# Vertically Integrated, Horizontally Open | AI Factory Insider Ep 5

## Executive summary

The discussion outlines the concept of 'vertically integrated, horizontally open' AI factories, emphasizing that a common, accelerated computing foundation is required to support specialized AI across diverse industries. NVIDIA's CUDA platform is presented as the core enabler, providing a full-stack architecture that allows for maximum performance optimization while remaining open to external ecosystem partners (ISVs, middleware, etc.). The evolution of the platform is driven by reusable libraries (CUDAX) and the emerging paradigm of Agentic AI, where these libraries will be exposed as 'skills' for autonomous agents.

## Key takeaways

- Understanding 'Vertically Integrated, Horizontally Open': Vertically integrated means designing the platform for maximum performance across the entire AI full-stack (hardware, software, libraries). Horizontally open means that every layer in this stack is open to the entire ecosystem, allowing partners to contribute their IP and workloads.
- The Role of CUDA and CUDAX Libraries: CUDA is the foundational platform connecting applications to GPU hardware. CUDAX libraries are reusable, highly optimized Intellectual Property (IP) built on top of CUDA, providing domain-specific solutions (e.g., QDF for data processing, cuGraph for graphs) that accelerate development and time-to-market.
- Agentic AI and Skills: The future of the platform involves exposing CUDAX libraries as 'skills' for autonomous agents. This allows agents to leverage specialized tools—such as a data processing pipeline or a visual profiler—to perform complex tasks, acting as a 'force multiplier' for developers.
- Hardware and Software Co-evolution: NVIDIA maintains a tight coupling between hardware and software roadmaps (e.g., Hopper $ ightarrow$ Blackwell $ ightarrow$ Rubin). This co-evolution ensures that software advancements can often be applied to older hardware, maximizing the lifespan and utility of the GPU architecture.

## Technical details

- AI Factory Architecture: The model is described as 'vertically integrated' (optimized full-stack performance) and 'horizontally open' (allowing diverse partners and industries to build upon a common foundation).
- CUDA Platform: CUDA is the fundamental platform that enables accelerated computing by connecting applications to GPU hardware. It forms a deep platform stack from frameworks down to hardware controllers.
- Library Interoperability: The platform relies on interoperability between core libraries (e.g., cuBLAS for linear algebra, cuDNN for AI) and specialized libraries, allowing new AI components to plug into established classical computing mechanisms.
- Agentic Workflow Enhancement: CUDAX libraries are being exposed as 'skills' for agents. This allows agents to autonomously identify and utilize the correct tool (e.g., QDF for data processing) to solve complex, multi-step problems.

## Practical implications

- System architects must design for interoperability, ensuring that specialized components (libraries) can plug into a common, high-performance foundation (CUDA).
- The shift in development focus is moving from writing boilerplate code to defining and managing complex workflows and 'skills' for autonomous agents.
- Build engineers must account for the co-evolution of hardware and software, utilizing the full stack capabilities to ensure maximum performance and longevity across different hardware generations.
- The open nature of the platform allows for modular development, enabling teams to focus their efforts on unique business logic rather than reinventing fundamental components.

## Topics

Accelerated Computing, AI Factory Architecture, CUDA Programming Model, Agentic AI, Software Interoperability, CUDA, CUDAX Libraries, QDF, cuGraph, Bioneo / Qaquarians, Metropolis, cuBLAS, cuDNN, Hopper, Blackwell, Rubin

Source: https://www.youtube.com/watch?v=JIeLDmAG8y0