en.Wedoany.com Reported - Campbell, California, USA, local time April 29, 2026 – Unstructured data management company Komprise officially announced it has been granted a U.S. patent (Patent No. US12,566,637) for a technology named "Elastic Sharing," designed to systematically solve the industry-wide problem of low GPU, memory, and network resource utilization when processing large-scale unstructured datasets through dynamic partitioning.
The resource idleness issue targeted by this patent is already supported by clear academic data. According to a 2025 Cornell University study, the average utilization rate of current GPU clusters is only 50%, and even during active jobs, GPUs remain idle 14% to 76% of the time. Mike Peercy, CTO and co-founder of Komprise, pointed out that enterprise unstructured data is growing at a rate exceeding 20% annually, while IT budgets are increasingly strained by the demand for new AI infrastructure. Improving the utilization efficiency of expensive computing resources has become an unavoidable cost control imperative for CIOs.
The core breakthrough of Elastic Sharing technology lies in replacing traditional one-time static load balancing with a dynamic reallocation mechanism. Static solutions can only perform data partitioning once at the start of a job and cannot adapt to the varying task completion speeds of different nodes during job execution. This causes GPUs that finish tasks first to enter idle waiting, while nodes that finish later become the bottleneck delaying the overall job. Komprise's Elastic Sharing continuously monitors the work progress of each computing unit, automatically assigning new tasks once a node completes its work. It can also automatically perform rebalancing scheduling based on the characteristics of unstructured data, such as unknown directory depth and varying file sizes, thereby achieving near-linear scaling speedup.
The commercial value generated by this patent directly targets the total cost of ownership of AI data pipelines. The workloads accelerated by Elastic Sharing cover key scenarios such as AI data ingestion, metadata extraction, large-scale data migration, hot-cold tiering, and sensitive data management. Komprise CEO and co-founder Kumar Goswami previously stated publicly that before investing in AI, enterprises must first transform unstructured data from a scattered state into structured data assets usable for model training and inference; otherwise, the reliability and return on investment of AI outputs will be threatened. Elastic Sharing addresses this from the dimension of computing power utilization efficiency, reducing the resource consumption and time cost of this data preparation phase.
From the perspective of the company's overall strategy, the Elastic Sharing patent is the latest piece in Komprise's puzzle to build a technological moat around "AI data lifecycle management." In February 2026, the company launched a serverless computing data service named KAPPA, which automates infrastructure management during metadata extraction and data preparation, allowing users to focus on file-level customized logic design without worrying about the scale and scheduling of computing clusters. The addition of Elastic Sharing further strengthens the underlying support for this strategy: KAPPA provides orchestration capabilities for business logic, while Elastic Sharing ensures the underlying GPU cluster achieves the most efficient parallel execution within the orchestrated task flow. Together, they form a complete technology stack from data governance to computing power optimization.
Founded in 2014 by Kumar Goswami, Krishna Subramanian, and Mike Peercy, Komprise is headquartered in Campbell, California, USA. It has raised a total of $85 million in funding, completing a $37 million Series D round in 2023. Its core platform is positioned for analytics-driven unstructured data management, providing discovery, classification, migration, tiering, and lifecycle management capabilities for file and object data across on-premises and cloud environments through three major components: Intelligent Data Management, Intelligent Data Workflows, and a Global Metadata Repository. The company was recognized as a Cloud Data Management Leader in the 2026 CRN Cloud 100 list.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
