# The AI Gold Rush: Are Investors Paying for Technology or Future Expectations?

Author: Norbert Biedrzycki
Published: 2021-11-15
Language: en
URL: https://norbertbiedrzycki.pl/en/the-ai-gold-rush-are-investors-paying-for-technology-or-future/
Summary: When capital is almost free, the future becomes very expensive. That is the environment in which AI companies are being priced this year.

**When capital is almost free, the future becomes very expensive. That is the environment in which AI companies are being priced this year.**

I cannot remember a year like 2021 for technology investment. Global venture funding is on track to exceed $600 billion, more than double the previous record. Rounds that used to take six months close in two weeks. Term sheets arrive before the data room is ready. Funds that once focused on late stage public equities are writing early stage cheques.

AI sits at the centre of this wave. Databricks raised at a $38 billion valuation in August. Data and AI infrastructure companies trade at revenue multiples that would have been unthinkable three years ago. Almost every pitch deck I see now contains the words "machine learning", "AI powered" or "proprietary algorithms".

**The labelling problem**

This is where I start to worry. A study by MMC Ventures in 2019 found that around 40% of European startups classified as AI companies showed no material evidence of using AI in their products. I see no reason to think the ratio has improved in a year when the label commands a premium.

The challenge for investors is not that these companies are dishonest. Most use some form of statistical modelling. The challenge is that "uses AI" and "has a defensible advantage because of AI" are two completely different statements, and only the second justifies a premium valuation.

In technology due diligence I now ask a very simple set of questions. What would happen to this company's revenue if a competitor used an off the shelf model from a cloud provider? How much of the claimed advantage comes from the algorithm, how much from the data, and how much from the distribution? Is the AI the product, or is it a feature that customers will soon expect for free?

The answers are often uncomfortable.

**What cheap money does to judgement**

There is a reason why near zero interest rates inflate the valuations of growth companies. When the discount rate falls, cash flows that are ten years away suddenly look almost as valuable as cash flows today. AI is the perfect asset for this environment, because almost all of its value lies in the future.

But that relationship works in both directions. If rates rise, the same arithmetic will compress the value of long dated expectations quickly. Investors who are paying fifty times revenue for an AI company today are not just betting on the technology. They are also, implicitly, betting on the cost of capital.

As a former CEO of listed companies I have seen what happens when the market changes its mind about a narrative. It rarely happens gradually. Multiples do not fall by ten percent. They fall by half, and the companies that raised at the peak face down rounds, layoffs and difficult conversations with their boards.

**Separating substance from story**

So how should a disciplined fund approach AI this year? In my view by looking for four things.

First, proprietary data that cannot easily be bought or replicated. Second, evidence that AI improves unit economics, not just the demo: lower cost to serve, higher conversion, better retention. Third, distribution, because the best model rarely wins without access to customers. Fourth, a team that understands the business domain as deeply as it understands the technology.

A company that has all four deserves a premium. A company that has none of them but uses the right vocabulary is a story, and stories are priced by the mood of the market, not by fundamentals.

*Investment question: How much of an AI company's valuation represents actual technological advantage, and how much represents expectations about the future?*
