deepset raises €30m Series B to expand LLM platform for enterprises

Berlin-based deepset has raised €30 million in Series B funding led by Balderton Capital, marking a significant milestone for a European company building critical infrastructure for enterprise AI. Google Ventures, Harpoon Ventures, System.One, and Lunar Ventures returned as investors in the round, bringing the company's total capital raised to €46 million.

The funding validates a shift underway in how large organisations approach artificial intelligence. Instead of experimenting with large language models in isolated teams, enterprises increasingly want to move AI projects into production at scale. Deepset's platform sits at that inflection point.

Deepset was founded in 2018 by Milos Rusic, Malte Pietsch, and Timo Möller. The trio initially bootstrapped the company by building custom natural language processing solutions for major clients. In 2019, they released Haystack, an open-source framework that gave other developers tools to build LLM-driven applications. The commercial product, the Haystack Enterprise Platform (formerly called deepset Cloud), launched to handle the full lifecycle of production AI systems: experimentation, deployment, and ongoing monitoring.

From open source to enterprise control

Haystack has become a critical component in the infrastructure stack for building AI applications. The framework lets teams assemble complex systems from modular components: different language models, vector databases, file converters, text embeddings. This avoids building everything from scratch. Airbus, for example, uses Haystack to build systems that recommend aircraft operations guidelines to pilots. The Economist and OakNorth Bank are also among customers using deepset's platform.

The enterprise platform extends Haystack by giving teams a unified interface to compare different language models, run experiments, scale infrastructure automatically, and monitor model performance. Unlike many AI platforms, deepset emphasises data sovereignty and privacy. The product is SOC 2-certified and includes functionality to run in customers' own virtual private cloud environments, which is essential for regulated industries dealing with sensitive information.

CEO Milos Rusic described the market opportunity in terms of an industry shift. "The industry is shifting from AI labs to AI factories," Rusic said, highlighting how data science teams are increasingly expected to support product teams trying to ship AI features. The pressure is real, as many organisations find that building and managing production AI systems requires skills most internal teams do not have.

What the new money funds

The €30 million will go toward three main areas. First, deepset is refining the platform specifically for retrieval-augmented generation (RAG), a technique where language models pull in external data to answer questions more accurately. The company is particularly focused on improving how the platform evaluates each component in a RAG pipeline, allowing teams to identify where problems occur.

Second, deepset is optimizing for customers with strict privacy requirements by improving support for virtual private cloud deployments, letting enterprises run the full platform within their own infrastructure without data leaving their systems.

Third, deepset is expanding LLM observability (the ability to monitor model performance in production). Language models drift over time and behave unexpectedly with real-world data, so organisations need visibility into what is happening. Hallucination, where models confidently state false information, remains difficult to predict and prevent. Better observability and detection are top priorities for enterprises running language models in production.

At the time of the announcement, deepset employed 50 people. The company planned to grow the team significantly, targeting 70 to 75 employees by the end of the year.

A European infrastructure play

Deepset's funding is notable in another context. European AI startups have struggled to raise large rounds compared with their US counterparts, particularly in infrastructure and tooling. Balderton Capital, a London-based venture firm, led this round, a sign that European investors recognise the value in the AI infrastructure category.

The company remains deeply committed to Haystack, its open-source framework. This dual approach—maintaining an open framework while building a commercial platform—is becoming a common pattern in enterprise software. It lets developers experiment with the free tool, building familiarity that eventually drives adoption of the paid platform.

The total addressable market for MLOps platforms, the broader category deepset competes in, is forecast to reach €18 billion by 2031, up from roughly €600 million in 2021. Deepset faces competition from large cloud providers like AWS and Google Cloud as well as specialised startups including Seldon, Arize, and Tecton. However, the emphasis on data sovereignty, modularity, and developer experience has carved out a distinct position.

For deepset's founders and investors, the next challenge is translating the company's strong traction with enterprise customers into sustainable growth. The platform needs to prove it can scale across regulated industries (finance, healthcare, legal, insurance) where data privacy requirements are non-negotiable and vendor lock-in is a serious concern.