AI infrastructure giant Nebius buys agentic search startup Tavily – News

Artificial intelligence infrastructure company Nebius Group N.V. said today it’s going to acquire an Israeli startup called Tavily Inc., which has developed a specialized search layer for autonomous AI agents.

The terms of the deal were not disclosed, but Calcalist reported that Nebius has agreed to pay $275 million initially, with the amount potentially rising to $400 million if certain milestones are achieved.

It’s an intriguing deal that underscores the growing significance of specialized AI infrastructure and the rapid evolution of autonomous AI agents. Tavily, founded in 2024, says it’s building the search infrastructure needed by AI agents to access real-time data and use it to reason and take actions. It enables large language models to retrieve accurate, up-to-date and contextually relevant data from both private and public sources.

Tavily’s platform acts as an interface between AI systems and the web, allowing developers to integrate powerful search capabilities into their applications easily. It does so via a developer-friendly application programming interface that combines high-performance search, intelligent crawling and structured web page extraction.

That differs from traditional search APIs because it’s designed specifically for AI use cases, injecting clean, structured data directly into the context windows of large language models. As a result, Tavily drastically reduces the likelihood of hallucinations and outdated answers, which are key hurdles in production-grade AI systems.

Although it’s still a relatively young company, Tavily claims to be generating significant revenue already from a broad customer base that includes a mix of fast-growing AI startups and Fortune 500 enterprises. Some of its most notable users include Cohere Inc., MongoDB Inc., Groq Inc., IBM Corp., Monday.com Ltd., LangChain Inc. and Amazon Web Services Inc.

Tavily has done this with minimal funding. Most recently, it raised $20 million in a Series A funding round in August, bringing its total funding to $25 million.

As for Nebius, it’s a fast-growing data center infrastructure company that spun out of the Russian search engine giant Yandex N.V. in February 2023 in order to focus on providing cloud compute for AI workloads. It sold the search engine business but kept its data centers, and has since expanded its infrastructure massively. It makes money by renting access to premium AI processors such as Nvidia Corp.’s graphics processing units.

But Nebius wants to do more than just provide the infrastructure for AI. It’s aiming to build a unified platform for companies to build, tune and run AI agents, and that explains why it’s interested in Tavily. It’s planning to integrate the startup’s agentic search tools into its platform so it can give developers everything they need to build and operate enterprise-grade AI systems. “We’re not just an infrastructure-as-a-service company — we’re building the complete platform for anyone who wants to build AI products, agents, or services,” said Nebius co-founder and Chief Business Officer Roman Chernin.

With Tavily’s capabilities, Nebius can solve one of the major limitations of AI agents. Advanced reasoning is not enough for business automation. In addition, AI agents must also be able to access verifiable information in real-time. Without any live search capabilities, agents cannot generate accurate insights or take actions when conditions change.

By acquiring Tavily, Nebius is bundling a key capability that many developers have previously had to stitch together using tools from separate vendors. Clearly, Nebius doesn’t view agentic search as something that should just be bolted on, but a core element of AI agent infrastructure.

“Tavily is solving a critical part of this stack with agentic search and has proven it with strong developer adoption,” Chernin said. “This acquisition brings the search layer directly into our stack, so developers can focus on their applications instead of managing multiple vendors.”

Image: Tavily