AI is moving fast, but the data pipelines feeding it are not. That is exactly the gap Encord is closing, and on February 26, 2026, the company announced a $60 million Series C round to scale what it calls an AI data infrastructure platform. The round brings Encord’s total funding to $110 million and signals growing investor conviction around data quality infrastructure for physical AI.
Wellington Management led the round, with participation from existing backers Y Combinator, CRV, N47, and Crane Venture Partners, alongside new investors Bright Pixel and Isomer Capital.
The Core Problem:
Most AI companies hit a wall they did not expect. Not compute, not model architecture, data. Specifically, the messy, unstructured, multi-format reality of real-world data that legacy enterprise tools were never built to handle.
Training a large language model on internet text is relatively straightforward. But training an autonomous vehicle, a delivery drone, or a humanoid robot on video feeds, LiDAR scans, and sensor data across thousands of edge cases, that requires a completely different class of tooling.
According to Encord’s co-CEOs, Ulrik Stig Hansen and Eric Landau, traditional enterprise infrastructure was built for cloud data management, not for the specific needs of AI model training, curation, and alignment. The result is that AI teams hit data quality issues precisely when they are preparing to go to production.
What the Platform Does:
Encord positions itself as a universal data layer that supports AI teams across the full model lifecycle. The platform handles four distinct functions.
Data curation helps teams identify and filter the most relevant training data from massive, messy datasets. Data management and orchestration gives teams a complete view of their data, indexed, integrated with storage, and traceable. Annotation and labeling covers the task of structuring training data across modalities including image, video, audio, document, LiDAR, and medical formats like DICOM. Post-training alignment handles model evaluation and the process of getting human feedback incorporated at scale.
The platform works across physical AI use cases as well as frontier and generative AI workflows. Customers include Woven by Toyota, Skydio, and Synthesia.
The Numbers Behind the Round:
Growth in the last twelve months gives the Series C context. Encord’s platform scaled from 1 petabyte to over 5 petabytes of data managed, a 5x increase in one year. Revenue from physical AI customers grew 10x over the same period.
Those figures reflect a shift in where AI investment is landing. Physical AI, robots, drones, autonomous systems, demands multimodal data at scale and in near real time. That is a fundamentally harder data problem than what most current infrastructure was built for.
Analysts project hundreds of millions of AI robots coming online in the next few years. Each one requires high-quality, continuously updated training data to perform reliably in the real world.
Why Data Infrastructure is Having its Moment:
There is a useful analogy in software history. When cloud computing scaled, a whole category of tools emerged to manage, secure, and orchestrate cloud data. AI is following a similar path. The model layer got most of the early attention, but the data layer is now catching up.
For teams building production AI today, the challenge is rarely “which model do I use”, it is “how do I make sure my model is trained on clean, labeled, representative data at every iteration.” If you are working on AI data annotation tools or multimodal data pipelines, Encord’s platform is built for exactly that workflow.
This also connects to broader conversations around AI reliability and governance. Poor training data, even a small number of malicious or mislabeled data points, can significantly degrade model performance. Encord’s platform is designed to address that throughout the development lifecycle, not just at the annotation stage.
What this Funding Enables:
The $60 million will go toward accelerating what Encord describes as a universal data layer for production-ready AI. That means expanding platform capabilities, scaling infrastructure to match customer growth, and supporting a larger team across its San Francisco and London offices.
The company is actively hiring across all teams. For AI engineers, data infrastructure builders, and ML ops practitioners, Encord’s growth trajectory makes it a relevant company to track, both as a platform and as an employer.
What to Keep in Mind:
For founders and operators building AI products, the Encord story highlights something practical: data infrastructure is not a post-launch problem. Encord’s own experience shows that AI teams often encounter data quality issues at the exact moment they are preparing to go to production, making investment in purpose-built data tooling a core part of the development process.
The companies already using Encord, including physical AI players in automotive and drone sectors, illustrate what high-quality AI data pipelines enable at scale. As physical AI applications expand, demand for AI-native data infrastructure will follow.
Encord’s Series C is a signal that this category is maturing, and the companies building foundational data layers for AI are attracting serious capital.
