Artificial Intelligence has been on my investor’s radar for a long time. I’ve invested in AI startups for many years and helped build successful companies that use the technology, such as Reface and Zibra AI.
Lately, much of the AI talk has been about China's rapid advancements — especially DeepSeek’s cheap, open-source model. But while some worry about these shifts, the reality is that AI investment isn't slowing down. Just weeks ago, Meta announced a $60 billion AI infrastructure push for 2025.
With global AI competition accelerating, European investors are right to rethink their approach. But rather than overhauling the investment playbook entirely, the key is adapting to new realities — without losing sight of what has worked.
Don’t buy the idea that European startups are falling behind
European AI is different — and that’s its strong side.
While the US and China lead AI research and development, European startups are better at finding applications for specific verticals.
This might be a more sustainable strategy in the long run because practical solutions have the potential to bring value right now. While AI research is essential, it’s the applications that will drive the market in the future.
To put it simply, while OpenAI is raising billions, it focuses mostly on theoretical domains. Many European startups are already finding applications for the models — and they are on the right path.
Embrace the fact that no one knows what will happen with AI
Not long ago, the media was swarming with news about how DeepSeek would burst the AI bubble. Weeks later, and ChatGPT is still around. OpenAI is even in talks to raise $40 billion.
In young sectors like AI, change is the only constant.
For us investors, it means that disruptive developments can happen anywhere and at any time. Accept this as a hard truth and remember it when investing in new AI startups.
Expect to see more open-source AI
With Chinese companies' open-sourcing models that rival proprietary AI from the US, it’s natural to expect American companies to follow suit with their own open-source versions.
Open-source models are important because startups that use them can be more cost-efficient, which makes them more attractive at early stages.
Adjust the way you evaluate your AI investments
Whatever framework you use to judge potential investments, fine-tuning it in response to AI advancements won’t hurt.
Here are two things to pay attention to when considering investing in an AI startup:
First, look at its value
According to Randy Bean's 2025 AI survey, organizational culture is still the primary reason organizations struggle to embrace AI.
With slow adoption, startups you invest in must bring enough value to the end user to overcome corporate inertia.
When talking to founders, make it your mission to determine if their product solves the client's problem and if the client will actually use their AI.
Questions to ask during the discovery call to assess the startup’s value:
1. Do you conduct customer interviews or rely on second-hand research?
2. How do you use feedback from real users in your product development?
Second, assess its efficiency
DeepSeek showed that creating an advanced AI model is possible even with limited resources. I expect that this precedent might lead to a general push for “more doesn’t equal better.”
Even if a startup’s AI has enough value, it still needs an efficient business model that will propel it to being self-sustainable.
Pay attention to startups that use efficient AI architecture and manage their funds efficiently instead of just throwing money and resources at a problem.
Price-to-performance ratio is very important in this context. It measures the balance between the cost of an AI system (hardware and software) and its performance (speed, accuracy, etc.).
I also expect that as more open-source AI models appear, many developers will decide to leave proprietary models behind and instead use open-source AI or their own models.
Some discovery call questions for checking the startup’s approach to efficiency:
1. Do you use open-source or proprietary models in your AI? Why? Have you considered other options?
2. What’s your estimated price-to-performance ratio? How are you planning to improve it?
The bottom line
After the appearance of DeepSeek and other cheap and open-source models from China, it’s tempting to say we should rethink how we invest in AI startups.
I partially agree, but it’s more about adapting our approach, as I described.
The most important thing to remember is that no one knows what will happen in the future. Disruptive developments can (and probably will) appear where and when no one expects them.
When considering investing in an AI startup, pay attention to its value and efficiency. Make sure to ask the founders if they talk to clients and how they plan to optimize their price-to-performance ratio.
And don’t buy into the idea that European startups are falling behind — they’re simply taking a different approach, focusing on specific vertical applications. In the long run, this might turn out to be a more sustainable strategy.
Good luck with your investments.
