The robotics industry is entering a new phase where success depends less on building impressive prototypes and more on solving the messy, unpredictable challenges of real-world deployment. Two major developments this week underscore this shift: Fujitsu and Carnegie Mellon University have launched a joint Physical AI Research Center, while Pudu Robotics secured nearly $150 million in funding to accelerate its embodied AI (AI systems embedded in physical robots) technologies across manufacturing, logistics, and service sectors.

Physical AI represents a fundamental departure from traditional robotics. Rather than relying on pre-programmed instructions or controlled environments, physical AI systems perceive their surroundings, reason about what they observe, and adapt their behavior in real time. This capability is reshaping what robots can accomplish and where they can operate.

What's Driving the Academic-Industry Partnership Model?

The Fujitsu-Carnegie Mellon partnership reflects a growing recognition that scaling physical AI requires bridging the gap between fundamental research and commercial reality. Carnegie Mellon's Robotics Innovation Center, a 14,000-square-meter facility in Pittsburgh, provides the infrastructure to test robotics and AI systems in realistic operating environments, not sterile labs.

The center will bring together researchers and engineers across multiple disciplines to tackle persistent challenges in physical AI deployment. These research priorities include action generation and learning, spatial perception, multi-robot coordination, human-robot collaboration, and integration between simulation and real-world environments.

"At this research center, Fujitsu will create new value through the convergence of AI, computing, networking, and robotics, and accelerate the societal implementation of reliable physical AI," said Vivek Mahajan, Corporate Executive Officer and Corporate Vice President, CTO in charge of System Platform at Fujitsu Limited.

Vivek Mahajan, Corporate Executive Officer and Corporate Vice President, CTO, Fujitsu Limited

The technologies developed through this collaboration will feed into Fujitsu's Kozuchi Physical OS platform, a software system designed to coordinate robots, sensors, and systems across physical environments. Initial integration of research outputs into the platform is expected to begin in fiscal year 2026.

How Are Companies Scaling Physical AI Beyond Research?

While Fujitsu focuses on the research foundation, Pudu Robotics is demonstrating what commercial-scale deployment looks like. The Shenzhen-based company raised nearly $150 million in its latest funding round, pushing its valuation above $1.5 billion and bringing cumulative funding to over $300 million.

Pudu's growth trajectory reveals the market momentum behind embodied AI. The company achieved a 100% year-over-year revenue surge in 2025, with its commercial cleaning segment now representing over 70% of total revenue. More tellingly, Pudu's industrial delivery robots have seen rapid adoption, with over 4,000 units shipped within just one year of their market launch.

The company's success stems from a diversified approach across multiple sectors and use cases. Pudu's solutions are deployed across industries including retail, hospitality, manufacturing, food and beverage, healthcare, and education. This multi-sector strategy contrasts with competitors focused on single applications like humanoid manufacturing or warehouse automation.

Steps to Unlock Physical AI's Potential in Your Organization

• Start with confidence-building use cases: Rather than attempting to deploy robots in highly unstructured environments immediately, organizations should begin with well-defined tasks where physical AI can demonstrate clear value and build internal expertise.
• Design through form exploration and avoid defaulting to humanoids: Not every robotics application requires a humanoid form factor. Organizations should evaluate whether specialized, mobile, or semi-humanoid designs better suit their specific operational needs.
• Redesign workflows for human-robot collaboration: Successful physical AI deployment requires rethinking how humans and robots work together, including safety protocols, task allocation, and communication mechanisms.
• Scale through platform-based architectures: Rather than deploying isolated robots, organizations should adopt platform approaches that enable coordination across multiple robots and systems, similar to Fujitsu's Kozuchi Physical OS model.

Why Is Industry Adoption Accelerating Now?

Market research from Capgemini reveals that physical AI adoption is already underway across industries. Nearly eight in ten organizations (79%) are already engaging with physical AI today, signaling momentum from experimentation to deployment. Labor shortages are a primary driver, cited by 74% of executives as a key reason for accelerating adoption of adaptive, AI-enabled robotics.

Beyond efficiency gains, over 60% of executives believe physical AI will make previously impractical use cases viable, expanding impact across productivity, resilience, safety, and growth. This perspective explains why companies like Pudu are expanding into industrial applications and why Fujitsu is investing in research infrastructure now.

"The Fujitsu-Carnegie Mellon Physical AI Research Center builds on CMU's focus on developing AI and robotics systems to tackle real-world problems and the university's collaboration with industry to put those innovations into practice and inspire what's next," noted Martial Hebert, Dean and University Professor of Robotics at Carnegie Mellon's School of Computer Science.

Martial Hebert, Dean and University Professor of Robotics, School of Computer Science, Carnegie Mellon University

The partnership model emerging between companies like Fujitsu and institutions like Carnegie Mellon suggests that the next wave of physical AI breakthroughs will come from organizations that can simultaneously conduct rigorous research and validate solutions in real-world conditions. Pudu's commercial success demonstrates that there is substantial market demand for these solutions, while the academic partnerships ensure that the underlying technologies continue to advance.

For organizations considering physical AI adoption, the message is clear: the technology is moving from experimental to deployable, but success requires careful planning, realistic timelines, and a commitment to human-robot collaboration rather than full automation.