Key takeaways
- Modular's partnership with AWS aims to enhance the MAX Platform's accessibility and power for a global user base.
- The AWS Marketplace will exclusively host the MAX Platform, optimally utilising Amazon's Graviton CPUs for AI applications.
- The MAX Platform offers significant performance improvements and cost savings over existing AI infrastructures.
- AWS executives highlight the collaboration's role in advancing AI capabilities for enterprises and startups.
- The MAX Platform features hardware portability, allowing for flexible deployment across various environments.
At ModCon 2023, Modular, the company behind the Mojo programming language, announced an exclusive collaboration with Amazon Web Services (AWS). The partnership aims to extend the reach of the MAX Platform to AWS production services worldwide, ushering in advanced AI capabilities for a vast user base.
Check out all the key announcements made at ModCon 2023.
Notably, the AWS Marketplace becomes the exclusive venue for leveraging the MAX Platform on Graviton CPUs, Amazon's ARM-based processors designed for compute-intensive workloads such as AI programs. The MAX Platform optimally enhances Graviton CPUs, delivering AI model execution with up to 5X higher performance and up to 80% cost savings compared to existing AI infrastructure.
Bratin Saha, AWS VP of machine learning & AI services, emphasised the significance of this collaboration in advancing AI capabilities. “At AWS, we are dedicated to shaping the future of AI by delivering services that reduce costs and accelerate progress for enterprises and startups. The MAX Platform amplifies this objective for millions of AWS customers, facilitating the rapid deployment of the latest GenAI innovations and traditional AI use cases,” remarked Saha.
One unique feature of the MAX Platform is its hardware portability, enabling seamless migration of workloads to Graviton without incurring any migration costs. This portability extends to the utilisation of MAX Serving and Mojo directly in the engine, offering full customizability and workload tuning options.
Furthermore, Modular announced a strategic technology partnership with NVIDIA to integrate the benefits of their accelerated compute platform into MAX, simplifying CPU+GPU development for AI developers. Key features unveiled include MAX Engine extensibility with Mojo, MAX Engine GPU support, the release of Mojo SDK v0.6, and the open-sourcing of Mojo documentation.
MAX GPU offers cutting-edge compatibility for NVIDIA H100, H200, A100, and L40 series GPU accelerators, as well as the recently introduced Grace CPU Superchip. This hardware integration extends across all facets of MAX, encompassing MAX Engine, MAX Serving, and Mojo, delivering unmatched heterogeneous computing capabilities to the field of AI.
Developers can now leverage a unified toolchain that caters to a spectrum of AI applications, spanning from GenAI to various other AI scenarios. This approach unlocks innovative CPU+GPU programming models, promising unparalleled performance and cost-effectiveness.
“Developers everywhere are helping their companies adopt and implement generative AI applications that are customised with the knowledge and needs of their business,” said Dave Salvator, director of AI and Cloud at NVIDIA.
ModCon 2023, Modular's inaugural developer conference, featured prominent figures in the AI landscape, including Bratin Saha (VP, AWS), Kari Ann Briski (VP, NVIDIA), Bryan Catanzaro, (VP, NVIDIA), Lex Fridman, (AI researcher, MIT), Shawn “Swyx” Wang (Latent.Space), Damien Sereni (director, Meta), Jeremy Howard (Fast.ai), and Michele Catasta, (VP, Replit).
Excitement surrounds this collaboration with early access to MAX on AWS Marketplace available at modul.ar/max, and further updates anticipated in Q1 2024. The partnership signifies a significant step towards ensuring the widespread availability of MAX, marking a pivotal moment in the evolution of AI capabilities.
