Sunny Ray talks with Leo Benkel, CTO of The Artificial Business, calling in from Luxembourg. Benkel explains why his company builds local, sovereign, on-prem AI for enterprises that must follow regulations, especially in European finance, rather than sending customer data to foreign data centers. After ten years in Silicon Valley, he moved back to Europe to be closer to family in France and to give back. He describes how requests from Luxembourg business audiences after ChatGPT launched led him to act. Benkel argues the AI model is becoming a commodity, so his team focuses on the harness installed next to a company's own data. He covers the self-funded, fixed-license revenue model, a roughly 5K Dell GB10 that can serve 5 to 15 concurrent users, and why Luxembourg is well placed for AI given its startup investment, supercomputer, and labor gaps. He also shares his hope that Europe produces its own AI tech giants.
A Luxembourg CTO explains why bringing intelligence to your data, not data to the cloud, is the future of enterprise AI.
What are you building at The Artificial Business, and what problem are you solving?
We solve the problem of AI being hosted only in the US. We make it local, sovereign and on-prem for enterprises that must follow regulations, especially in Europe and in finance. You don't want to send your customer data and secrets all the way to foreign data centers.
Why did you move from Silicon Valley back to Europe, and what did you take with you?
I wanted to be closer to France, where I'm from and where my family is, and I wanted to give back to Europe after ten years in Silicon Valley. From America I kept the straight-to-the-point way of doing business and the ambition of going as far as possible.
When did you decide AI had to run inside the company rather than in someone else's cloud?
The Chamber of Commerce of Luxembourg asked me to give classes and keynotes when ChatGPT came out. After the talks, people kept telling me they didn't feel safe sending their data there. Hearing it once is one thing, but after dozens and dozens of times, I knew I had to do something.
Can you describe The Artificial Business in one sentence?
We bring intelligence to the data rather than sending the data to the intelligence.
You separate the harness from the AI. Which one do you sell, and why?
We focus on the harness. The AI model is becoming a commodity, with many open source models available, and you can switch them with little visible difference. The real difference is what the harness can do in your company. We install it on your infrastructure, next to where your data is.
What does your revenue model look like when the customer owns the server?
We charge a fixed license for using the platform, including maintenance, support and new versions. Compute and GPU costs are the customer's, so we don't price on that. We then build services on top, such as new API connections or paid prioritization of roadmap features.
At what point is a company too small for on-prem AI to make sense?
The bar to entry is very small. I have a Dell GB10 here that costs about 5K, and it runs pretty good AI for 5 to 15 concurrent users, no problem. The real obstacles are hardware supply, which took three months, and having a proper harness with permissions connected to everything.
What happens if a cloud provider offers sovereign regions and makes your pitch obsolete?
I wouldn't be surprised how they would pull that off. On-prem would mean selling hardware to customers and building a pipeline that deploys at every single customer's infrastructure. It's a whole different business, and they're not ready for it.
Building something daring? Sunny talks to founders like this every day. Fifteen minutes to see if your story belongs on the stage.
Claim your pre-interviewSend a founder here and we will know it was you.