At the heart of every conversation between AI startup founders and potential investors, lies a question that neither side feels particularly confident answering: “What is the market size?”

Pinning down the market size for a technology as fluid and expansive as artificial intelligence feels like capturing a cloud in a net. The challenge is not just in the complexity of the task, but in the dynamic nature of AI itself, where today’s innovation could be tomorrow’s standard.

Most founders struggle with the market size question. Market size is difficult to estimate, varies according to what the company is currently offering, or planning on offering, and opens up an ambition risk, making founders look either too bold and unbelievable or too conservative and under-ambitious.

So why do investors ask it all the time?

From the point of view of an investor, few questions are more important than the market size one for determining the potential return of their early-stage investment. Early stage startups can pivot at any time, but one of the hardest things to change is the target market. An investor can investigate the market, estimate the potential market share of the startup and get a very very rough and uncertain – but still useful – view of where the startup can be if everything goes right. They then estimate the probability of it getting there from past experience, and the return they want to achieve on the top performers. This process, also called the VC method, is one of the few methods that allows investors to look at the value of a startup, and not just accept the given market price.

Understanding this, founders might see the market size question in a new light, not as a hurdle but as an opportunity to discuss the true potential of the company. By focusing on value, not just numbers, founders can engage VCs in a meaningful conversation about what makes the startup stand out. Of course, the question is doubly difficult to answer in the case of the new, highly uncertain, but full-of-potential, AI market. In this article, I’ll have a crack at this problem, and explain what I’d do if asked that question.

Tricky to estimate, but $203.72B. Or between $14B and $600B

Estimating market size is tricky, it is tricky already for current markets, it is even trickier for long shot, new markets like AI (but we could include crypto, VR, Space and many others.) And that is partially why it’s hated by founders.

The estimation of the market size of AI is particularly tricky as it presents itself as a productivity paradigm shift, comparable in my opinion to the invention of the steam engine.

In the following paragraphs, I’ll argue a market size for AI services of between $14B and $600B. Of course this is a huge range, and it reflects the variety of definitions and potential markets that, despite the sector’s early stage, are proliferating and confusing the picture. Through the following interpretations of the technology and the potential markets, we’ll see how the smaller (but not small) estimate is probably related to AI replacing specific use cases, markets that are already active today but that are performed in a more inefficient way. The largest estimate is where the uncertainty mainly lies, but maybe also the opportunity. By looking at different lenses, we try to estimate a number that has sufficient context to be useful to investors and founders that rightfully want to pursue these large, world-changing ideas.

How do we get to this number

We need to look into comparable industries. All prices and all products in the world are competitive, and the best solution will steal away market share from worse solutions. Of course, better solutions can open new markets and extend itself beyond the current parameters. I believe that if we consider this in our estimations, we’ll still get some interesting results.

In order to triangulate as much as possible an interesting market size, we can progress by looking at AI under different lenses.
Starting from the bottom of the stack, AI the hardware product -> faster chips to train and run AI models.

AI: the hardware product

The current market size of chip manufacturing is around 600B per year [source].

We can assume downstream AI services will increase the demand here (the main assumption behind the valuation of NVIDIA right now), but it is an interesting starting point. If we want to consider growth in this segment, we also need to consider lag times in creating production capacity. Chip manufacturing is one of the most difficult, expensive, time consuming and advanced manufacturing processes there are, with new factories lead times of up to 3 years.

AI: the motorcycle for the mind

This definition might be a bit of a stretch, it goes something like this. We use computers everyday to make us more productive. Whether it is designing a bridge or a website, but nowadays also curing a patient, caring for a crop, or catering a table, computers are fundamental to all these interactions, because they make us more productive. In the words of Steve Jobs, “Computers are bicycles for the mind”.

AI can be used in much the same way, in fact I’d argue this is the main way it will be used, and to a much larger extent. There is no “market” for this type of usage of computers, in a sense the whole hardware and software market is geared towards making humans more productive, indeed that is the definition of technology. However, at a minimum, we can look at the market for computers, as designated productivity machines, to get a baseline estimate.

The market size of computers is around 200B [source].

AI: the new interface for the internet

The closest thing we have now to a unified internet interface is search. People want to consume the internet in several ways, AI could be a powerful way to extract information, and that is why search engine leader Google has been intensely focused on this potential alternative for at least three years [source].

If we look at the market size of search, it is currently between 180B and 200B [source]

AI: the personal assistant

While it’s great to look at world defining changes, it is also good to balance them with smaller, more conservative counter-examples. In this view we look at current AI LLMs in its maybe more reductive scope, helping people as a personal assistant. Right now the market for personal assistants is severely limited by their cost, and what we saw time and time again is that when technology brings down the cost of a service, more people can make use of it. The net effect of this is normally expansionary when it comes to market size. For this reason, the following estimate should likely be a floor to our ideas of the overall market size of AI, but still useful.

The market size for personal assistants in 2023 was 14B [source]

AI: the copywriter or stock photo generator

Keeping in line with floor estimates and current beachhead use cases for AI LLMs and other current models, we find markets such as copywriters and stock photos. The market sizes for both are: 25B [source] & 4.2B [source] respectively.

Other articles estimating AI market size

Given the prominence of the AI topic and expansion, it is only natural that this article is not the first to estimate the market size. So far, among others, these articles are, in my opinion, relevant:

Interestingly every report takes a different approach in estimating the market size, and it does feel like we are already close to the point where just discussing AI in general is not going to be enough, we’ll have to look at specific products and specific customer budgets. We are already seeing important segments in hardware production, model training and development, and model usage/exploitation. This last one is the most likely to need its own category very soon.

The final estimate

Given the early stage of the market, I’d argue a broad approach is necessary, and, while taunting AI as the Next Big Thing seems very appropriate these days, we should be wary of inserting too much bias on this number. A simple average of the numbers above includes 3 big, all-encompassing visions of AI, and 3 more limited ones.

The simple average is $203.72B but we need to look at a range, a range is more helpful given all the assumptions we have inserted so far.

I believe a range between the 14-15B of “small”-targeted use cases such as replacing copywriters or stock images, and $600B for the most all-encompassing projects are the numbers that should be used for startup estimations and calculations. For a reality check, the latest Microsoft-Nvidia-Google yearly revenue is ~$140-230B, companies of this scale can come out of the AI market, the market definitely has the breath to sustain them. However both founders and investors should keep reality well in check, especially when it comes to world changing technologies.

$600B as the top of the range feels both incredibly large and quite conservative. It’s the largest current market in terms of comparable services as investigated in this article. But the smallest when compared to other literature on the topic. I believe the reason for the difference is the different purpose of analysis of the other market size articles. The purpose of this analysis is to find a practical reference for valuation. When looking at valuation, investors and founders will take these numbers and investigate the market share that a potential company will be able to get in that market. They’ll then use that market share to estimate revenue, profits and valuation. Figures that already set AI up for a world changing outcome are, in my opinion, premature and bubble prone. Even if true, the percentage and the sheer number of companies after that vision (and with serious potential to get there) is extremely small. However, these are the companies that rack up all the media attention. For the “regular” AI company, if there is such a thing, this range should be not only more realistic, but more useful, practical and informative.