Condensed look at the European health Core AI market

What speaks against an investment in the European health core AI market and personalized cancer immunotherapy

In this post, I looked at what speaks against an investment in the European health core AI market and personalized cancer immunotherapy. Please note that I have not accounted for any trends (like, for example, “data-less” ML based on cognitive neuroscience) or whether AI is truly here to stay (or if we are headed into another AI winter).

Why investing in European personalized cancer immunotherapy startups might not make sense

Based on a condensed view of European health Core AI startups (see Appendix) I thought that personalized cancer immunotherapy would be a very promising category for VCs because it addresses a very large problem that is highly valuable for multiple stakeholders and very urgent, especially for patients (see Hunter Walk’s Why I Care About Problem Size More Than Market Size).

However, from a non-medical POV, I would not invest because the application of AI in immunotherapy still looks experimental and hasn’t been proved to work. Also, there are several other cancer treatment methods that could be supported by AI and should, therefore, be compared to AI in immunotherapy. This comparison should look at the application’s suitability, the level of “AI-ness”, availability of scientific evidence, maturity of research and ecosystem, difficulty of reaching ecosystem buy-in and flexibility of scope.

  • The application’s suitability: The question is which kind of implementation would make the most sense whereas these can be separated by type of improvement (making research easier, new research methods or directly healing patients) or health stage (prevention, healing …).
  • Level of “AI-ness” in used technology: It should be determined how much AI lies within the startups’ technology to see, for example, whether the company is applying real AI or “just” data/business analytics marketed as AI.
  • Availability of scientific evidence: VCs might want to find out whether there is (scientific) evidence proofing that AI can indeed help in the chosen field.
  • Maturity of research: Finding out how progressed the startups’ scientific work is, can help investors determine whether they are funding a research project or a company and if that is aligned with their expectations.
  • Maturity of ecosystem: It is important to evaluate how much resources the startups have at their disposal because several factors surrounding them (such as availability of knowledge or funding) can “make or break” the market. See below for a more detailed discussion on ecosystem requirements.
  • Difficulty of reaching ecosystem buy-in: This analysis, closely related to the previous one, should evaluate how much which stakeholders (governments, other organizations along the value chain, patients…) in the ecosystem will exercise resistance on the startup and whether and how that influence can be managed. For example, would AI-based cancer research cause resistance from research labs because applying AI in that area could eliminate certain jobs?
  • Flexibility of scope: It should be determined whether the company could pivot by reusing its resources (data, algorithms, knowledge…) in case the chosen solution turns out to be less effective than alternatives (or totally ineffective). This flexibility-issue relates to the notion of whether the company should be product- (focus on one type of therapy) or customer-oriented (focus on healing cancer patients)?

What, however, speaks for personalized cancer immunotherapy, is that its parent category, European health Core AI, makes sense in the long-term. In the long run, the issues that I have listed below will change provided that AI is indeed capable of solving a large, valuable and urgent problem at an “acceptable” price. These issues include hurdles from market structure, rejection by stakeholders and lacking financial and operational resources in the ecosystem.

Why European health Core AI might not make sense in the short-term

The combination of a moderately startup-friendly market, stakeholder-rejection as well as a missing ecosystem make the market challenging.

Moderately startup-friendly market makes entering easy but long-term sustaining difficult

Although startups would be able to carve out a market position and defend it against direct competitors due to the absence of first mover advantages, and the presence of high-level segmentation, the indirect competition through forward integration and adjacent segments, however, results in a only moderately startup-friendly market.

  • Absence of first-mover advantages allows also late entrants to succeed: When using funding as a proxy for success there is no reason to believe that older companies are in an advantage (see table below) although one could assume that, for example, longer training time and larger data sets for algorithms result in more appealing companies for investors.
Correlation between age and funding across three AI segments shows that age is not an indicator for funding volume (see “Fundamentals for Core, Healthcare and BI & Analytics AI startups” in the Appendix)
  • Due to high-level segmentation the market is easy to break into: My overview of the European health Core AI startups (see Appendix) shows that these companies differentiate themselves on a very high level, like for example, type of product, but not necessarily performance. Nevertheless, it must not be forgotten, that performance plays an important role and will become the center of differentiation when the market gets more crowded.
  • Competitive threat through forward integration: Depending on how much impact AI will have, the startups’ B2B customers might fear becoming data suppliers and therefore start developing their own AI technology. Whereas this is similar to competition within their own segment, startups must not overlook that threat by concentrating too much on their own industry.
  • Threat of cannibalization from adjacent segments: In the segments in which core AI startups operate (e. g. health or finance) they also compete with industry-specific startups which might now or in the future offer similar products. Here, the same caution applies as above; startups must not overlook that threat by concentrating to much on their own industry.

Stakeholder-rejection could slow down adoption of AI

I further consider the segment risky because enablers (bearers of healthcare costs and governments), as well as customers and end consumers, might simply be against the new technology.

Rejection through enablers
Possible reasons for rejection are increased healthcare costs as well as regulatory hurdles.

  • Increased healthcare costs might scare off insurers: Adoption rate of AI in healthcare depends upon the raising in healthcare costs and the degree to which insurers (governments or insurances) will be able to pass that increase onto consumers. This accounts only, of course, if AI will raise healthcare costs.
  • Regulatory hurdles might derail adoption of AI: Adoption speed of AI in healthcare also depends on how many regulations startups will have to overcome before they can offer their technology.

Rejection through customers and end consumers
Patients, as well as doctors, might reject the technology because they are technology-avers. Further opposition might stem from increased healthcare costs and stakeholders fearing job loss.

  • Rejection due to technology adversity in B2C and B2B: As in many other industries, the likelihood of consumers (B2c) accepting a technology is a measure of potential success. Depending on the new technology’s severity of change, transparency and ease of understanding I believe that patients’ technology adversity could be higher than in other industries (admittingly, the adoption of fitness trackers and similar is a solid counter-argument to my assumption). The case is even more challenging in the B2B environment where stakeholders (doctors, insurances…) might be unwilling to cross the chasm by testing a new, possibly experimental technology that might not only yield no output but also cost them time.
  • Rejection due to increased healthcare costs: Similarly to the above-mentioned price-issue in regards to insurers (“Increased healthcare costs might scare off insurers”) I believe that increased healthcare costs resulting through AI might lead to rejection by consumers. As above, this accounts only, of course, if AI will raise costs.
  • Rejection due to fear of job security: Considering that AI might result in lay-offs, startups might run into significant adoption hurdles stemming from stakeholders fearing job loss. (Please note that this assumption contradicts my assumption of increased healthcare costs)

Lacking financial and operational resources in the ecosystem might hinder startups building their businesses

Startups might not be able to grow quickly enough due to the absence of skilled workforce and VCs

  • Lack of workforce: Considering that the intersection of AI and medicine demands for some very skilled workers I am convinced that the absence of suitable employees could hinder the proliferation of AI in healthcare.
  • Lack of knowledge in VCs: Similarly to above, I believe that it is further safe to assume that VCs might as well lack the adequate knowledge to support startups financially as well as operationally.


Overview of data I have used for my decisions

European Core AI Healthcare AI startups

Source: Top 25 companies from AngelList based on the filters Europe, Artificial Intelligence and health care

American Core AI Healthcare startups

Source: Core AI startups from CB Insights’ AI 100

Fundamentals for Core, Healthcare and BI & Analytics AI startups

Fundamentals for Core AI startups

Age and total equity funding for Core AI startups based on AI 100 from CB Insights

Fundamentals for Healthcare AI startups

Age and total equity funding for Healthcare AI startups based on AI 100 from CB Insights

Fundamentals for BI & Analytics AI startups

Age and total equity funding for BI & Analytics AI startups based on AI 100 from CB Insights

More posts like this one here: