).

I therefore use being before or after the launch of ChatGPT as a proxy for using GenAI. This is not an ideal identification mechanism for obvious reasons but I think it’s a reasonable proxy. Importantly, Garry Tan talks about “the whole batch” growing, not “the GenAI startups in the batch”.

Sources of error

I expect there are two major sources of error in this project.

1. Company valuations are generally not easily available for privately held companies.
   1. This unequivocally means that the results for smaller companies are wrong. Indeed, for many companies, I was unable to find any information about their valuation at all. However, as companies get larger, they tend to publicly disclose their valuation. I would be surprised if there were many billion dollar companies, which I was unaware of because their valuation was kept so completely private.  And since startup outcomes tend to be so fat-tailed, the average is driven almost entirely by these larger companies, and therefore I expect that the aggregate statistics are more or less correct. But it is hard to say for sure.
   2. If you have access to a better data set, I would love to hear from you!
2. The LLM may have made errors when extracting data.
   1. This most commonly occurred when there are multiple companies with the same name. For example, there are no less than three companies that are backed by YCombinator whose name is “Apollo,” and another three whose name is some minor variant like “Apollo.io.”
   2. My guess is that this would tend to inflate results for newer companies because they would have valuations which are incorrectly attributed to them, despite actually being the valuations of previously founded companies. But again, this is a non-trivial source of noise and should give people caution when interpreting these results.

These sources of errors tend to more heavily affect the results for the average statistics and don't affect the estimates of the top companies as much. I therefore would suggest believing the top 20 lists more than the information about the average statistics.

Results

The most valuable companies 2 years after doing YC

Tennr is the only 2023+ company on this list.

The most valuable companies 1 year after doing YC

Legora is the only 2023+ company on this list.

Average Growth

The fastest growing YC batch of all time was Winter 2009. This is because it was a small batch (16 companies), one of which was Airbnb.

YC has shifted to having much larger batches (Summer 2025 was 159 companies, 10x W09). This brings down the average, and I’m not sure one should read that much into the average statistics, but I will include them for completeness.

Discussion

Garry Tan’s comments

CNBC reports:

Y Combinator CEO Garry Tan told CNBC that this group is growing significantly faster than past cohorts and with actual revenue. The winter 2025 batch of YC companies in aggregate grew 10% per week, he said.

“It’s not just the number one or two companies -- the whole batch is growing 10% week on week,” said Tan, who is also a Y Combinator alum. “That’s never happened before in early-stage venture.”

Tan does not say what metric he is using to measure startup growth, and the cynical reader may suspect if he was using a real metric like “revenue” he would have said so instead of leaving it unspecified.

If Garry or someone else from YC is reading this, I would value more insight into what exactly has been growing 10% week on week.

Has YCombinator just lost the mandate of heaven?

I do think that there is a meaningful sense in which Y Combinator startups are just not the right reference class for someone who is thinking about generative AI startups. My mental brainstorm of the most valuable GenAI startups turns up mostly companies which are not YC-backed (Anthropic, OpenAI, Cursor, Harvey, Wiz, Windsurf, etc.). 

My guess of the YC companies which have benefitted the most from the GenAI boom are Scale AI and Replit; both of which were in the 2016 batch. This seems consistent with the view that AI will make a lot of valuable companies, but they will still take 10+ years to exit.

The only two post-ChatGPT companies on the top-20 lists are both GenAI companies (Tennr and Legora). This is consistent with a view that GenAI startups are doing well and it’s just that YCombinator startups as a reference class are doing worse.

I therefore do think that these results are not as broadly applicable as they might seem. That being said, it's also hard for me to find examples of quick billion-dollar AI exits outside of YCombinator. Cursor, for example, was almost an example of a company that got acquired for billions of dollars after only two years after being founded, but didn't quite make it. So I would still generally back the claim that we have not seen many $1B+ exits <4 years after founding.^[4] [5]

Stripe data disagrees, showing that AI companies are growing revenue more quickly

Stripe published this report. A key graph:

There are a couple of possible explanations for why their results disagree with my findings:

1. YCombinator companies are worse than the Stripe Top Companies. This seems likely. For example, none of the three companies highlighted in that report (Cursor, Lovable, Bolt) are YC-backed.
2. Margins for AI companies are worse, which means that higher revenues don't translate into higher valuations.  It seems unquestionably true that AI companies have unusually bad margins by software company standards. For example, it's rumored that Cursor has negative gross margins, which is almost unheard of amongst pure software companies.
   1. Cursor raised at a revenue multiple of ~20x. This is definitely on the larger side of revenue multiples, but not extremely so (Nvidia is trading at around 30x, for example). Given that Cursor has had absurd growth but is still raising at only a 20x multiple, I take this as some evidence that revenue multiples are indeed lower for AI companies, ceteris paribus.
3. Secular trends affect valuation and not revenue. The poor unit economics of AI companies mean that they need to raise large amounts of money which, in a high interest rate environment, depresses valuation but does not similarly depress revenue.
4. There is an error in my or their data or there is some statistical noise. I would be surprised if this explains the entire discrepancy but it presumably explains some of it.

Carta’s data agrees, showing that companies aren’t growing faster

Carta’s Q2 report shows that companies are taking longer to raise initial rounds:

An optimistic interpretation of this would be that companies are able to reach profitability more quickly and therefore don’t need to raise money as aggressively. Unfortunately, if this were true I think we would see that companies valuations are increasing, but Carta actually reports a decrease in average deal size YoY:

The one exception is that average seed round size seems to have increased. My guess is that Q2 just had some outliers, but if this trend holds true in Q3 and Q4 it might indicate a real sustained shift.

Overall, this data seems consistent with the view that startups are not growing more rapidly than they used to be (although Carta’s dataset doesn’t go back very far, is fairly noisy, and is subject to various other caveats^[6]).

You can still get rich quick though

While generative AI might not make it easier for people to have billion dollar companies quickly after going through Y Combinator, it also probably doesn't make it dramatically harder. My guess is that most entrepreneurs should basically assume that the results for generative AI companies are roughly in line with historical YCombinator software companies and the fact that I found them to be lower is mostly due to noise. (Whether this trend will continue is, of course, more debatable.)

The declining influence of YCombinator is maybe even a positive sign about the strength of AI: maybe AI means that you don't even need to go through YCombinator anymore and that's why the YCombinator companies aren't doing so well (the best AI companies are just successful without YC).

Further Research

If you are interested in researching this area more, I would be curious to know:

1. Does this same trend exist if you use “startups funded by {set of prestigious VC firms}” as the sample instead of YC companies?
2. Is AI companies having a lower revenue multiple actually the explanation for their seemingly lower valuation? (e.g. measure whether GenAI companies have consistently lower revenue multiples.)
3. Cherry pick the fastest-growing GenAI startups (e.g. Wiz) and compare them to the historically fastest-growing historical non-AI startups. Are they actually growing faster?

Conclusion

There are a bunch of factors which could explain the relatively sluggish performance of YC startups post-ChatGPT: high interest rates, YC being less attractive, etc. I have made no attempt to control for these and expect that, if I did, the 2023+ startups would perform better.

However, this does mean that even the most favorable interpretation of the data implies that the newer startups have their growth boosted by an amount smaller than the harm caused by factors like interest rates, which means that the benefit is relatively small.

1. ^{^}

    All dollar figures are in 2025 dollars, unless otherwise stated

2. ^{^}

    I include averages for completeness, but think that changes in e.g. how many companies YC accepts per year make those figures hard to interpret.

3. ^{^}

    Scale AI is, I think, the major exception, but they are perhaps the exception that proves the rule as they were founded in 2016 and attribute their early growth to self-driving cars.

4. ^{^}

    Which is a high bar! But it is one that founders might need their company to meet if they want to exit before their labor gets substantially automated away.

5. ^{^}

     I haven't measured the base rate here at all, so I don't feel very confident that this rate isn't actually higher than it was previously.

6. ^{^}

     Notably their dataset covers all companies, not just GenAI ones.