Habana Labs
Habana Labs is an Israeli AI chip startup that developed purpose-built deep learning processors for training and inference workloads in data centers.
Visit WebsiteCompany Overview
Habana Labs developed two key AI processor families: Gaudi for training and Goya for inference, both designed to deliver high-throughput deep learning compute in data center environments. The company emphasized programmability, power efficiency, and integration with standard deep learning frameworks.
Founded in 2016 in Tel Aviv by David Dahan (CEO) and former Primesense team members, Habana Labs raised approximately $120M in venture funding from Bessemer Venture Partners, Battery Ventures, and WRV Capital. Intel acquired Habana Labs in December 2019 for approximately $2 billion, making it one of the largest Israeli AI acquisitions.
Dual-use relevance is strong: purpose-built AI processors are foundational to defense and intelligence applications requiring large-scale model training, real-time inference for ISR analytics, and autonomous systems AI compute in secure data center environments.
Dual-Use Assessment
Purpose-built AI training and inference processors are dual-use for commercial cloud AI workloads and defense/intelligence applications including ISR analytics, autonomous systems AI, and secure data center AI compute.
Key Technologies
- Purpose-built AI training processor architecture (Gaudi)
- Dedicated AI inference processor (Goya)
- High-bandwidth memory integration for deep learning
- Standard deep learning framework support (TensorFlow, PyTorch)
- RoCE-based scale-out interconnect for distributed training
Use Cases & Applications
- Data center AI model training at scale
- Real-time deep learning inference for cloud services
- Defense/intelligence ISR image and signal analytics
- Autonomous systems AI compute for model training and deployment
- Secure AI processing in classified data center environments
Strategic Value to U.S.-Israel Alliance
AI processor technology is critical for defense AI infrastructure, enabling large-scale model training and inference for intelligence, surveillance, and autonomous systems.
Interested in this startup?
Learn more about our investment approach or get in touch to discuss opportunities in dual-use technology.