AWS re:Invent 2025 - Speed to Impact: AI Factory in the Cloud with NVIDIA on AWS (AIM116)

Summary of "AWS re:Invent 2025 - Speed to Impact: AI Factory in the Cloud with NVIDIA on AWS (AIM116)"

AI Adoption and the Need for Scalable Infrastructure

• AI adoption is widespread across industries, driving a new industrial revolution powered by "AI factories"
• Three key scaling laws for AI are emerging:
  1. Pre-training scaling: Leveraging internet data to teach models
  2. Post-training scaling: Model learning and reasoning
  3. Test-time scaling: Longer, more thoughtful reasoning before responding
• As AI models become more intelligent, there is a higher demand for scalable compute infrastructure

The NVIDIA-AWS Partnership for AI Factories

• NVIDIA's full software stack is integrated with AWS, including CUDA libraries, DGX Hub, NVIDIA AI Enterprise, Omniverse, and inference microservices
• This ecosystem is built upon NVIDIA's high-performance GPU and networking infrastructure

Introducing NVIDIA DGX Cloud on AWS

• NVIDIA and AWS co-engineered DGX Cloud to provide an optimized, full-stack AI platform:
  • Infrastructure: High-performance GPUs, networking, and Lustre storage
  • Orchestration: Optimized EKS and Run:AI GPU orchestration
  • Software: Enterprise-grade AI software with production support
  • Expertise: Access to NVIDIA AI and cloud experts

Service Now's AI Goals and Experiences with DGX Cloud

• Service Now's goals:
  • Build efficient, cost-effective "agentic AI" systems to power their enterprise platform
  • Develop powerful foundation models that can be reused across use cases
• Experiences with DGX Cloud:
  • Enabled rapid model development and training without infrastructure management overhead
  • Leveraged Run:AI for optimized GPU utilization
  • Benefited from the enterprise-grade software and support

SLB's AI Goals and Experiences with DGX Cloud

• SLB's goals:
  • Develop domain-specific foundation models to generate synthetic data and responses for geophysical and petrophysical applications
  • Integrate AI agents into their existing portfolio of physics-based and machine learning tools
• Experiences with DGX Cloud:
  • Accelerated model development and deployment compared to on-premises infrastructure
  • Ability to fine-tune models with customer-specific data to build trust
  • Leveraged the elastic, scalable nature of the cloud to respond to evolving needs

The Path Forward: Flexibility and Continued Innovation

• Both companies plan to continue leveraging the NVIDIA-AWS partnership, utilizing a mix of cloud and on-premises infrastructure as needed
• The flexibility and portability of the NVIDIA ecosystem across cloud and on-premises environments is a key advantage
• As AI continues to evolve, the ability to rapidly innovate and respond to customer needs will be critical

Key Takeaways

• The emergence of scalable "AI factories" powered by cloud infrastructure is enabling new levels of AI adoption and innovation across industries
• The NVIDIA-AWS partnership provides an optimized, enterprise-grade platform for developing and deploying AI at scale
• Companies like Service Now and SLB are leveraging DGX Cloud to accelerate their AI goals, from building efficient agentic systems to developing domain-specific foundation models
• Flexibility, portability, and continued innovation in the NVIDIA-AWS ecosystem will be crucial as AI becomes an increasingly integral part of business operations