AI investment activity: trends of 2018 - Issue #8

Hi all,

Those of you who have subscribed to my newsletter via metaverse.vc, please don’t be surprised, I’m migrating from a personal website to getrevue. There will be no emails from metaverse.vc, so there’s no need for you to subscribe from it.

Also, I am moving from weekly/biweekly reviews towards more thoughtful monthly or so researches.

This review highlights trends in launching/investing in artificial intelligence (AI) startups from 2018. It contains an analysis of 47 AI startups that were launched in 2018 and managed to raise at least $1M each. Also, it includes a more broad overview of some longer term trends. I cover the US, Canada, Europe, and Israel in this review. 

You also may find interesting my overview of Technologies and startups that hack the brain, and a research on Data network effects for an artificial intelligence startup. These pieces done quite well on Medium and were viewed more than a thousand times each.

Best,

Peter

Having reviewed data of AI startups launched in 2018, I could highlight the following five trends:

1.    AI hype slowdown – less startups were launched, more thorough investment approach took place;

2.    Emergence of a new wave of AI startups that solve challenges created by other AI startups;

3.    Continuous build-up of the AI tech stack;

4.    Vertical integration;­­­

5.    Limited popularity of the marketplace model. 

2018 witnessed the lowest number of new AI startups launched since 2010, according to Pitchbook (Chart 1). Even given the fact that not all startups launched in 2018 could be included in the database[1], the trend is likely there and we are seeing a slowdown of AI companies formation. 

Partially, this slump may be explained by the fact that startups are now less motivated to pitch their AI-ness and therefore are being less likely spotted by data providers. However, a more careful search reveals those startups who “hide”. For example, one may spot more AI startups via searching for multiple keywords associated with AI, rather than by relying on a static classification applied to a startup by a data provider (Chart 2).   

Nevertheless, even adjusted by those AI startups which do not  advertise their AI component, the number of newly formed AI startups is significantly below the level seen in previous years. 

A slowdown in AI startups launches is mirrored by investment activity. Previously, investors were spreading their bets signing smaller checks to multiple startups. Nowadays, they seem to double down on supposed winners and invest in newcomers in a more focused manner. For example, out of a sample of 121 AI startups that launched in 2018[2], 7% (or 8 companies) raised first rounds of $10M or more, while only 3% of 801 startups launched in 2016 raised large initial rounds. 

Even if large initial rounds became more frequent, the median size of the AI seed rounds is just slightly higher than the US average, $2.3M vs. $2M respectively. The majority of sampled newcomers managed to raise only once during 2018, with just 9 startups raised several times each (Chart 3).

Brian W. Arthur once noted, that ‘…needs [for technologies] themselves derive more from technology itself than directly from human wants; they derive in the main from limitations encountered and problems engendered by technologies themselves’[3].

The past year demonstrated how AI, robotics, blockchain and other technologies have created demand for supporting technologies. Fore example, emergence of self-driving cars opened an opportunity for Ottopia to build a platform to remotely assist them. Proliferation of telemedicine demanded a new type of recruitment companies like Enzyme, that are able to connect clinicians with telemedicine providers. A wave in robotic process automation (RPA) increased demand for tools to describe and map business processes like what Mimica develops. 

On a separate note, I wanted to highlight a trend of platformisation. It was widely noticed in banking and now it seems to approach the legal industry. For example, Reynen Court whose ‘…application accelerates inter-operability between and among legal technology applications, enabling law firms adopt artificial intelligence and smart contracts…’. Opportunities opened by smart contracts and AI push law firms towards becoming tech companies/platforms. 

Overall, out of 48 sampled AI startups launched in 2018, 15 were more or less clearly associated with the needs of new technologies/sectors (Chart 4, and a Google sheet for a larger version ). 

DataOps and DevOps functions are the most popular to tackle by listed AI startups. Twelve startups out of 48 work in these fields: 

15 out of selected 47 startup demonstrate various aspects of vertical integration, ‘… an arrangement in which the supply chain of a company is owned by that company’. AI strartups usually do not directly own the supply chain, but control several elements of it, and offer a holistic service, rather than pure software. Several ways of ‘owning the supply chain’ may be identified:

Marketplaces are not very popular among AI founders who launched their ventures in 2018. Just 8 companies took this approach. For example, Enzyme ‘…uses machine learning algorithms to match doctors and nurse practitioners with positions that suit their expertise and availability’, while Lumen App uses an algorithm to verify selfies, uploaded by members of its dating application, designed for people over-50s age group. 

1)   We might be at the point where pitching/highlighting AI-ness of a startup does not make much sense. Partially, because it was compromised by false claims. Another reason is that differentiation based on value of a product, rather than its technology may be a better selling strategy;

2)   Fundraising for an AI startup may become harder, as investors have already made their bets. Now they double downon them and cautiously invest in new ones;

3)   AI, as other technologies, creates new problems, not only solve them. Therefore, spotting these problems early on may point to new market opportunities;

4)   Cognitive tech stack is a good example, when one wave of tech generates demand for supporting technologies, that help to solve challenges associated with the first wave;

5)   Vertical integration appears to be popular among AI startups, therefore when founding a company, entrepreneurs should be prepared to go outside of their comfort zone. The team will likely engage not only in research and software development, but in hardware engineering and large operation activities, e.g. managing a group of call center operators. 

[1] There is a lag between the launch of a startup and its appearance in the database. Chart 1 shows the difference between startups identified by Pitchbook as queries from January and March 2018. The further back we go, fewer startups remain uncaptured by the database;

[2] Here and for 2016 dataset, only AI startups with known amounts of capital raised were included in the sample (i.e. not all AI startups that were launched in these years);

[3]Brian Arthur, W. The Nature of Technology: What It Is and How It Evolves (p. 204). Penguin Books Ltd. Kindle Edition;

[4] Google sheet of Chart 4. Sampled AI startups launched in 2018 and raised $1M or more each.