Key Highlights
- Chip giant Nvidia completed the acquisition of enterprise AI company Kumo AI in a deal exceeding $400 million
- An Nvidia executive initially revealed the transaction through a LinkedIn post
- Kumo specializes in predictive AI systems that leverage structured corporate data, serving clients including DoorDash, Reddit, and Sainsbury’s
- All three Kumo co-founders have transitioned to roles at Nvidia
- Shares of NVDA dropped 1% Thursday after the acquisition became public
The chip giant Nvidia has completed the purchase of Kumo AI, an enterprise artificial intelligence firm established four years ago in Mountain View, for a sum surpassing $400 million, as reported by The Information. The transaction came to light after an Nvidia executive shared details on the professional networking platform LinkedIn.
The three individuals who co-founded Kumo — Vanja Josifovski, Jure Leskovec, and Hema Raghavan — have already transitioned into their new roles at Nvidia. Their updated LinkedIn profiles confirm their employment status at the semiconductor company, while Kumo’s corporate website remains unchanged regarding the ownership transfer.
Shares of NVDA declined 1% during Thursday’s trading session after the acquisition news emerged.
Kumo AI was established in 2022 by a team of researchers with connections to Stanford University. Prior to being acquired, the company had secured $37 million through venture capital funding, with Sequoia Capital among its notable backers.
Understanding Kumo’s Technology
Kumo specializes in developing artificial intelligence systems engineered to address predictive business inquiries by analyzing structured enterprise information — including customer databases and transaction records. This represents an area where conventional large language models typically face limitations.
The company developed a solution by merging synthetic data generation methods with graph-based machine learning innovations. Its AI systems can generate predictions for scenarios such as customer attrition, credit default probability, and comparable business use cases without requiring retraining for each new application.
“With the foundation model, you point it to your data, you define what you mean by churn, and a second later, you get the prediction,” co-founder Jure Leskovec said previously.
Kumo released its latest offering, KumoRFM-2, in April 2026. The company’s client roster and partnership network has featured prominent names such as DoorDash, Reddit, Databricks, and Snowflake.
Latest Addition to Nvidia’s Growing Portfolio
This transaction represents another chapter in an ongoing strategy. Nvidia has completed acquisitions of over 100 startups in recent years as part of its effort to construct a comprehensive AI technology stack.
Other notable recent transactions include a $20 billion talent and asset acquisition from AI inference provider Groq in December 2025, the purchase of data semantics specialist Illumex in February 2026, and a $700 million agreement to acquire orchestration software provider Run.ai in April 2024.
The specific integration plans for Kumo’s technology platform remain undisclosed. According to The Information, potential applications include incorporating Kumo’s models into Nvidia’s AI foundry software suite or leveraging Kumo’s research talent to create additional enterprise-focused foundation models.
This acquisition broadens Nvidia’s collection of AI models designed to work optimally with its hardware infrastructure while providing enterprises with enhanced capabilities for building and tailoring models using its technology platform.


