[ryanmercer]

But, won't most anything startups be out of business in 3 years?

Data from the BLS shows that:

"approximately 20% of new businesses fail during the first two years of being open, 45% during the first five years, and 65% during the first 10 years. Only 25% of new businesses make it to 15 years or more."

https://www.investopedia.com/financial-edge/1010/top-6-reaso....

Purely guessing that startups fail even faster. This Hubspot article states:

"All these reasons bring up one question: How many startups fail? The reality is that 90% of startups fail."

https://blog.hubspot.com/the-hustle/how-many-startups-fail#:....

[blagie]

It also depends on how one defines "fail." A majority of startup I know* are "living dead." This means:

- There are 1-20 employees

- Big-O, investors have seen no returns

- Big-O, they're cash-flow neutral and growth/decline neutral

A VC defines that as a "fail." Founders are often very happy running a small business in their domain of passion. It's a lot more fun than a big business.

* Yes, that's a very strong sample bias. I don't mean to imply a statistical sample, and explaining the types of startups I typically interact with would be an off-topic essay.

[itsoktocry]

>Purely guessing that startups fail even faster

Do they? Most businesses in the real world have to generate cash or they fold. Venture capital can keep small startups going far beyond their useful life.

[ryanmercer]

But how many startups receive any appreciable amount of VC money? How many are operating on the savings of the founder(s) and/or money from family and friends rounds? How many founders are going into personal debt to fund their startup?

And how many of them that do raise funds go crazy on a hiring spree once they get any remotely reasonable funding, while not having a functional product or a remotely sustainable user base?

According to Crunchbase only 1 in 3 startups even make it to Series A (between 2011 and 2018, the % was fairly consistent each year)

https://news.crunchbase.com/liquidity/seed-funding-series-a-....

Looking at "funding data from around 15,600 U.S.-based technology companies founded between 2003 and 2013" TechCrunch comes to the conclusion that only about 40% that close a Pre-Series A round make it to a Series A.

https://techcrunch.com/2017/05/17/heres-how-likely-your-star...

[anonylizard]

Most businesses can get loans, including unviable ones, by either

1. Burning existing equity: Owner uses own house as security to get loan, common story 2. Political pressure to get loans: This is how zombie companies are born, and they are very common outside of anglo saxon countries.

I would not say startups are subject to more or less business discipline compared to normal companies of the same caliber. Your average tech startup founder has a lot more resources and credentials to burn in emergencies compared to an immigrant starting up a restaurant, so it has to be a like for like comparison.

[Avshalom]

There's an important difference here in that AI company is basically a new market. If 50% of Mexican restaurants or low-cost database companies that start this year go out of business by 2026 it doesn't matter much because there's a pre-existing mass of them. If 85% of AI startups go out of business by 2026 that's basically 85% of ALL AI companies.

[halfcat]

On the flip side, also consider low barrier to entry as drastically inflating the denominator. The cost to start an “AI company” is virtually zero.

The reason “99% of day traders fail” isn’t because day trading is harder than being a doctor.

[slowmovintarget]

If, last year, we had seen a headline that "85% of all Blockchain Startups Will Be Out of Business in 3 Years" I don't think anyone would be shocked at that prediction.

[tlogan]

A startup is a specific category within the broader framework of "new businesses". Based on my knowledge, that about 90% of startups do not survive beyond their initial 5 years. If a startup cannot find repeatable and scalable business model in 5 years then it should close: it is just waste of time.

[Oras]

Exactly. The article seems to focus on AI startups making their own algorithms and platforms (OpenAI, Meta, Google) but doesn't mention startups who are using these platforms for niche use cases. This is similar to saying a hosting company will fail because AWS, Azure and GCP will always have more resources.

[wslh]

This. It would be great to have an AI product to catch this reasoning-baits.

Offtopic: I understand this is not a fallacy but how is it call in English when someone applies a logic that is truth for the universe but only to a subset? It is not a tautology.

[infotropy]

Cherry-picking maybe? You’re selecting only the data that supports your argument.

[swyx]

exactly. this investor is unwittingly more bullish on AI than he thinks lol