Follow Datanami:
November 4, 2020

Intel Buys Another AI Startup

Intel Corp. has quietly acquired another AI platform developer, Israeli-based Cnvrg.io.

The acquisition, confirmed by Intel late Tuesday (Nov. 3) to the web site TechCrunch.com, is the latest in a flurry of deals by the chip maker (NASDAQ: INTC) as it fills out its AI and machine learning portfolio.

Terms of the acquisition were not disclosed. Intel did say the Cnvrg.io would continue to operate as an independent company.

Founded in 2016, Jerusalem-based Cnvrg.io’s data science platform is designed to help development teams manage and scale AI models. Early backers include Prashant Malik, co-creator of the Cassandra NoSQL database management system.

The startup’s most recent release is an MLOps dashboard designed to boost machine learning server utilization, a gap the startup refers to as “computational debt.” The framework is designed to boost utilization of CPUs, graphics processors and memory resources by as much as 80 percent.

The startup’s MLOps dashboard “increases utilization by providing an advanced resource management solution with our meta scheduling tools, and straightforward attachment of compute resources to jobs,” Cnvrg.io co-founder Yochay Ettun noted in a blog post.

That framework along with the startup’s partnerships with key AI vendors likely caught Intel’s attention as the chip maker beefs up its AI capabilities in key market segments like computer vision. Those machine learning applications also would help it maintain is lead in the enterprise server market.

Cnvrg.io brings with it a growing list of cloud infrastructure and networking partners, including Nvidia, NetApp and IBM Red Hat. It recently announced an AI operating system designed to accelerate workloads via native support for Nvidia A100 Tensor Core GPUs, thereby delivering multiple instances of a single GPU for machine and deep learning workloads. The startup claims its machine learning platform is the first to integrate the “multi-instance GPU” capability.

The capability partitions A100 GPUs into as many as seven accelerators, with each instance running simultaneously with dedicated computing, memory and cache resources.

Also this summer, the startup announced a partnership with NetApp (NASDAQ: NTAP) billed as a machine learning data science pipeline. The collaboration includes data set caching running on clusters managed via the Kubernetes orchestrator. The framework allows attachment of models to data sets via integration with Nvidia’s registry of GPU-tuned AI software.

The deal for Cnvrg.io is Intel’s second AI-related acquisition in as many weeks. The chip maker initiated its latest foray in the AI market with its late October deal for AI optimization software vendor SigOpt. The San Francisco-based vendor focuses on metrics used to boost the performance of machine learning models.

Hence, both acquisitions address what Intel notes is the growing complexity of machine learning models and the need to squeeze more performance from hardware accelerators used for model training and inference.

The chip maker’s stock has been trending up since it announced the SigOpt acquisition.

Recent items:

Intel Acquires Model Optimizer SigOpt

Staying on Top of ML Model and Data Drift

AI Startup Uses FPGAs to Speed Training, Inference

Datanami