Physical AI hits the warehouse floor: five industrial robotics moves from June 2025

Ambi Robotics and Pickle Robot announced on June 30 the first commercial integration of their respective systems into a single automated inbound logistics workflow, covering truck unloading through palletizing without human intervention. That milestone, reported by RoboticsTomorrow, is the clearest sign yet that physical AI is crossing from proof-of-concept into production-grade supply chain deployments.

Inbound logistics gets its first end-to-end automated stack

The Ambi-Pickle integration tackles a persistent gap in warehouse automation: the two hardest manual steps, unloading trailers and building pallet loads, have historically required different vendor systems that don't coordinate. The combined deployment closes that gap by linking Pickle Robot's trailer-unloading capability directly to Ambi Robotics' sortation and palletizing intelligence.

For distribution center operators, the practical implication is that inbound dock labor can be redeployed rather than simply supplemented. The integration also sets a precedent for how physical AI vendors with complementary hardware may go to market together rather than compete for the same line item.

Kawasaki and Dexterity deepen their warehouse bet

A week earlier, Kawasaki Robotics and Dexterity announced an expanded collaboration built around the RL030N, an 8-degree-of-freedom robot arm platform. The system combines Kawasaki's industrial robot engineering, Dexterity's Mech hardware, and Dexterity's Foresight World Model, a physics-informed AI that helps the arm predict and adapt to object behavior in real time.

Eight degrees of freedom matters operationally because it allows the arm to reach into constrained spaces, a requirement for mixed-SKU warehouse picks that 6-DoF arms frequently struggle with. Procurement teams evaluating robotic picking for high-SKU environments should treat this specification as a baseline comparison point.

No-code cobot painting removes the specialist requirement

Hirebotics introduced the Cobot Painter on June 25, built on FANUC's CRX-10iA/L Paint hardware and the company's Beacon Platform, which requires no code to configure. The unit carries an explosion-proof rating, which means it can operate in solvent-heavy spray environments that would otherwise require a hardened, dedicated paint cell.

The practical upside for contract manufacturers and job shops is significant. Dedicated paint cells are expensive to build and maintain, and traditional paint robot programming demands specialized integrators. A no-code, explosion-proof cobot changes both the capital expenditure and the staffing model for finishing operations.

Edge inference moves off the cloud for real-time robotics

DEEPX and Sixfab launched the DEEPX AI HAT, an AI acceleration module designed for Raspberry Pi hardware. The product eliminates cloud dependency for real-time inference in robotics and smart automation, according to RoboticsTomorrow. The partnership ships a production-ready Starter Kit alongside SDKs targeted at global developers.

For operations technology teams, the significance is architectural. Real-time robotic control loops that rely on cloud inference introduce latency and connectivity risk. Hardware-level inference at the edge, at Raspberry Pi price points, lowers the barrier for deploying autonomous guidance in environments where network reliability is constrained, including cold storage, outdoor yards, and older facilities with limited Wi-Fi infrastructure.

X Square Robot's $2.8B raise signals foundation model ambitions

X Square Robot disclosed that it has completed four consecutive financing rounds, surpassing a $2.8 billion valuation, according to RoboticsTomorrow. The company is developing embodied AI foundation models and is the only company in the embodied AI space backed by all four major Chinese internet technology leaders.

Foundation models for physical AI represent a different bet than task-specific robot software. If they mature, they could enable a single trained model to generalize across robot types and tasks, much as large language models generalized across text tasks. That prospect is driving investment concentration, and it will likely influence how incumbent automation vendors position their own AI roadmaps over the next 18 to 24 months.

German manufacturers add pressure on humanoid timelines

A survey cited by RoboticsTomorrow found that 82 percent of respondents within the German economy consider humanoid robots drivers of future innovation, with calls for faster adoption. Germany is a major buyer of industrial automation globally, and demand signals from that market tend to pull forward investment and product timelines from major robot manufacturers.

What this means for your team

• Evaluate inbound dock automation as a system, not separate line items. The Ambi-Pickle integration shows that end-to-end contracts covering unloading and palletizing are now commercially available; benchmark any existing point-solution quotes against integrated alternatives.
• Add DoF and world-model AI to your cobot RFP criteria. The Kawasaki-Dexterity RL030N raises the bar for mixed-SKU picking; 6-DoF systems without adaptive AI may underperform in high-variety environments.
• Re-examine finishing cell capex. If spray painting is on your automation roadmap, the Hirebotics Cobot Painter's no-code, explosion-proof design may eliminate the need for a dedicated paint cell build-out.
• Audit your inference architecture. Any robotic deployment that currently depends on cloud inference should be assessed for edge-hardware alternatives, particularly in connectivity-constrained facilities.