NVIDIA Acquires Predictive AI Startup Kumo AI

NVIDIA has acquired Kumo AI, a Mountain View, California-based startup that develops artificial intelligence models for business prediction, according to reporting by Fortune. Financial terms were not disclosed, though The Information reported the deal was valued at more than $400 million.

Kumo's three co-founders, Vanja Josifovski, Chief Executive Officer, Hema Raghavan, Head of Engineering, and Jure Leskovec, Chief Scientist, have joined NVIDIA. Their LinkedIn profiles now list NVIDIA as their employer. Neither company has publicly announced the acquisition, and NVIDIA declined to comment to Fortune.

Founded in 2022, Kumo raised approximately $37 million from investors including Sequoia Capital. The company built foundation models designed to generate business predictions directly from enterprise data. Customers cited in public reporting include DoorDash, Reddit, Sainsbury's, and Black Entertainment Television (BET).

Kumo Focused on Enterprise Prediction

Kumo positioned itself as an alternative to traditional predictive analytics systems that often require extensive data engineering and model development. According to the company, customers can connect the platform directly to enterprise data warehouses and generate predictions related to customer churn, demand forecasting, fraud detection, and credit risk.

The company's technology is built around Graph Neural Networks (GNNs), a machine learning approach that models relationships between entities such as customers, products, transactions, and services. Kumo argues that understanding these connections can improve predictive accuracy compared with models that analyze records independently.

In an earlier interview with Fortune, Leskovec described the company's approach as allowing organizations to point the model at their data, define an outcome, and receive a prediction without additional model training.

The technology reflects a broader push across enterprise AI toward systems that operate directly on business data. As enterprises move AI initiatives into production environments, spending has increasingly shifted toward operational infrastructure and deployment platforms.

Another Addition to NVIDIA's AI Stack

The acquisition continues NVIDIA's effort to expand beyond chips into software, infrastructure, and enterprise AI applications.

In April 2024, NVIDIA completed its acquisition of Run, a company focused on GPU orchestration and workload management. More recently, it acquired data semantics startup Illumex. The company has also pursued investments and acquisitions across AI infrastructure, enterprise software, and industry-specific deployments.

Kumo adds technology that sits closer to business decision-making than many of NVIDIA's previous acquisitions. Rather than helping train or deploy models, Kumo's software generates predictions from enterprise datasets that organizations can use for customer retention, forecasting, and risk management.

The deal also brings a team with deep experience in machine learning and large-scale data systems. Before co-founding Kumo, Josifovski held leadership roles at Google, Pinterest, and Airbnb. Leskovec is a Stanford University professor known for his work in graph machine learning, while Raghavan previously held senior roles at IBM, LinkedIn, and Yahoo.

The acquisition aligns with NVIDIA's broader effort to build capabilities across the AI stack, from infrastructure and model development to enterprise applications. That strategy has helped position the company as a supplier of AI systems that extend beyond hardware into software and operational workflows.