Skip to main content

kausable

Category: Foundation Models / LLMs

kausable is a German deep-tech AI startup developing reasoning-first frontier AI models that learn from minimal data and adapt to new situations without requiring retraining, using causal world models trained on synthetic data. kausable was founded in 2025. The company is led by Johannes Haux. Based in Heidelberg, Germany. Team size: 1-10. Total funding raised: $14.9M. Latest round: Seed. Key investors include UVC Partners, Entourage, HTGF (High-Tech Gründerfonds), Mätch VC.

Founded
2025
Headquarters
Heidelberg, Germany
Team size
1-10
Total funding
$14.9M

Value proposition

Building reasoning-first frontier AI that adapts to changing contexts without costly retraining — a causal world model that learns like humans from just a few examples, trained on synthetic data, enabling adaptation across robotics, energy, finance, and healthcare domains.

Products and solutions

TipPFN (zero-shot forecasting model for complex dynamic systems predicting tipping points and black swan events), Causal world model (reasoning-first foundational AI that learns from minimal data and adapts without retraining)

Unique value

Reasoning-first causal world model that requires no retraining — AI that adapts to new situations from a handful of examples, trained on synthetic causal data rather than massive real-world datasets, enabling sovereign European frontier AI.

Target customer

Enterprises in highly dynamic domains including robotics, energy sector, finance, healthcare, and industrial forecasting — organizations with complex systems that are difficult to predict and control

Industries served

Robotics, Energy, Finance, Healthcare, Industrial forecasting

Technology advantage

Reasoning-first causal world model trained on synthetic data using Bayesian learning; zero-shot transfer across domains; no retraining required; TipPFN model for predicting rare high-impact events; co-authored research with Columbia University validating causal reasoning architecture; founded by three physicists with ties to Heidelberg University and Black Forest Labs

How they differentiate

Unlike conventional AI models that require massive datasets and constant retraining (scaling-law approach), kausable builds reasoning-first causal world models that learn from minimal examples and adapt without retraining. Trained on synthetic causal data rather than real-world data, enabling privacy, control, and cross-domain transfer. Challenges the scaling-law orthodoxy with a Bayesian learning approach.

Main competitors

Sakana AI (Japan - also building alternative AI paradigms), Aleph Alpha (Germany - European sovereign AI), Mistral AI (France - European frontier LLMs)

Key partnerships

Heidelberg University (research roots), Black Forest Labs / BFL (alumni ties), Columbia University (co-authored research paper validating causal reasoning architecture), ELLIS (European Laboratory for Learning and Intelligent Systems)

Major milestones

2024: Initial idea development, 2025: Company incorporation, raised €1.5M pre-seed, 2025: Developed TipPFN zero-shot forecasting model, 2026: Co-authored research paper with Columbia University, 2026-07: Raised €12M seed round led by UVC Partners and Entourage

Market positioning

European sovereign frontier AI lab positioned as an alternative to US/China-dominated AI development. Competes with other frontier AI labs but differentiates on reasoning-first causal approach vs. scaling-law paradigm. Backed by top European VCs as a strategic digital sovereignty play.

Geographic focus

Europe (with emphasis on German/French/EU digital sovereignty)

About Johannes Haux

Ex-researcher at Heidelberg University Computer Vision Group (Prof. Björn Ommer); Co-authored computer vision papers (Unsupervised Robust Disentangling of Latent Characteristics for Image Synthesis); background in deep learning and computer vision; physicist by training

Latest news about kausable

More Foundation Models / LLMs companies

Official website: