Our IPU lets innovators make new breakthroughs in machine intelligence

Explore Products

IPU-POD₁₆

Pre-configured with a 4 petaFLOP AI system, IPU-POD₁₆ is where you experience the power and flexibility of larger IPU systems.

Learn More

IPU-POD₆₄

16 petaFLOPS of AI-compute for both training and inference workloads, the IPU-POD₆₄ is designed for AI at scale.

Learn More

IPU-POD₁₂₈

With a powerful 32 petaFLOPS of AI compute, the IPU-POD128 is designed to provide cutting-edge performance for machine intelligence at datacenter scale.

Learn More

Graphcloud

A secure IPU cloud service to add state of the art AI compute on demand - no on-premise infrastructure deployment required.

Request Access

WORLD-CLASS PERFORMANCE IN MLPERF RESULTS

State of the art performance for natural language processing and computer vision as seen in latest MLPerf 1.0 training results

Best for Natural Language Processing

Train BERT in 9.39 minutes. These blistering fast results on an IPU-POD64 measured in industry standard MLPerf benchmarks for BERT in the commercially available, open category for production hardware.

Even the closed result for IPU-POD16 is seriously impressive - just under 35 minutes. We’re getting into supercomputer territory with a 5U machine you can buy from our partners or use in Graphcloud today.

MLPerf v1.0 Training Results | MLPerf ID: 1.0-1098, 1.0-1099

Best for Computer Vision

Train ResNet-50 in 14.48 minutes. In the tightly regulated closed category for MLPerf, we achieve stunning performance on an IPU-POD64 and a really impressive 37 minutes on our mainstream, commercially available IPU-POD16 system.

MLPerf v1.0 Training Results | MLPerf ID: 1.0-1026, 1.0-1028

More Benchmarks

Our Customers

We benchmarked with a wide range of interesting state-of-the-art networks and we see the IPUs are at least multiple times faster than the common GPUs. To give a number, I think it would be at least 10x faster