Red Hat, the IBM-owned open source software company, is acquire Neural Magic, a startup that optimizes AI models to run faster on standard processors and GPUs.
Terms of the deal weren’t disclosed.
MIT researcher Alex Matveev and Professor Nir Shavit founded Somerville, Massachusetts-based Neural Magic in 2018, inspired by their work on high-performance execution engines for AI.
Neural Magic's software goals to process AI workloads on processors and GPUs at speeds corresponding to specialized AI chips (e.g. TPUs). By running models on commodity processors, which usually have more available memory, the corporate's software can realize these performance gains.
Big tech firms like AMD and plenty of other startups including NeuReality, Deci, CoCoPie, OctoML and DeepCube offer some form of AI optimization software. But Neural Magic is certainly one of the few with a free platform and a group of open source tools to enhance it.
Neural Magic had thus far managed to lift $50 million in enterprise capital from backers resembling Andreessen Horowitz, New Enterprise Associations, Amdocs, Comcast Ventures, Pillar VC and Ridgeline Ventures.
Red Hat CEO Matt Hicks said Neural Magic's work on vLLM, an open source model deployment project, is of particular interest to Red Hat. With Neural Magic, the open source company advantages from a vLLM-based “enterprise-grade” stack that allows customers to optimize and deploy models in cloud environments with complete control over infrastructure and security, Hicks said.
Red Hat is already involved within the vLLM project and uses it to run models in products resembling Red Hat Enterprise Linux AI and Red Hat OpenShift AI. Hick said that through the joint work of Red Hat and Neural Magic, its infrastructure partners will have the ability to raised scale AI across platforms, and its integrated service provider partners will gain more robust inference and performance to integrate into their offerings.
“AI workloads have to run wherever customer data resides within the hybrid cloud. This makes flexible, standardized and open platforms and tools a necessity as they permit organizations to pick the environments, resources and architectures that best fit their unique operational and data needs,” Hicks said in a press release. “We are pleased to enhance our hybrid cloud-focused AI portfolio with Neural Magic's groundbreaking AI innovation, further advancing our ambition to be not only the 'Red Hat' of open source, but in addition the 'Red Hat' to be of AI.”
Red Hat's acquisition of Neural Magic comes as the corporate is making quite a lot of AI-related announcements this week at KubeCon, the annual computing conference in Salt Lake City. Red Hat also introduced Climatik, an information center energy efficiency optimization tool developed in collaboration with Intel, Bloomberg and IBM, in addition to latest versions of its OpenShift AI and Device Edge development platforms.
As the ability demands and costs of making and deploying AI rise, major technology providers have moved to snap up firms with products that might help optimize AI algorithms. In April, Nvidia acquired Run:ai, an organization that makes it easier for developers to administer and optimize their AI infrastructure. And in October 2020, Intel bought SigOpt, a model optimization specialist.
