James Wu, the former vice president of NVIDIA's DRIVE Sim and DeepMap business, has launched a new startup called Calder that targets a critical gap in robotics development: the scarcity of high-quality training data for embodied AI systems. Wu's move signals where the next wave of autonomous vehicle and robotics innovation may depend less on algorithms and more on access to real-world data that teaches machines how to understand and navigate physical spaces .

What Is Spatial Intelligence and Why Does It Need So Much Data?

Spatial intelligence refers to a machine's ability to understand, reason, perceive, and act in the three-dimensional physical world. Unlike large language models (LLMs), which are AI systems trained on text data from the internet, spatial intelligence requires machines to learn from embodied experiences: how robots move through spaces, interact with objects, and respond to real-world conditions .

The challenge is stark. Fei-Fei Li, a leading AI researcher at Stanford University and founder of World Labs, has emphasized that robotics research faces a fundamental data problem. "Robotics research is still in its early stage, and there is really a lack of data," Li explained, noting that the field "basically has no commercial application scenarios, especially in daily use. So it is difficult to collect data" .

This data scarcity has become a consensus issue across the industry. Wang Xingxing, founder of Unitree Robotics, identified "the dependence on real machine data and the scarcity of real robot data" as one of three major obstacles restricting embodied intelligence development .

How Is the Industry Addressing the Data Collection Challenge?

• Corporate Data Centers: JD.com announced plans to build the world's largest embodied intelligence data collection center, leveraging its supply chain and business operations across retail, logistics, health, industry, takeout, and housekeeping to accumulate over 10 million hours of high-quality data within two years .
• Specialized Data Infrastructure Companies: Ropedia, a data infrastructure provider established in 2025, completed seed-stage funding worth tens of millions of dollars and focuses on providing data collection solutions for robotics, space, and physical intelligence applications .
• Wearable Collection Devices: Qianxun Intelligence announced 1 billion yuan in new funding and has iterated its wearable data collection device to a fifth generation, reducing collection costs to one-tenth of traditional methods .

Who Is James Wu and Why Does His Move Matter?

Wu's career trajectory reveals why he is positioned to tackle this problem. In 2016, he founded DeepMap, a high-precision mapping company that built mapping platforms for autonomous vehicles and smart cities. The company attracted investment from venture firms including a16z and GSR Ventures, as well as from NVIDIA itself, and secured customers like Ford, Honda, and Bosch .

When NVIDIA acquired DeepMap in 2021, the company stated that "DeepMap's technology will enhance the mapping and positioning functions on NVIDIA DRIVE, ensuring that autonomous vehicles always know exactly where they are and where they are going." NVIDIA DRIVE is NVIDIA's full-stack autonomous vehicle platform and a cornerstone of the company's bet on intelligent driving .

Wu remained at NVIDIA as vice president and general manager of the DRIVE Sim and DeepMap business until December 2025, when he departed to launch Calder. His new venture targets spatial data services, a broader market than mapping alone, though both depend fundamentally on collecting and delivering high-quality data .

Before founding DeepMap, Wu worked at Google, Apple, Baidu USA Research Institute, and the personal cloud storage company Upthere, giving him deep experience with digital mapping businesses including Google Earth, Google Maps, Apple Maps, and Baidu's autonomous driving initiatives .

What Does Calder Actually Do?

Calder positions itself as "the foundation of spatial intelligence" and describes its mission as providing embodied intelligent spatial data services. The company's name references sculptor Alexander Calder, known as the "pioneer of kinetic sculpture" for his work integrating movement and environmental interaction, a fitting metaphor for a company focused on teaching machines to move and interact with physical spaces .

Currently, Calder is in active recruitment and working with a small number of laboratories and robot teams. The company has not yet released detailed information about its specific services or business model, though Wu's background suggests it will focus on collecting, curating, and delivering high-quality spatial data to robotics and autonomous vehicle developers .

How Large Is the Market Opportunity?

The embodied intelligence data collection market is experiencing explosive growth. According to market research firm QYResearch, the global market size was approximately $753 million in 2024 and is projected to reach $6.752 billion by 2031, representing a compound annual growth rate of 36.8% from 2025 to 2031 .

This growth reflects a fundamental shift in how the AI industry views the path to advanced robotics. As data becomes the constraint that determines how intelligent robots can become, both venture capital and established tech companies are racing to secure access to high-quality training datasets. For companies like Calder, the window to establish dominance in spatial data collection and delivery may be narrow but lucrative .

Wu's departure from NVIDIA and launch of Calder illustrates how the autonomous vehicle and robotics industries are evolving. While NVIDIA DRIVE remains a critical platform for autonomous vehicle development, the real competitive advantage may increasingly depend on who controls the highest-quality data for training spatial intelligence systems. Wu's new venture positions him to compete directly in that emerging market.