Cambridge, UK, July 15th 2020 –, a recognized leader in accelerating machine learning inference, today announced the industry’s most efficient accelerator for recommendation models, potentially saving hyperscale and tier one data center companies hundreds of millions of dollars every year.

Deep learning based recommendation models are one of the most common data center workloads. Use-cases include search, news feeds, adverts and personalized content. Recommendation models always contain a mix of dense and sparse features. This leads to complex memory access challenges for up to 80% of the time. With throughput constrained by memory bandwidth in typical compute infrastructure, expensive compute resources are highly under-utilized.

SEAL™ accelerates the memory-bound inference operations in recommendation models. This delivers large gains in latency-bounded throughput within existing infrastructure, enabling data center companies to scale rapidly and halve the infrastructure cost of their peak traffic capability. Energy consumption, a high-profile issue for data center companies, can be reduced by more than half.

“In designing SEAL, we knew we had to make it practical and easy to adopt”, said Peter Baldwin, CEO at “SEAL works seamlessly from within the deep learning framework PyTorch, fully preserves existing model accuracy and supports model co-location and sharding. It’s also complementary to existing compute accelerators and scalable, so adoption is as straightforward as possible.”

SEAL is available initially in the Open Compute Project M.2 Accelerator Module form factor, intended for use in Glacier Point carrier cards, delivering up to 384GB of DDR4 memory per carrier. First customer evaluations are anticipated in Q3 2020. Alternative form factors are being reviewed.

SEAL represents the lowest power, smallest form factor, easiest-to-deploy method of adding memory bandwidth to existing infrastructure used for recommendation models.

For more details go to or contact today at


About optimizes inference workloads such as recommendation models, recurrent neural networks and other deep neural networks with sparse features. This enables businesses to rapidly scale and improve their services while reducing capital costs and energy consumption. is proud to be a founding member of MLCommons, the benchmarking organisation driving machine learning innovation through For more information, please visit


Press Contact:

Giles Peckham