Deci AI

What Deci AI does

Deci AI, acquired by NVIDIA, enabled deep learning to reach its potential by using AI to build better AI. Its end-to-end deep learning acceleration platform allowed AI developers to build, optimize, and deploy faster and more accurate models for various environments including cloud, edge, or mobile. The platform could increase deep learning model inference performance by 3x-15x. NVIDIA itself is a world leader in artificial intelligence computing, focusing on cybersecurity, AI model pre-training, specialized AI agents, simulation, and AI infrastructure development. They provide compute power, tools, and algorithms for various AI applications across industries.

Products

• NVIDIA GB300 NVL72A platform used for AI training, setting world records in pre-training large language models like DeepSeek-V3.
• NVIDIA Vera Rubin platformA platform, accelerated by the NVIDIA Vera CPU, for post-training frontier open models to build specialized AI agents, offering high intelligence per dollar for domain-specific agents.
• NVIDIA Agent ToolkitExpands with NVIDIA PhysicsNeMo and CUDA-X, used by industry leaders to accelerate chip-to-systems design workflows.
• NVIDIA Vera CPUSpeeds up next-gen CPU and GPU designs with up to 1.5x higher performance and accelerates the Vera Rubin platform.
• NVIDIA Synthetic Video Detector NIM™Helps support source verification and media integrity by detecting synthetic video.
• NVIDIA Cosmos EdgeBrings world models to Edge GPUs, capable of understanding and generating text, image, video, ambient sound, and action.
• NVIDIA NemoClaw™ on DGX Station™Provides engineers with a local agent stack for continuous project development.
• DiffusionGemmaAn open model developed by Google DeepMind, optimized by NVIDIA to run on NVIDIA RTX PRO platform, NVIDIA DGX Spark systems, and GeForce RTX GPUs, for parallel text generation.
• NVIDIA DGX GB300An AI supercomputer used for AI education and research, particularly in the public sector.

Key capabilities

• End-to-end deep learning acceleration platform
• Build, optimize, and deploy AI models
• Faster and more accurate models
• Increased deep learning model inference performance (3x-15x)
• Support for cloud, edge, and mobile environments
• AI computing leadership
• Cybersecurity solutions for the AI era
• World record AI model pre-training capabilities
• Specialized AI agent building
• Simulation tools for chip-to-systems design
• AI infrastructure development
• Data analytics
• Machine learning
• Predictive analytics
• Recommender systems
• Speech AI
• Apache Spark acceleration
• RAPIDS acceleration
• Synthetic video detection
• Omni-model capabilities (text, image, video, sound, action)
• Local agent stack for engineers
• Optimized open models for local AI
• Real-time systems governed by physics for AI worlds

Use cases

• Building and deploying deep learning models
• Optimizing deep learning model performance
• Accelerating deep learning inference
• Securing AI systems
• Pre-training large language models
• Building domain-specific AI agents
• Accelerating chip-to-systems design workflows
• Expanding AI capabilities and innovation
• AI education and research
• Strengthening American manufacturing
• Accelerating healthcare and scientific discovery
• Generative AI applications
• Speech recognition
• Medical imaging
• Improved supply chain management
• Data analysis
• Machine learning development
• Predictive modeling
• Developing recommender systems
• Speech AI applications
• Verifying video authenticity
• Edge AI applications with world models
• Local AI agent development
• Real-time AI systems in graphics

AI approach

Deci AI, acquired by NVIDIA, focused on using AI to build better AI, providing an end-to-end deep learning acceleration platform to optimize and deploy faster and more accurate models. NVIDIA itself is a leader in AI computing, offering platforms and tools for various AI applications, including generative AI, speech recognition, medical imaging, and supply chain management. They develop specialized AI agents, accelerate chip design, and provide infrastructure for AI development and research.

Tech named: DeepSeek-V3 671B, NVIDIA Vera CPU, NVIDIA PhysicsNeMo, CUDA-X, NVIDIA Synthetic Video Detector NIM, NVIDIA Cosmos Edge, NemoClaw, DiffusionGemma, Apache Spark, RAPIDS

Industries served

• Software Development
• Cybersecurity
• Public Sector
• Healthcare
• Scientific Research
• Manufacturing
• Supply Chain Management
• Chip Design
• Creative Industries

What it says sets it apart

• Uses AI to build better AI
• End-to-end deep learning acceleration platform
• Significant inference performance improvement (3x-15x)
• World record holder in AI model pre-training
• Comprehensive AI computing leadership
• Focus on open and secure AI alliances
• Broad range of AI tools and platforms for various applications and environments (cloud, edge, mobile)
• Integration of hardware (GPUs, CPUs) and software for optimized AI performance

Funding rounds we track

From the AI funding tracker — rounds as reported by the linked publications.