Infrastructure Trends and Strategies: Artificial Intelligence Workloads

Related Links



Meet the Experts

Photo of Ashish Nadkarni
Ashish Nadkarni

Group Vice President and General Manager, Infrastructure Systems, Platforms and Technologies and BuyerView Research

Photo of Heather West, PhD
Heather West, PhD

Research Manager, Infrastructure Systems, Platforms and Technologies Group

Photo of Peter Rutten
Peter Rutten

Research Vice-President, Infrastructure Systems, Platforms and Technologies Group, Performance-Intensive Computing Solutions Global Research Lead



IDC's Infrastructure Trends and Strategies: Artificial Intelligence Workloads service looks at the impact of performance-intensive computing (PIC) workloads and workflows on the infrastructure hardware and software markets. IDC defines PIC as a collection of machine and deep learning, big data and analytics, and modeling and simulation (M&S) workloads (aka "HPC") and use cases. Specific focus is put on the infrastructure needs of newer technologies (e.g., SAP HANA, Greenplum, and Oracle Advanced Analytics), nonrelational analytic data stores (e.g., Hadoop, Spark, MongoDB, and Cassandra), continuous analytic tools (e.g., Amazon Kinesis, Splunk Universal Forwarder, and Microsoft Azure Data Factory), relational data warehouses, analytic and performance management applications, and business intelligence and analytic tools and platforms (including AI software platforms). Also included are the infrastructure needs for AI data preparation, AI model training, and AI inferencing from edge to core to cloud. The service also covers M&S workloads/use cases delivered via specialized (supercomputing) and general-purpose (including cloud-based) infrastructure and infrastructure as a service. The service covers new computing paradigms such as quantum computing including enabling technologies, platforms, systems, and services. The impact to infrastructure is examined across compute and processor architectures, storage interfaces and system types, data organization, and storage capacity. These data points will be used for segmentation and forecasting.


Markets and Subjects Analyzed


  • Infrastructure (and infrastructure-as-a-service) trends, strategies and market outlook for AI workloads/use cases
  • Infrastructure (and infrastructure as a service) trends, strategies, and market outlook for modeling and simulation workloads/use cases.
  • Infrastructure (and infrastructure-as-a-service) trends, strategies, and market outlook for big data and analytics workloads/use cases
  • Quantum computing technologies, platforms, systems, services, and use cases

Core Research


  • Infrastructure PIC Workloads Taxonomy
  • Infrastructure PIC Market Size and Forecast
  • Infrastructure PIC Best Practices and End-User Adoption Trends
  • Infrastructure PIC Use Cases and Evolving Applications Requirements
  • PIC Adoption Trends in Shared and Dedicated Cloud Infrastructure
  • Tracking Supercomputing Trends (Top 500) and Events
  • Hardware Accelerators Used for Compressed Time to Value from Data Sets Used in PIC Workloads
  • Storage Systems Trends for PIC Workloads
  • Quantum Computing Developments, Ecosystems, Vendors, and Technologies

In addition to the insight provided in this service, IDC may conduct research on specific topics or emerging market segments via research offerings that require additional IDC funding and client investment.


Key Questions Answered


  1. What is the build and services revenue from PIC workloads?
  2. What are the infrastructure hardware and software requirements imposed by PIC workloads?
  3. What are some of the data life-cycle challenges associated with PIC workloads?
  4. What are the optimal compute and storage configurations for PIC workloads?
  5. What is the role of accelerated computing (GPUs, FPGAs, ASICs, manycore processors, and emerging acceleration technologies), NVMe, tiering, deduplication, and compression as they are related to PIC?
  6. How is quantum computing disrupting PIC? What are the trends associated with shifting classical computing workloads to quantum computing?