Nvidia has made another strategic software bet, paying over $400 million to acquire Kumo AI, a predictive AI startup focused on enterprise structured data. The acquisition — first reported by TechStartups on June 4, 2026 — represents a premium over Kumo's pre-deal valuation of roughly $250 million. Three co-founders have already joined Nvidia, signaling swift integration.
A Stellar Team Targeting Structured Data
Founded in 2022 and based in Mountain View, Kumo AI brings a founding team with deep Silicon Valley roots: CEO Vanja Josifovski (ex-Airbnb/Pinterest CTO), engineering head Hema Raghavan (ex-LinkedIn AI lead), and chief scientist Jure Leskovec (Stanford professor and GNN authority). The startup's core offer is predictive models for relational data — churn prediction, fraud detection, demand forecasting, and credit risk — without requiring data prep or feature engineering.
Its flagship product, KumoRFM, uses a Predictive Query Language similar to SQL. Clients include DoorDash, Reddit, and UK supermarket chain Sainsbury’s.
Filling a Gap in Nvidia's Enterprise Stack
The buyout is the latest in Nvidia's push to build a full-stack AI platform, following acquisitions of data semantics firm Illumex and AI orchestration company Run (the latter for ~$700 million). Kumo's technology will likely integrate with Nvidia's NIM inference microservices, making enterprise adoption smoother. As cloud giants Microsoft, Google, and Amazon dominate the software layer, Nvidia aims to offer customers a complete hardware-plus-software package — and structured data prediction was a missing puzzle piece.
Industry analysts note that predictive analytics on structured data accounts for a large chunk of corporate IT spending. Nvidia's move signals its ambition to compete not just on chips but on end-to-end enterprise AI solutions.