How to run AI workloads on mixed GPUs quickly and affordably
News
Kim Martineau
AI Hardware
The world is generating reams of data each day, and the AI systems built to make sense of it all constantly need faster and more robust hardware. We’re developing new devices and architectures to support the tremendous processing power AI requires to realize its full potential.
Our work
Bringing quantum-centric supercomputing to Illinois
Q & AKim MartineauIBM demonstrates extreme scale for content-aware storage with a 100-billion vector database
NewsPeter HessBuilding PyTorch-native support for the IBM Spyre Accelerator
Technical noteViji Srinivasan, Raghu Ganti, Mudhakar Srivatsa, Elpida Tzortzatos, Tabari Alexander, and Vaidyanathan SrinivasanLifting the cover on the IBM Spyre Accelerator
Technical noteJeffrey BurnsIt takes a village to make open infrastructure for AI a reality
NewsPeter Hess- See more of our work on AI Hardware
Analog AI: A New Design Paradigm
At IBM Research we’re developing a new class of Analog AI hardware, purpose built to help innovators realize the promise of the next stages of AI. Journey inside this unique architecture.
Publications
HRL-Topo: Hierarchical Reinforcement Learning with Large Language Models for Analog Circuit Topology Synthesis
- Prashanth Vijayaraghavan
- Luyao Shi
- et al.
- 2026
- DAC 2026
Conference paperStepPRM-RTL: Stepwise Process-Reward Guided LLM Fine-Tuning for Enhanced RTL Synthesis
- Prashanth Vijayaraghavan
- Apoorva Nitsure
- et al.
- 2026
- DAC 2026
Conference paperDeepView: A Generalizable Debugging Methodology for Enabling Large Language Models on Emerging AI Accelerators
- Kaoutar El Maghraoui
- Jordan Murray
- et al.
- 2026
- SERVICES 2026
Conference paperEnterprise-Scale RAS in IBM Z Memory Subsystem
- Pat Meaney
- None None
- et al.
- 2026
- DSN 2026
Conference paperVSIMO: A DC-DC SIMO-based Current Mismatch Compensator Supporting Per-Output Bidirectional Power Flow for Stacked Voltage Domains Achieving >82% System Efficiency under up to 360-mA Load Mismatch
- Yichen Xu
- Baoqi Zhu
- et al.
- 2026
- VLSI Technology and Circuits 2026
Conference paperA 1-to-1050-mA DLDO with Rising-Edge Computational Control and Load-Dependent Feedback Achieving 46-mV/ns DVS Rate and 0.15-ps FoM
- Yichen Xu
- Mao Li
- et al.
- 2026
- VLSI Technology and Circuits 2026
Conference paper
AI Hardware Center
The IBM Research AI Hardware Center is a global research collaboration hub dedicated to creating the next generation of systems and chips for AI workloads, as well as expanding joint research efforts across technology, architecture, and algorithms.
Explore the center