Our verdict
ML6 has built AI systems for European enterprises since 2013, reports more than 400 delivered projects, and became one of OpenAI’s European services partners in June 2025 (ML6 history; ML6). It suits a European company or public body that wants an experienced engineering team to take an AI use case from strategy to a production system under EU rules, and it is less suited to a US buyer or a startup that wants a cheap first prototype.
The right buyer
- Public bodies and regulated organizations in the Benelux and Germany. For the Dutch National Cyber Security Centre it built an advisory generator and internal assistants, set up an AI governance framework for the EU AI Act, and cut the cost of each new assistant by 85% with reusable components (NCSC case).
- Media, energy, and industrial companies with large data sets. For Mediahuis it built NewsSearch, a Gemini-based archive assistant that went from idea to live newsroom tool in six months (Mediahuis case).
- Buyers who want the model choice left open. Its partners include Google Cloud, AWS, Microsoft, OpenAI, Anthropic, Mistral, NVIDIA, and Hugging Face (homepage).
Reasons to look elsewhere
- Its offices and most public cases are in Belgium, the Netherlands, and Germany, so US buyers will find few local references.
- It is built for enterprise programs, so a small company that wants a low-cost prototype may find it more than it needs.
ML6, At A High Level
ML6 started in Ghent in 2013, was an early Google machine learning launch partner, and later added partnerships with AWS, Azure, and OpenAI; it now has offices in Ghent, Amsterdam, Eindhoven, Berlin, and Munich (ML6 history). Nicolas Deruytter founded the firm and is its chief executive, and Matthias Feys, its co-founder, is chief technology officer (leadership page).
- Headquartered at Esplanade Oscar Van de Voorde 1 in Ghent, Belgium, through its operating company Skyhaus BV (privacy notice).
- Skyhaus BV reported revenue of €16.55 million (about $18 million) and a profit of €1.52 million for 2025, up 32% from 2024, with 60.9 full-time staff on its Belgian payroll; ML6 counts more than 100 AI experts across its offices (filed accounts; ML6 history).
- More than 150 clients and 400 projects by its own count, with an average time to value of 11 weeks, and its engineers spend 17% of their time on internal research (ML6 history).
- Clients across retail, manufacturing, energy, finance, and the public sector include Walmart, BASF, ING, AXA, Pfizer, and Transport for London (ML6).
Scores by criterion
We scored ML6 from its history, leadership, and case pages, its OpenAI partner announcement, and the filed 2025 accounts of its operating company. Production track record and technical depth score highest; commercial clarity scores lowest because it publishes no terms.
| Criterion | Weight | Score | Basis |
|---|---|---|---|
| Production track record | 20% | 8.0 | More than 400 projects since 2013; detailed public cases for the NCSC, Mediahuis, and Scout24. |
| Technical depth | 15% | 8.0 | Machine learning, generative AI, agents, MLOps, and vision and voice AI. |
| Industry depth | 10% | 7.0 | Public sector, media, energy, manufacturing, and retail cases. |
| Delivery model and handoff | 15% | 7.0 | Strategy, build, governance, and training, with clients keeping ownership of products. |
| Commercial clarity | 10% | 5.5 | No published rates or minimums. |
| Platform independence | 10% | 7.0 | Partners with Google, AWS, Microsoft, OpenAI, Anthropic, and Mistral. |
| Stability and scale | 10% | 6.5 | Founded 2013; Belgian entity revenue of about $18 million in 2025. |
| Independent evidence | 10% | 6.5 | FT1000 listing and four Deloitte Fast 50 awards; attributed client quotes. |
| Overall | 100% | 7.1 | Weighted average |
Where ML6 fits
ML6 is used by large European companies and public bodies that want AI built into their operations, from choosing use cases to running the systems and training staff. On price it is a senior European engineering firm, below the global consultancies and above nearshore development shops. Competitors on this site include deepsense.ai and Unit8 for European AI engineering, AE for Belgian IT and AI programs, and Fuzzy Labs for MLOps.
| Need | Fit | Why |
|---|---|---|
| A generative AI assistant on company data | Strong | It built NewsSearch for Mediahuis and internal assistants for the NCSC. |
| AI governance under the EU AI Act | Strong | It set up the NCSC’s AI governance framework and risk process. |
| Classic machine learning and forecasting | Strong | Its energy cases cover anomaly detection and shutdown risk prediction. |
| A low-cost prototype for a US startup | Weak | Its teams and references are in Europe and it works at enterprise scale. |
What clients and leaders say
ML6’s leaders and clients describe its work in published announcements. One buyer concern is written from public facts about the firm.
“It was a real pleasure having a party like ML6 guiding us with how AI can be leveraged in a secure, compliant and trustworthy way and enhance our operations to provide better service to our stakeholders.”
“The time it takes to find information has been drastically reduced: what sometimes took hours is now done in seconds or minutes.”
“This validates our deep technical expertise and unwavering commitment to building ethical and impactful AI solutions for our clients.”
“The engineering was strong and they knew the EU rules, but we had to ask for rates early because none are published, and most of their case work is under NDA.”
Where customer reviews couldn’t be found, our analysts attempted to give a voice to buyer sentiment, drawing on real information from review sites, forums and product documentation. Based on: ML6’s cases page, which says most of its cases are under NDA, and its NCSC case.
Who leads ML6
Nicolas Deruytter founded ML6 in 2013 and is its chief executive (ML6 history).
