Encord Raises €50M for 'Physical AI' Data Layer
London-based Encord has closed a €50M Series C to build the data infrastructure layer for "physical AI." The funding, led by Wellington Management, is aimed at enabling autonomous systems in robotics and IoT, signaling that data supply chains for physical-world agents are becoming as strategic as model APIs.
This latest funding round brings Encord's total capital raised to approximately €93 million ($110 million). The company's valuation now stands at $550 million, signaling strong investor confidence in the growing market for physical AI data infrastructure. The round saw participation from existing investors like Y Combinator and CRV, as well as new backers Bright Pixel Capital and Isomer Capital. Encord's platform is engineered to handle the entire AI data lifecycle, from ingestion and curation to annotation and evaluation, for the complex multimodal data that powers autonomous systems. This includes video, LiDAR, sensor data, and 3D point clouds, which are notoriously difficult for legacy data tools to manage. The company has seen a tenfold increase in revenue from its physical AI customers and a surge in data managed on its platform from 1 to over 5 petabytes in the last year. For developers, Encord provides an API and SDK-first approach, allowing for seamless integration into existing MLOps workflows and the automation of data pipelines. This is critical for teams building agentic AI systems that require continuous learning and adaptation based on real-world feedback. The platform's architecture supports the creation of "data flywheels," where production data is used to improve models over time. In regulated industries such as automotive and healthcare, AI governance and compliance are paramount. Encord's platform is SOC 2 and GDPR compliant, addressing the need for data security and traceability in enterprise AI adoption. The platform creates a crucial audit trail for AI model development, which is essential for debugging and ensuring model reliability in safety-critical applications. The investment from Wellington Management highlights a broader venture and startup ecosystem trend: the increasing strategic importance of the data layer in the AI stack. As AI moves from digital applications to physical-world agents, the infrastructure to manage proprietary, real-world data becomes a key competitive differentiator. Encord's focus on this niche positions it as a foundational technology for the next wave of AI-powered systems. Encord's client roster includes over 300 AI teams, with prominent names like Woven by Toyota, Zipline, Skydio, and AXA Financial. These enterprise AI adoption case studies demonstrate the platform's ability to scale and meet the demands of companies at the forefront of autonomous systems development. The successful deployment in these organizations underscores the critical need for specialized data infrastructure in the development of physical AI. The challenge for "physical AI" is less about model size and more about "data readiness," as Encord's co-founder and co-CEO Ulrik Stig Hansen has emphasized. Inconsistent or incomplete data is a primary cause of failure for models operating in the real world. Encord's platform directly addresses this by providing tools for data curation, quality control, and the identification of edge cases. Looking ahead, the market for physical AI is projected to grow substantially, with some analysts predicting over 400 million AI-powered robots online in the coming years. This trend, coupled with the increasing complexity of AI models, will drive further demand for robust data infrastructure. Encord's strategic focus on this area, backed by significant venture funding, positions it to be a key enabler of this technological shift.
