Why This Is The Best Time To Build A Generative AI Startup


I said it in the title of this post, and I’ll say it again.

This is the best time to be building a Generative AI startup.

Most people will probably say the exact opposite though, because of reasons like:

  1. There’s no moat (a competitive advantage- technological or otherwise).
  2. Anyone can build a competitor.
  3. Larger incumbents simply need to incorporate a new AI feature to kill your product.

These points are all true, and yet, I disagree with their conclusion.

I believe (and I’ll explain why through this post) that there is a small window of opportunity right now for small teams/solo founders to exit for low tens of millions of dollars in the next few years.

But before I dive in, here’s a quick overview of the sections of this article:

  1. State of the current Generative AI landscape
  2. Who should start a Generative AI company?
  3. How much will Generative AI startups be worth in 3 years?
  4. What are ‘Special’ AI startups?

State Of The Current Generative AI Startup Landscape

Sequoia Capital, one of the world’s top VC firms recently released this graphic recently which went viral:

The instant feeling one gets from seeing this image is that, this space is REALLY crowded. As with any ‘hot’ new frontier- there is this initial barrage of companies that raise money and try to make it during the initial goldrush days.

But look a little closer.

Just peer under the hood.

…and you start to realise a lot of them are basically the same product.

They all use the same set of models, such as Open AI GPT-3, DALL-E-2, Stable Diffusion, etc.

And that’s something many people say about this space. That there is very little technological differentiation between Generative AI companies. They can be competed with anytime, very easily, because they don’t own the underlying technology that they use.

Which is true.

It’s companies like Open AI, Facebook, etc that own these models are driving the core innovation here.

That doesn’t mean these Generative AI applications aren’t driving value for consumers. They are.

Think of the underlying models as the application ecosystem built by Apple (iOS). Apps built on iOS are valuable to varying degrees- some earn zero money, while others have millions of users and are billion dollar companies.

So basically- while Apple profited immensely from owning the ecosystem in which apps are built, companies that own AI models will profit obscene amounts of money- because all startups will build on top of those.

So does this mean that everyone should be starting a Generative AI company?

Will it be as lucrative as making mobile apps back in the early 2010s?

Let’s dissect.

Should You Be Starting A Generative AI Company?

The answer isn’t simple and there are 3 things to think about.

  1. Ease of starting up in this space
  2. Current trend towards unbundling existing applications
  3. What happens in crowded markets.

Let’s look at each of these closely.

#1 It’s Never Been Easier To Build An AI Startup.

The barrier to entry to building a software startup used to be the ability to code.

Now that barrier is falling.

Every.

Single.

Day.

As someone who wants to learn to code, but always hesitates- this tweet by Steph Smith saying it took her 300 hours to learn to build several projects on her own made it seem…doable.

And the best part is that she did this before the explosion of tools like Github Co-Pilot and CHAT GPT-3.

For those who don’t know- Co-Pilot is like Google’s autosuggest feature which ‘suggests’ the completion for your search terms- only, for code.

It will prompt you to finish your line of code with a suggestion and, as per my research, upto 40% of code being created on Github is with the help of Co-Pilot suggestions.

Chat GPT takes this to another level – you can ask it to write a piece of code, troubleshoot particular lines of code, etc and it’ll do it for you.

The crazy part?

The introduction of Co-Pilot and then GPT-3 didn’t happen over the course of years.

No, all this happened within 6 months of each other.

My point being that today, if she were to start, she would probably be able to get the same results with ~100 hours of effort with the help of these AI tools.

Which in turn means that the barrier to entry to build Generative AI companies just went down drastically.

Soon enough, the only programming language remaining will be English- you tell the model what to do, and it’ll write the code, and convert it into a piece of software- just for you.

But forget about that for a minute. That’s still probably 5-10 years away (most people’s guess).

Just remember the main point which is- it’s getting easier by the day to build a start-up in this space.

#2 Unbundling Of Existing Applications

Jasper AI came out at the start of the Generative AI boom, and has built itself as an all-purpose writing tool.

It’s used for blogging, copywriting, social media posts- you name it. It came out as a smart writing companion tool for a wide array of written content formats.

Recently, more AI tools have come out that have unbundled Jasper AI and offer even more specific and tailored solutions.

Just in the last 1.5-2 months, I have personally tested out beta versions of over 3 AI tools specifically tailored towards writing SEO optimised blog articles- which is just one part of the entire solution Jasper offers.

This is happening across Industries. Unbundling of existing applications, such as Adobe Photoshop, into AI offering specifically for underserved, niche markets.

This is happening, at scale, across Industries, right now.

In the extreme case- this means a glut of options for the consumer to choose from- and this makes the market more and more competitive.

Very good for consumers.

Not so good for startups.

#3 Consolidation Of Fragmented Markets

Look at the entire thing from a business standpoint.

Since generative AI companies don’t have a technological moat- what differentiates them?

Business metrics like – branding, distribution, and customer loyalty.

Like any other fragmented market, the companies that are able to get ahead in the game will start acquiring smaller companies, until the various Generative AI applications consolidate and only a few large ones survive that can service the entire market’s needs.

Ultimately, one could even argue that incumbent software companies like Adobe just need to wait, watch the market and strategically acquire promising AI startups that might mount a credible challenge to their product later on.

The large incumbents already have distribution, a solid customer base… and can simply acquire the smaller players which fit them best, strategically.

So, with regards to Generative AI:

In a market where there is very little to differentiate between offerings, the one with the biggest distribution will win.

Which points to large incumbents being the ultimate winners.

That is, as long as they don’t allow themselves to be out-innovated by a start-up, like Open AI has disrupted Google Search, recently.

But I digress.

So, now that we understand the state of the Generative AI startups, what does this mean for founders?

Is starting a company in this landscape a good bet?

Will they be able to make any meaningful amounts of money in such a competitive space?

Here’s my take:

How Much Will Generative AI Startups Be Worth In 3 Years?

As per me, here are the various scenarios for Generative AI startups over the next 3 years (in order of likelihood):

  1. Go bust due to immense competition.
  2. Get acquired by competitors ($0.5-10 M range)
  3. Get acquired by a PE firm($10-50 M range).
  4. Get acquired by a large company – ($50 M plus range).

A lot of companies will likely go bust.

Very few companies will qualify for pt 4, i.e large payouts greater than $50 M, but I do think a sizeable number of ‘special’ companies will be picked up by competitors or PE firms for the low tens of millions.

And that’s where the opportunity exists today.

To be a part of that sizeable chunk of companies though, I believe the startups have to be ‘Special’.

What Does A ‘Special’ AI Company Mean?

As per me, a ‘Special’ AI company is one that:

  1. Uses AI as an unlock to solve unsolved problems, or make existing solutions even more frictionless.
  2. Focuses on solving the customer pain points and is obsessed with the customer journey, understanding their target customer personas, etc.
  3. Builds features/solutions that customers care about.

It all comes down to the same fundamentals of building businesses- AI is just the new medium.

The sooner startups realise that they aren’t building cool new ways of doing stuff- they are building the next generation of solutions that are hyper targeted towards various customer personas, solving their problems, at scale, that is when they’ll start on the path to becoming special.

So- why did I say at the beginning of this post that this is the best time to build a generative AI startup?

Because, over the next few years, we have a short window where:

  1. Large incumbents would be rolling out AI platforms relatively slow.
  2. The performance of Large Language Models like Chat GPT will improve exponentially.
  3. It’ll be easier for ANYONE to build a company (this is good and bad both- depending on how you see it).
  4. If you can find a niche, serve it well and gain customers- you will win.
  5. You don’t need big teams or lots of funding to build cashflow positive startups in this space.

All in all, if executed correctly- founders that raise little money/bootstrap will make life-changing sums of money and exit relatively quickly (within 3 years).

That’s what, according to me, makes this space so exciting.

Anyway- that’s my analysis. What do you think?

Whether you agree or disagree- comment below and let me know.

Shubhankar Chaudhary

I used to operate a Defence Startup. In my free time, I like to write about personal growth, entrepreneurship and my journey on both these fronts.

Recent Posts