Unless you’ve been hiding under a rock, you will have noticed there’s a lot of heat around AI as an investment theme right now. Octopus’s recent announcement of a £120m dedicated AI fund is one of many recent events I could cite as evidence.

In that same announcement Octopus mention that they have had three AI exits (Swiftkey, Magic Pony and Evi) so this is not a new investment trend.

It is, however, a trend that is changing. Up until this point AI exits have largely been driven by a desire to acquire talent. Even Deep Mind’s $400m sale to Google in 2014 is, I think, best understood as an acqui-hire.

Going forward two things will be different. Firstly, universities have responded to the demand for AI PhDs. Hence talent will be less scarce going forward and acqui-hires will be less necessary.

Second, and perhaps more interesting, is that it’s becoming much easier and much cheaper to build AI driven products and we are seeing an explosion in the number of AI startups with a clear path to delivering value to their customers and making profits. There were, of course, numerous companies in the previous generation of AI startups that were on this path, just nothing like as many as we are seeing now and expect to see in the years ahead.

AI startups are becoming cheaper and easier to build, because many of the underlying technologies are now mature enough to apply predictably, and because of the declining cost of cloud computing – including many AI as a service products on AWS and Google Cloud.

I liken this development to the time when cloud computing first emerged around ten years ago. Resources that were previously the preserve of cash rich companies became available to anyone who could pull together a few grand and a thousand flowers bloomed. I think we will see something similar again now.

  • Gary Mulder

    Fairly sure we’re approaching the Gartner “Peak of Inflated Expectations” when it comes to AI. There are some tasks that Machine Learning can best human performance with ease. In contrast, there are many representational issues to solve before we move on to the “Slope of Enlightenment” where AI becomes a commodity service.