Latest News

How IBM is Helping Clients Deploy Foundation Models and AI Workloads with New GPU Offering on IBM Cloud

Tailored for AI, latest infrastructure from IBM Cloud will support WatsonX services and is designed to support compute-intensive workloads

By Rohit Badlaney | GM IBM Cloud Product
May 09, 2023

Across the globe, adoption of artificial intelligence (AI) is steadily growing. Most recently, we’ve seen the emergence of performance intensive computing as a service (PICaaS) to support foundation model workloads. While foundation models can represent a drastic change in how businesses can create and scale AI, few organizations have the skills and infrastructure needed to build or utilize foundation models. With the combination of our AI stack, cloud technology and industry expertise, IBM is committed to bringing the power of foundation models to enterprise clients and to help them optimize outcomes and responsibly tap into AI to digitally transform.

Performance Intensive Computing at IBM: Our Roadmap to Success

IBM Research is applying performance intensive computing solutions for training and executing foundation models. We recently introduced Vela, IBM’s first AI-optimized, cloud-native supercomputer hosted on IBM Cloud, for the IBM Research community. To support this initiative, we partnered with industry innovators like NVIDIA. IBM Research designed Vela to scale up at will and readily deploy similar infrastructure into IBM Cloud data centers. Vela is now our go-to environment for IBM researchers creating our most advanced AI capabilities, including our work on foundation models, and where we collaborate with partners to train many kinds of models.

By using IBM Cloud for its GPU requirements, Vela can potentially help tackle a variety of real-world problems across science, healthcare, manufacturing and more. For example, IBM recently trained a geospatial model on Vela. Built from IBM’s collaboration with NASA, the model is designed to convert satellite data into high-resolution maps of floods, fires, and other landscape changes to reveal our planet’s past and hint at its future.

On the heels of Vela’s success, IBM is directing even more focus towards empowering the future of AI for business by providing more access to GPU-based computing and GPU-accelerated watsonx services. Today, IBM is announcing the availability of additional GPU offerings, featuring NVIDIA GPUs, on IBM Cloud, which bring innovative GPU infrastructure designed to train foundation models for enterprise workloads, and which will be used to serve enterprise-class foundation models via watsonx services. IBM’s GPU offerings can be used for many workloads, including analytics, training, and serving large language models (LLMs). Later this year, IBM will offer full stack high-performance, flexible, AI-optimized infrastructure, delivered as a service on IBM Cloud, for both training and serving foundation models. This full stack approach aims to provide a one-stop approach for building enterprise-grade foundation models, encompassing software, middleware, and infrastructure.

How We Help Clients Lead with Speed: Our Technology and Collaborators for Success

We have seen that the market interest in performance intensive computing as a service has steadily been growing in recent years. However, many enterprises seem to struggle with the costs, performance needs, and scalability issues associated with foundation models. That is why we are offering an end-to-end performance-intensive computing as a service, built on an infrastructure that includes the resiliency, performance, security, that our clients demand, particularly those in regulated industries, such as financial services.

We’ve also enlisted several collaborators to help us deliver success, including using PyTorch in our stack. PyTorch is a machine learning framework for building deep learning models. We are also working with, an open-source unified compute framework that is helping IBM Research to streamline the data pre- and post-processing steps of the AI workflow. This includes cleaning data as well as simplifying model adaptation and validation after the model is trained.

Clients Across Industries Can Benefit from IBM’s GPU Infrastructure

Enterprises all over the world and across sectors, including heavily regulated industries, can use IBM’s foundation model stack for AI applications to help improve business outcomes to better meet the needs of their customers. For example:

  • Financial Services – Financial service institutions have vast amounts of data on client interactions. They can use this data to fine tune foundation models, which can help provide better client experience and use data-driven trends to identify fraudulent transactions. Foundation models can also be used to improve operations and compliance based on historical data. IBM Cloud for Financial Services is built to help regulated industries manage the complexity of data privacy, security, resiliency, and their compliance needs. IBM’s GPU infrastructure can help financial institutions tackle complex transactions more quickly and utilize risk adverse applications while providing value for their clients from foundation models. 
  • Manufacturing – AI can have a positive impact on the manufacturing industry, helping to improve everything from delivery time to inspection quality. Manufacturers can benefit from IBM’s foundation model stack designed for accelerated time-to-market and faster innovation, which is critical to maintaining competitive supply chain operations.

Learn More

As part of IBM’s mission to make AI for business more accessible, IBM also unveiled WatsonX earlier today, which leverages IBM's GPU capabilities on IBM Cloud. For more information on IBM’s GPU capabilities, please visit HERE.

Statements regarding IBM’s future direction and intent are subject to change or withdrawal without notice and represent goals and objectives only.

Article Categories