This week, the signal is clear: physical AI is no longer a robotics feature, it’s an infrastructure play. Compute, connectivity, sensing, and skills are becoming the constraint set.

🤖 The new automation stack is “Physical AI + Infrastructure + Skills”

If you’re planning warehouse, manufacturing, or transportation automation for 2026–2028, the conversation is shifting from which robot to which stack.

1) Physical AI is consolidating into platforms (and that matters for your roadmap)

Alphabet’s robotics software company Intrinsic is moving under Google, explicitly to accelerate “physical AI” by pairing robotics tooling with frontier AI and cloud resources.

This is the same strategic arc we’ve seen in other waves: once a capability becomes foundational, it gets pulled closer to the core platform (models, developer tools, deployment/runtime, data).

What to do with that insight (operator lens):

– If your automation program depends on “robot apps” (multi-vendor), start mapping portability risk: what breaks if the platform layer changes pricing, interfaces, or supported hardware.

– Push vendors to commit to integration contracts (APIs, telemetry, simulation interfaces) that survive model updates.

2) AI infrastructure is now an operational constraint, not just an IT line item

​McKinsey & Company frames the bottleneck: AI growth depends on data centers, fiber, power, and distributed compute, and the value pools are increasingly measurable. They estimate global data center demand could more than triple by 2030 and call out fiber connectivity, “intelligent networks,” and GPU-as-a-service as major pools.

For logistics automation, this is practical, not theoretical:

– Inference at the edge (vision, safety envelopes, motion policies) is sensitive to latency and reliability.

– Distributed sites (DC networks, cross-docks, yards) need consistent connectivity and governance.

We should treat AI infrastructure like you treat MHE utilities: do a site-by-site “compute & connectivity readiness” audit (power headroom, network redundancy, on-prem GPU options, cloud egress sensitivity).

3) Funding is flowing to “foundation layers”, sensing, models, and autonomy

Several rounds this week reinforce what investors think the stack is:

– Wayve raised $1.2B with a plan tied to robotaxis in London and broader AV partnerships.

– RLWRLD raised $26M to scale industrial robotics AI (robot foundation models for industrial environments).

– ZaiNar emerged with $100M+ invested, positioning itself as a wireless-network-based sensing layer (location without GPS/cameras).

– AI2 Robotics announced new Series B funding to advance embodied AI and its AlphaBot platform.

Why operators should care: these “layer” bets tend to become the default building blocks vendors embed. Expect faster capability jumps, and faster obsolescence of point solutions.

4) The hidden limiter: skills and decision density

Even with more AI tools, freight teams are making more decisions, not fewer. A survey reported by The Supply Chain Xchange indicates:

– 74% make >50 operational decisions/day

– 50% make >100/day

– 18% exceed 200/day

…and 83% say they operate in reactive mode, driven by fragmented systems and validation work.

That aligns with the organizational reality McKinsey & Company highlights: scaling AI is usually a skills, leadership, and culture challenge more than a pure tech challenge.

Measure and attack decision density. Don’t just automate tasks, redesign workflows so fewer human approvals are required (clear policies, confidence scoring, exception bands).

🧠 Supporting Insights

🚚 Fleets are optimizing for resilience, not just efficiency

​Geotab’s 2026 report frames the “pandemic echo” (asset retirement cycle) plus inflation and rates as a perfect storm, pushing resilience to the top. Their longitudinal view shows a 38.7% reduction in collisions per million miles (2021–2025) in the U.S./Canada.

In parallel, Verizon Connect reports fleets are making AI a core tool, with adoption rising across safety, maintenance, and dispatch use cases.

Operator move: treat telematics AI as a control system: define which alerts trigger action, and which are logged for trend review (avoid alert fatigue).

📦 Warehouse automation demand is up — but price inflation is part of the story

New data reported by The Supply Chain Xchange shows warehouse automation order intake rose 7% in 2025, while noting rising steel and labor costs inflated project values even when unit demand was flatter.

Operator move: separate “volume” from “value” in your vendor pipeline. Track throughput per $ CapEx and time-to-value as first-class metrics.

🔁 Reverse logistics is turning into a competitive battleground

​McKinsey & Company estimates U.S. consumers returned nearly $1T in merchandise in 2024 and that retailers spend roughly $200B annually recovering value. Their case is blunt: modernize policies + disposition decisions with AI and automation to convert returns from cost center to advantage.

Operator move: start with dispositioning. Add item-level decisioning (sellable vs. refurbish vs. liquidate) and optimize routing to the best recovery node, not the closest.

🤝 Cross-border commerce is getting re-wired (EU–China)

​DHL Group and JD.com signed an MoU to support German brands in China and expand JD’s European retail platform Joybuy.

Operator move: if you run cross-border fulfillment, watch how partnerships reshape returns routing, customs brokerage, and inventory placement for Europe↔China flows.

🧰 Sensors and components keep getting better (and cheaper)

​Teledyne FLIR launched the Lepton XDS: a compact dual thermal-visible module (160×120 thermal + 5MP visible) designed to bring richer thermal context to embedded systems.

Operator move: thermal is no longer “nice-to-have” for automation safety and condition monitoring. Revisit use cases like overheated motors, electrical panels, and fire detection in high-density storage.

⚡️ Snippets

• 🤖 Humanoid component race:Tesollo commercialized a lightweight robotic hand designed for compact humanoid form factors. a sign that end-effectors are turning into a fast-moving product category.
• 🛠️ Better motion stacks, faster commissioning: Integrated motion control is increasingly packaged as a cohesive stack (control + sensing + safety), reducing the integration tax for sophisticated trajectories and reactive behavior, as described by The Robot Report.
• 🏭 Regional show-as-signal:The Robot Report flagged Seoul’s Automation World spotlight on Chinese humanoid makers, less spectacle, more signal that component ecosystems are standardizing.
• 📱 Agentic UX is going mainstream:Google’s Gemini can now automate some multi-step Android tasks, the same “intent → plan → execute → confirm” loop will shape how warehouse and TMS tools evolve.
• 👷 Frontline ops software is consolidating:Humand raised $66M for a platform connecting deskless workers. Adoption will hinge on integration into daily workflows (shift handoff, safety, training), not feature count.
• 🚆 Rail footprint expansion continues:Cando Rail & Terminals acquired Utah railroad assets from Savage, extending first/last-mile rail and terminal reach where shippers want modal options.