At Rise of AI 2019 in Berlin, Christian Nagel, Partner at Earlybird Venture Capital, argues that Europe is more competitive in AI than commonly assumed and lays out how his firm evaluates AI startups. He notes Europe produced 17 unicorns against China's 20 in the prior year, with UiPath, a Romanian company, becoming Europe's most valuable AI company and the fastest-growing enterprise software company ever, reaching an almost 400 million dollar run rate and a 7 billion dollar valuation. He says roughly a third of Earlybird's portfolio and deal flow involves AI, though the firm avoids investing on AI hype alone. He distinguishes descriptive, predictive and increasingly autonomous or self-driving AI applications, and separates vertical approaches (deep in one industry, citing a payment-fraud detection company) from horizontal ones (UiPath's robotic process automation across industries). His central investment criterion is data-driven lock-in: whether a company's algorithm improves enough with more data and customers to create a forward-looking advantage competitors cannot catch. He closes on the resulting chicken-and-egg problem of needing customers to get data and data to get customers.
Key takeaways
- Nagel cites 17 European AI unicorns against 20 in China the prior year, with far less capital invested (22 billion versus 60-70 billion).
- UiPath, from Romania, reached a 7 billion dollar valuation and a nearly 400 million dollar run rate, calling it Europe's fastest-growing enterprise software company.
- Roughly a third of Earlybird's portfolio genuinely uses AI to differentiate its product, though the firm invests in business models, not AI as such.
- Nagel's core investment criterion is a forward-looking, data-driven lock-in effect where more data and customers reinforce each other.
- He estimates European AI investment lags the US and China by a factor of five to eight.
Detailed notes
Europe's AI position relative to China
Nagel opens by countering fear of China, citing 20 Chinese unicorns against 17 European ones the prior year, despite China raising 60 to 70 billion dollars in investment against Europe's 22 billion. He highlights UiPath, a Romanian company, as Europe's most valuable AI company and the fastest-growing enterprise software company ever, reaching close to a 400 million dollar run rate and a 7 billion dollar valuation, up from almost 3 billion just six months earlier.
How much of Earlybird's portfolio is really AI
Nagel says Earlybird does not invest in AI or blockchain as such, but rather in business models; still, 13 to 20 of its portfolio companies, almost a third, genuinely use AI to differentiate their product, and about a third of deal flow claims AI. He says nearly every pitch now claims AI or blockchain, making it hard to distinguish real AI from hype, since the same system can look impressive to a founder and unconvincing to an AI researcher.
Descriptive, predictive and autonomous AI
Nagel frames AI maturity in stages: early descriptive systems (no longer interesting on their own), predictive systems that anticipate what might happen (where much of Earlybird's portfolio sits), and systems that make suggestions or act autonomously, which he sees as the more interesting next step, comparing it to self-driving cars, which work but are not yet at mass adoption.
Vertical versus horizontal AI strategies
Nagel distinguishes a vertical approach, going deep in one industry, citing a portfolio company applying AI narrowly to payment-fraud detection with global, industry-disrupting impact, from a horizontal approach applicable across industries, exemplified by UiPath's robotic process automation, which the company argues every desktop will eventually need. He says both can produce unicorns and Earlybird has no preference between them.
Data lock-in and the chicken-and-egg problem
Nagel says his key investment criterion is whether a company can build a data-driven lock-in effect: as an engine trains on more data, it improves, attracting more customers and more data in an upward spiral competitors cannot easily catch, creating a forward-looking lock-in beyond simple customer retention. The obstacle is a chicken-and-egg problem: more data requires more customers, and more customers require enough data to be useful. He says startups can overcome this by finding ways to use customer data even when it must stay on-premise, for example training shared models without moving raw data, and cites a portfolio company combining AI and blockchain to let customers share data on their own terms. He looks for early evidence that the engine already works, even on small data, and for teams with genuine AI expertise built in from day one, rather than rule-based systems presented as AI.
Q&A highlights
Asked whether Europe can compete in the AI infrastructure stack (chips, cloud, APIs), Nagel says Earlybird avoids that layer since each hardware generation is quickly superseded. On Europe's broader competitiveness, he says European investment lags the US and China by a factor of five to eight and needs more capital flowing into the ecosystem, though quality companies are already emerging. Asked about investing in general-purpose intelligence, he says it depends on how advanced the underlying technology is, since a horizontal approach requires a genuine technical edge. Asked how Earlybird itself uses AI in sourcing or evaluating deals, he says the firm is still deciding whether to build a model for pattern recognition in investment decisions, beyond its current module-based system.
Quotes
You can't eat AI, and we rather look for business models.
Christian Nagel
No one comes through our door without saying I have AI or blockchain, this is just something everybody has.
Christian Nagel
You can create these locking effects, which we as a VC like very much.
Christian Nagel
Questions answered in this talk
How does Earlybird Venture Capital decide if a startup's AI is real?
According to Christian Nagel at Rise of AI 2019, Earlybird looks for evidence the AI engine already works, even on small data, and for founding teams with genuine AI expertise from day one, rather than rule-based systems presented as AI.
What is the difference between a vertical and a horizontal AI startup strategy?
Christian Nagel at Rise of AI 2019 defines a vertical approach as going deep in a single industry (such as payment-fraud detection), and a horizontal approach as applying one AI capability across many industries, citing UiPath's robotic process automation as an example.
How does Europe's AI investment compare to the US and China?
Christian Nagel said at Rise of AI 2019 that European AI investment lags the US and China by a factor of five to eight, but that Europe can still produce competitive AI unicorns, citing UiPath and comparable unicorn counts to China.