Intel will acquire model optimization startup SigOpt

Optimizing AI models

SigOpt's platform and expertise will likely help Intel broaden its capabilities for bringing machine learning and deep learning models to scale in production, said Dave Schubmehl, research director for cognitive/artificial intelligent systems and content analytics at IDC.

The deal also provides Intel with an opportunity to expand the work it has done with Intel's oneAPI, the tech vendor's unified programming model designed to simplify development across multiple architectures, and OpenVino, an open source toolkit for optimizing deep learning models, Schubmehl noted.

By expanding on the work it has done with those products, the deal enables Intel "to make more software services available to machine learning developers, thereby encouraging them to consider and use Intel infrastructure when running their models," Schubmehl said.

Terms of the deal were not disclosed.

Founded in 2014 and based in San Francisco, SigOpt sells a platform that automates model optimization by tuning hyperparameters. The platform can optimization models with up to 100 hyperparameters, enabling users to speed up model training, improve model performance and increase computational efficiency.

While SigOpt lists at least several customers, its platform was still in beta. According to business information provider Crunchbase, SigOpt has raised $8.7 million in funding. SigOpt's team, including CEO and co-founder Scott Clark and CTO and co-founder Patrick Hayes, will join the Machine Learning Performance team at Intel.

The deal will likely not affect current SigOpt customers, Schubmehl said.

Competitive edge

The acquisition comes as Intel makes new efforts to improve its AI offerings. Earlier this year, Intel Capital, the tech vendor's investment and venture capital arm, invested $132 million in 11 AI, automation and chip design startups.

[The acquisition enables Intel] to make more software services available to machine learning developers, thereby encouraging them to consider and use Intel infrastructure when running their models.

In February, Intel said it would end support for its Nervana NNP line in favor of building AI processors based on technology from Habana Labs, an AI hardware vendor Intel acquired for $2 billion in 2019. The sudden decision to end support for the Nervana NNP line, which include chips for AI training and inference, surprised analysts, who expected a slow wind down for the products.

Intel's acquisition of SigOpt could help Intel against its battle with competitors, such as Nvidia, for AI dominance.

"The reality of the market is that this is not just about hardware, but also about providing tools and software to help machine learning developers and data scientists create and deploy models with more accuracy, speed and flexibility," Schubmehl said. "The acquisition of SigOpt by Intel definitely will help with this."