Nvidia invests 400 million USD to acquire Kumo AI! Completing the "enterprise prediction model" puzzle and accelerating the transformation into a full-stack AI giant

According to foreign media TechStartups in a report on June 4, 2026, chip giant NVIDIA continues to expand its software footprint, with reports stating that it has spent more than $400 million to acquire the predictive AI startup Kumo AI. The company has a star-studded Silicon Valley founding team and focuses on enterprise structured data prediction models that can be applied without cumbersome data preparation. This move will significantly strengthen NVIDIA’s capability to address real enterprise business pain points and positively take on major cloud giants.
(Background: AI stock guru Serenity: SIVE targets the “NVIDIA optical upstream bottleneck,” after Ayar Labs announced that it has joined NVLink Fusion.)
(Additional background: Intel: The AI chips we will launch will be cheaper than NVIDIA and AMD, and Crescent Island focuses on inference with air cooling.)

As artificial intelligence (AI) technology moves from concept validation to practical commercial applications, chip leader NVIDIA is accelerating the expansion of its software infrastructure footprint. According to TechStartups, NVIDIA has spent more than $400 million to acquire Kumo AI, a predictive AI startup founded only about four years ago, at a high price. Insider sources say that Kumo AI’s three core founders had quietly joined NVIDIA before last month, showing that this strategic acquisition has entered a real integration phase.

A star-studded team is on board, specializing in structured data prediction

Founded in 2022, Kumo AI is headquartered in Mountain View, California, and its founding team background is extremely impressive. CEO Vanja Josifovski previously served as Chief Technology Officer at Airbnb and Pinterest; engineering lead Hema Raghavan was the AI business lead at LinkedIn; and chief scientist Jure Leskovec is a professor at Stanford University and an internationally renowned expert in graph neural networks (GNN).

Different from the text-generation AI widely discussed in the market, Kumo AI’s core technology lies in developing foundational models specifically tailored for enterprise “Relational Data.” Enterprise customers do not need to spend a large amount of time and money on data preparation or feature engineering—they can generate predictions directly from their data warehouses. Its flagship product, KumoRFM, adopts a SQL-like “Predictive Query Language,” and is widely used in key business scenarios such as customer churn prediction, fraud detection, demand forecasting, and credit risk analysis. It is currently adopted by large enterprises such as DoorDash, Reddit, and the UK supermarket chain Sainsbury’s.

Completing the software ecosystem, NVIDIA targets a full-stack AI platform

This premium acquisition (Kumo’s valuation before the acquisition was about $250 million) is of profound strategic significance to NVIDIA. Although NVIDIA holds absolute dominance in the GPU hardware market, large enterprise customers care more about end-to-end solutions that bring measurable return on investment (ROI). Kumo AI’s technology precisely fills NVIDIA’s technical gap in the structured data business prediction domain.

NVIDIA is actively pushing its “full-stack AI platform” strategy, hoping that when customers purchase hardware, they can also fully adopt its software stack. In the future, Kumo’s predictive models are expected to be deeply integrated and optimized with infrastructure such as NVIDIA’s inference microservices (NIM), greatly lowering the barrier for enterprises to adopt the technology. Combined with NVIDIA’s recent consecutive acquisitions of the data semantics company Illumex and its spending of about $700 million to acquire AI coordination firm Run, NVIDIA is actively equipping its enterprise-level software ecosystem to compete with cloud giants such as Microsoft (Microsoft), Google (Google), and Amazon (Amazon) in the software services arena.