Oracle Cloud Infrastructure is strategically deepening its partnership with AMD, integrating both EPYC processors for flexible, cost-effective compute and advanced Instinct GPUs to power its burgeoning AI and high-performance computing services, positioning itself as a formidable competitor in the rapidly evolving cloud landscape.
In a significant strategic move, Oracle Cloud Infrastructure (OCI) is solidifying its position in the competitive cloud market by forging a deeper alliance with Advanced Micro Devices (AMD). This collaboration extends across two critical frontiers: enhancing general-purpose cloud computing with AMD’s EPYC processors and making a substantial push into the high-demand AI and HPC segments with AMD’s Instinct GPUs. This dual-pronged approach aims to offer OCI customers unparalleled balance between cost and performance, directly challenging established cloud providers and catering to the surging demand for AI infrastructure.
The Foundation: AMD EPYC Processors Powering OCI Compute
At the core of Oracle’s revamped cloud strategy is the adoption of 4th generation AMD EPYC processors for its new OCI Compute E5 instances. This integration is designed to offer organizations a better balance between cost and performance, a perennial challenge in cloud environments. Unlike the more rigid instance options often found with other providers, OCI’s flexible instances empower customers to allocate CPU cores and memory precisely according to their specific requirements, significantly reducing expenses associated with unused computing resources. This granular control has already resulted in substantial savings, with existing flexible instances on AMD processors collectively saving OCI customers approximately $40 million per year, as reported by Oracle’s official announcement.
The upcoming OCI Compute E5 instances bring a suite of enhancements across various instance types:
- Standard Instances: These are the preferred platform for web servers, enterprise applications, and development environments. They are expected to deliver 33% better performance per core and 50% better memory bandwidth compared to prior E4 instances.
- High-Performance Computing (HPC) Instances: Catering to complex mathematical and scientific problems, the E5 HPC instances offer a 40% better price-performance ratio than their predecessors. By leveraging cluster networking, these instances can combine processing power to tackle tasks traditionally requiring supercomputers, such as AI model training and complex simulations.
- Dense-IO Instances: Designed for demanding workloads like large databases and big data, these instances boast 50% higher storage capacity and 63% higher storage performance than the prior E4 Dense-IO instances.
These instances will support a wide array of operating systems, including Oracle Linux, Windows, and Red Hat, along with numerous installable images from the Oracle Marketplace. High-profile clients like Uber, Phenix, SoundHound, and Xactly have already expressed their intent to leverage or are currently benefiting from OCI’s AMD-powered compute offerings, underscoring the practical advantages of this collaboration.
The AI Revolution: Oracle’s Bet on AMD Instinct GPUs
Beyond traditional compute, Oracle is making a bold play in the burgeoning artificial intelligence market by strategically adopting AMD Instinct GPUs. This move comes at a time when the demand for AI capacity is skyrocketing, and cloud providers like OCI have been grappling with Nvidia GPU shortages, which have impacted short-term growth. Instead of attempting its own silicon program, Oracle has opted to partner with AMD to deploy its high-performance AI accelerators.
Oracle is set to be among the first to deploy AMD’s Instinct MI300X AI and HPC GPUs in its cloud services, with initial deployments expected in early 2024. Looking further ahead, OCI has announced plans to deploy a staggering 50,000 MI450 processors, with the deployment commencing in the third quarter of 2026 and expanding into 2027 and beyond, according to Reuters. This massive scale signifies Oracle’s deep commitment to building robust AI infrastructure.
Introducing AMD Instinct MI355X and Zettascale AI Clusters
Oracle and AMD have also detailed the availability of AMD Instinct MI355X GPUs on OCI, promising over 2x better price-performance for large-scale AI training and inference workloads compared to the previous generation. OCI plans to offer Zettascale AI clusters accelerated by the MI355X, capable of scaling up to 131,072 GPUs. These superclusters are designed with a high-throughput, ultra-low latency RDMA cluster network architecture to support the most demanding AI applications.
Key features of the MI355X-powered shapes on OCI include:
- Significant Performance Boost: Up to 2.8x higher throughput for AI deployments, enabling faster results and lower latency.
- Larger, Faster Memory: Equipped with 288GB of High-Bandwidth Memory 3 (HBM3) and up to eight terabytes per second of memory bandwidth, allowing large models to execute entirely in memory.
- New FP4 Support: Cost-effective deployment of modern large language models (LLMs) and generative AI models with the new 4-bit floating point compute (FP4) standard.
- Dense, Liquid-Cooled Design: Maximizes performance density at 125 kilowatts per rack, supporting 64 GPUs per rack (1,400 watts each) for faster training times.
- Open-Source Stack: Leverages AMD ROCm, an open software stack that provides flexible architectures and easy migration of existing code without vendor lock-in.
- Network Innovation with AMD Pollara: Oracle will be the first to deploy AMD Pollara AI NICs on backend networks, providing advanced RoCE functionality and supporting open industry standards from the Ultra Ethernet Consortium (UEC) for high-performance networking.
Strategic Implications and Community Impact
Oracle’s deep integration of AMD’s processors and GPUs marks a significant shift, signaling its determination to compete aggressively in the cloud and AI sectors. The company’s focus on AI is evident in its intention to integrate generative AI capabilities into its application lineup, including a new collaboration with Canadian AI software firm Cohere. This positions Oracle alongside tech giants like Meta Platforms, Alphabet, and Amazon in the race to invest in and leverage AI, with Oracle specifically focusing on its cloud infrastructure tools.
This strategic move also has broader implications for the AI ecosystem. OpenAI, the creator of ChatGPT, has made significant moves to diversify its computing power. It not only signed a multi-year deal with AMD for processors, giving it an option to buy up to a 10% stake in the chipmaker, but also inked a massive five-year cloud deal with Oracle that could be worth up to $300 billion. OpenAI’s pursuit of multiple computing sources, even planning to design its own AI chips with Broadcom, highlights the industry’s critical need for diverse and powerful hardware. Oracle’s “AI superclusters” with AMD, powered by AMD’s “Helios” rack design, directly address this need for large-scale, integrated AI solutions.
For the fan community—developers, engineers, and tech enthusiasts—Oracle’s AMD strategy translates into more choice and potentially better price-performance in the cloud. The flexible instance options and the open-source compatibility of AMD ROCm offer greater freedom and efficiency for developing and deploying AI models and other demanding workloads. This move solidifies Oracle’s commitment to being an AI-ready cloud provider, capable of supporting the next generation of intelligent applications and services.
