AMI Labs
Category: AI Infrastructure
A frontier AI research company developing world models based on Joint Embedding Predictive Architecture (JEPA) that enable intelligent systems to understand the real world through continuous, high-dimensional sensor data, rather than just predicting text tokens. AMI Labs was founded in 2025. The company is led by Alex LeBrun. Based in Paris, France. Team size: 10-50. Total funding raised: €500.0M. Latest round: Series A (€500.0M, Dec 2025). Key investors include ["Cathay Innovation","Greycroft","Hiro Capital","Bpifrance"].
- Founded
- 2025
- Headquarters
- Paris, France
- Team size
- 10-50
- Total funding
- €500.0M
Value proposition
Eliminates LLM hallucinations by learning abstract representations of real-world data and predicting in latent space, enabling action-conditioned world models that can plan sequences, reason about consequences, and operate safely in physical environments—critical for healthcare, manufacturing, and robotics where reliability and controllability are paramount.
Products and solutions
["JEPA (Joint Embedding Predictive Architecture) - core learning framework","I-JEPA (Image JEPA) - foundation model for image understanding","V-JEPA / V-JEPA 2 (Video JEPA) - foundation model for video understanding and physical reasoning","Action-Conditioned World Models for autonomous planning","Open-source world model infrastructure and research publications"]
Unique value
Pioneered by Turing Award winner Yann LeCun as a contrarian bet against LLM-centric AI. World models learn by predicting representations of missing future information in abstract latent space rather than generating pixels or tokens—enabling systems that observe and interact with physical environments like humans do, with persistent memory and action consequence prediction.
Target customer
Healthcare technology companies, robotics manufacturers, industrial process control and automation firms, wearable device makers, and enterprises requiring reliable, auditable AI systems in high-stakes environments
Industries served
["Healthcare & Medical AI","Robotics & Autonomous Systems","Industrial Process Control & Automation","Manufacturing","Wearable Devices","Enterprise AI Safety & Reliability"]
Technology advantage
Combines 12+ years of LeCun's research at Meta (I-JEPA, V-JEPA architectures) with a world-class founding team including Alex LeBrun (3x AI unicorn founder: Wit.ai→Facebook, VirtuOz→Nuance, Nabla). The approach avoids the computational inefficiency and hallucinations of generative models by predicting in representation space. First-mover advantage in the emerging 'world model' race alongside competitors like Fei-Fei Li's World Labs, DeepMind's Genie, and NVIDIA Cosmos.
How they differentiate
JEPA (Joint Embedding Predictive Architecture) approach predicts in latent representation space rather than generating pixels or tokens, eliminating hallucinations. Focuses on deterministic, auditable AI systems for high-stakes applications with exclusive healthcare partnership and 12+ years of LeCun's proprietary research at Meta.
Main competitors
["World Labs (Fei-Fei Li)","Google DeepMind (Genie 3)","Verses.ai (AXIOM)"]
Key partnerships
["Nabla (exclusive strategic partnership) - privileged access to world models for FDA-certifiable healthcare AI, with Alex LeBrun serving as CEO of both entities during transition","Global academic research community - commitment to open publications and open-source contributions","Industry partners for applications in manufacturing, automation, and wearables (recruiting via amilabs.xyz)","Multi-continent research operations - Paris HQ with offices in New York, Montreal, and Singapore"]
Notable customers
["Nabla (exclusive partnership for FDA-certifiable healthcare AI)"]
Major milestones
["Company founded and announced in December 2025","Exclusive strategic partnership with Nabla for first access to world model technologies","Alex LeBrun appointed CEO (former Nabla CEO with 3 successful AI exits)","Recruiting team from OpenAI, Google DeepMind, and xAI","Targeting FDA-certifiable agentic AI systems for healthcare"]
Growth metrics
Pre-revenue R&D phase with €500M target at €3-3.5B valuation; exclusive strategic partnership secured with Nabla for healthcare applications
Market positioning
Frontier AI research challenger positioned as contrarian bet against LLM-centric approaches, targeting enterprise healthcare, robotics, and industrial automation with world models that can plan, reason about consequences, and operate safely in physical environments.
Geographic focus
Global with Paris headquarters and offices in New York, Montreal, and Singapore
Patents and IP
JEPA architecture was developed by LeCun at Meta with open research publications. No specific AMI Labs patents publicly disclosed yet, but proprietary implementations and applications in healthcare, robotics, and industrial control are being developed. V-JEPA 2 (1.2B parameter model) published by Meta in 2025 demonstrates the foundational technology.
About Alex LeBrun
Alex LeBrun is a serial AI entrepreneur with three successful exits. He co-founded and served as CEO of Wit.ai (acquired by Facebook in 2015) and VirtuOz (acquired by Nuance Communications in 2012). Most recently, he co-founded and led Nabla, a leading healthcare AI company supporting over 85,000 physicians. Between his ventures, he served as Head of Engineering at Facebook AI Research (FAIR) in Paris from 2015 to 2018, where he worked directly under Yann LeCun.
Official website: https://amilabs.xyz