What recent R&D moves signal a step-change in robotics capabilities?

NVIDIA’s Project GR00T pushes foundation models for humanoids, ABB’s acquisition of Sevensense brings AI-powered 3D vision to mobile robots, and Figure AI’s $675 million raise accelerates language-guided manipulation research. Together, these advances close the gap between perception-heavy lab work and reliable, factory-ready systems.

How are enterprise buyers approaching robotics development in 2025?

Manufacturers are adopting a buy-vs-build strategy: purchasing mature perception and navigation stacks from vendors like NVIDIA and ABB while developing domain-specific policies in-house. This approach reduces integration time and concentrates internal R&D on the tasks that differentiate their operations.

What role do foundation models play in robotics R&D?

Foundation models provide reusable representations for vision, language, and control that can be fine-tuned for specific robot tasks. By leveraging large-scale pretraining and simulation, teams can transfer skills across robots and environments, shortening data collection cycles and improving robustness.

Which safety and regulatory issues are shaping robotics research roadmaps?

The EU AI Act will classify certain robotics applications as high-risk, requiring documentation, monitoring, and data governance that must be considered during development. In the U.S., OSHA guidance and evolving cobot standards influence how startups and incumbents validate human-robot interaction before pilots and scale-up.

Where are the most promising near-term applications for these R&D advances?

Warehouse mobility and piece-picking, flexible electronics assembly, and inspection tasks are seeing the fastest translation from research to production. Pilots by Amazon with mobile and humanoid platforms, plus ABB’s vision-enabled AMRs, indicate measurable gains in throughput and uptime without sacrificing safety.

Robotics R&D Surges as NVIDIA Unveils GR00T, ABB Buys Sevensense, Figure AI Raises $675M

A fresh wave of robotics research is moving from lab demos to factory floors as NVIDIA pushes foundation models for humanoids, ABB folds 3D vision into its mobile robots, and Figure AI secures a blockbuster round. The stakes: faster deployment, safer collaboration, and new software revenue streams riding on smarter perception and planning.

Published: November 23, 2025 By Marcus Rodriguez Category: Robotics

Robotics R&D Surges as NVIDIA Unveils GR00T, ABB Buys Sevensense, Figure AI Raises $675M

Lab Breakthroughs Start Shipping

Industrial robotics is hitting a new R&D stride, with software-heavy advances in perception and planning finally clearing the path to deployment at scale. Global robot installations reached a record 553,000 units in the latest reporting year, according to the International Federation of Robotics, underscoring a demand pull that is reshaping research agendas inside engineering groups and corporate venture arms alike.

NVIDIA put a stake in the ground with Project GR00T, a generative model initiative aimed at teaching humanoid robots complex tasks via multimodal learning. For more on related robotics developments. The effort, revealed at GTC, positions the chipmaker to sell not just GPUs but full-stack robotics software through its Isaac platform, as widely covered in the press following GTC announcements. In parallel, ABB agreed to acquire Zurich-based Sevensense to integrate AI-powered visual navigation into its autonomous mobile robots, expanding its software IP and moving perception R&D closer to the warehouse and shop floor, Reuters reported.

Capital is chasing these shifts. Figure AI closed a $675 million round backed by a who’s who of strategic investors to accelerate a general-purpose humanoid capable of language-guided tasks, TechCrunch reported. The money is earmarked for scaled data collection, teleoperation training, and safety validation—areas that have historically limited robotics R&D from translating into durable commercial wins.

Foundation Models Meet the Assembly Line

The next frontier in robotics research is model-centric: large, pre-trained policies that can be specialized for manipulation, navigation, and inspection. Microsoft has pushed Azure edge services deeper into robotic controllers and simulators to support data pipelines for iterative policy training, while OpenAI is working with hardware partners on language-to-action interfaces for embodied agents. Early pilots suggest that codifying tacit knowledge from skilled operators into reusable policies can cut integration times by months.

Tesla...

Read the full article at AI BUSINESS 2.0 NEWS