If you have been paying attention to the AI landscape beyond chatbots and image generators, you already know that the next frontier is physical. Robots assembling cars. Drones inspecting bridges. Autonomous trucks crossing state lines. All of these systems depend on massive streams of video, sensor, and multimodal data, and none of them can function without infrastructure purpose-built to manage that data at production scale.

That is exactly where Encord has positioned itself. The London and San Francisco-based company just closed a $60 million Series C led by Wellington Management, bringing total funding to $110 million. The round included existing backers Y Combinator, CRV, N47, and Crane Venture Partners, alongside new investors Bright Pixel and Isomer Capital.

Co-founded by CEO Eric Landau and Danish-born Ulrik Stig Hansen, Encord has deep Nordic roots despite its London headquarters. The company's trajectory from Y Combinator graduate to $110 million in total funding makes it one of the most significant AI infrastructure plays to emerge from the European startup ecosystem.

The numbers behind Encord's growth are striking. Over the past year, the company's managed data volume surged from roughly one petabyte to more than five petabytes. Revenue from physical AI customers grew tenfold. Those are not vanity metrics. They reflect a genuine inflection in demand as AI systems move from controlled lab environments into messy, unpredictable real-world settings.

Encord's platform addresses three core functions that every physical AI deployment requires. First, data curation to identify the training inputs that actually matter from oceans of raw sensor footage. Second, data management to index, trace, and version datasets across distributed storage systems. Third, annotation and alignment to ensure models are trained and evaluated against specific performance benchmarks.

If that sounds like a data engineering problem rather than an AI research problem, that is precisely the point. The bottleneck for physical AI is not model architecture. It is the plumbing underneath, the infrastructure that ensures the right data reaches the right model at the right time, with full traceability from raw input to production output.

Encord's customer roster reads like a who's who of companies building AI systems that operate in the physical world. Woven by Toyota, the autonomous driving subsidiary of the Japanese automaker, uses the platform. So does Skydio, the leading US drone manufacturer, and Synthesia, the AI video generation company. Each of these customers works with massive volumes of multimodal data that legacy enterprise tools simply cannot handle at production scale.

The common thread is complexity. Autonomous vehicle developers need to annotate millions of video frames with pixel-level precision. Drone operators must manage sensor data from cameras, LiDAR, and thermal imaging simultaneously. AI video companies require alignment between generated outputs and training inputs across thousands of scenarios. Encord's platform was built from the ground up for these workloads.

Having Wellington Management lead a Series C is a significant signal. Wellington is one of the world's largest independent investment management firms, with more than $1 trillion in assets under management. When a firm of that scale leads a growth round, it typically indicates that the due diligence went deep into unit economics, customer retention, and competitive moats rather than stopping at top-line growth metrics.

The participation of existing investors like Y Combinator, CRV, and Crane Venture Partners suggests the cap table is clean and the insiders are confident. New entrants Bright Pixel and Isomer Capital add European strategic depth. This is not a round assembled from scattered checks. It is a concentrated bet by sophisticated capital allocators.

Encord's most obvious competitive benchmark is Scale AI, the San Francisco-based data labeling giant valued at over $13 billion. Scale AI has built its business primarily on government and defense contracts alongside large language model training. Encord is targeting a different wedge: the multimodal, sensor-heavy data workloads that define physical AI.

That distinction matters more than it might appear. Physical AI data is fundamentally different from text or web-scraped image datasets. It is sequential, multimodal, and often safety-critical. A mislabeled frame in an autonomous driving dataset is not just an inconvenience. It is a potential liability. Encord's platform is designed around this reality, with built-in traceability, version control, and quality assurance workflows.

The $60 million will accelerate development of what Encord describes as a universal data layer, a single infrastructure backbone that can handle curation, management, and annotation for any physical AI use case. If the company executes on that vision, it could become as essential to robotics and autonomy companies as cloud providers are to SaaS.

While Encord is headquartered in London, its Nordic connection through co-founder Ulrik Stig Hansen places it squarely in the orbit of the Scandinavian AI ecosystem. The company joins a growing roster of Nordic-linked AI companies that are building foundational infrastructure rather than end-user applications.

For the broader physical AI market, Encord's trajectory is a signal of maturation. When early-stage startups dominate a category, it means the market is speculative. When Series C rounds of this size start flowing into infrastructure layers, it means production workloads are real, customers are paying, and the scaling phase has begun. That is where physical AI stands today, and Encord is positioned to be the data layer that holds it all together.