Humanoid Factory Testing, Warehouse Reshoring and AI Infrastructure Constraints — September 22, 2026
Automation History does not claim authorship or ownership of the third-party articles featured in our Daily Articles section. All articles, research, and reporting are credited to their original authors and publishers, with links provided to the original sources. Automation History shares these links solely for educational and informational purposes, helping readers discover and learn from relevant content. All rights remain with the respective authors and publisher
1. Boston Dynamics opens an Atlas training ground inside Hyundai’s Georgia plant
Axios reported September 21 that Boston Dynamics opened its Robotics Metaplant Application Center inside Hyundai’s vehicle plant near Savannah, Georgia. Initial Atlas training will focus on logistics and parts sequencing before progressing to assembly and heavy lifting; Boston Dynamics did not disclose how many robots are already operating there.
Why it matters: A production-site training center should expose the faults that controlled demonstrations miss—part variation, blocked aisles, handoff delays and recovery after an interrupted task. Repair leaders should ask for intervention rates, safe-state behavior, tooling wear and technician access before treating Atlas as production equipment.
Read the September 21 Axios report
2. Daifuku doubles U.S. capacity as ASRS demand rises
Daifuku told Automated Warehouse on September 21 that a 300,000-square-foot expansion in Hobart, Indiana, doubled its local capacity and is now fully operational. The company is bringing more Japanese technology into North American production, adding staff and developing modular software around rising demand for ASRS, mini-load and shuttle systems.
Why it matters: Local manufacture can shorten delivery and parts lead times, but the real system still spans storage machines, conveyors, controls and integration software. Supervisors should verify which components are genuinely sourced and supported locally, then test diagnostic coverage and recovery across subsystem boundaries.
Read the September 21 Automated Warehouse interview
3. EnerVenue opens a 95%-automated battery factory in China
EnerVenue opened its first commercial factory in Changzhou on September 22 after abandoning a planned Kentucky site, Reuters reported. The $20 million–$50 million facility is about 95% automated, targets 250 MWh of annual nickel-hydrogen cell capacity this year and is planned to reach 1 GWh in the third quarter of 2027.
Why it matters: The decisive advantage was an engineering cluster able to iterate on hydraulics, pneumatics and automation—not incentives alone. Operations leaders evaluating locations or suppliers should measure the surrounding troubleshooting, tooling and vendor ecosystem because difficult ramp problems rarely stay inside the factory fence.
Read the September 22 Reuters report
4. Alibaba links a new AI chip to a 20-gigawatt infrastructure plan
Alibaba unveiled the Zhenwu V900 on September 22, saying it delivers three times its predecessor’s performance and can connect in clusters of as many as 500,000 chips; mass production is planned for the first quarter of 2027. The company is also training larger Qwen models and wants Alibaba Cloud’s global data-center capacity to exceed 20 GW by 2032.
Why it matters: A vertically integrated roadmap can reduce outside dependencies while creating a large, tightly coupled stack of chips, firmware, networks, cooling and models. Industrial buyers should separate announced capability from qualified availability and demand lifecycle, compatibility and failure-recovery evidence before committing critical workloads.
Read the September 22 Reuters report
5. Texas freezes new data-center permits during a grid audit
Texas Governor Greg Abbott directed the state environmental agency on September 21 to stop issuing new data-center permits until regulators complete an audit of grid, water and tax impacts. More than 470 GW of data centers and other large users were waiting for grid connections when the state had already paused new connections—over five times Texas’ peak demand.
Why it matters: Power, permits and public acceptance can become the critical path after equipment has already been specified. Leaders should make utility capacity, curtailment rights, water reporting and permit milestones explicit gates in every AI-infrastructure project rather than treating them as external assumptions.
Read the September 21 Reuters report
6. U.S. public pressure for stronger AI safeguards reaches 73%
A Reuters/Ipsos poll published September 22 found that 73% of 1,277 U.S. adults believe AI companies have not done enough to prevent serious societal harm. The same share prioritized safe, responsible development over U.S. leadership, while 55% said slowing AI development would be good and 13% said it would be bad.
Why it matters: Trust is an operating constraint: a technically successful deployment can still fail if employees, customers or regulators cannot see its boundaries. Plant leaders should publish who authorizes automated actions, how incidents are logged and stopped, and which decisions remain accountable to a person.
Read the September 22 Reuters/Ipsos report
7. Ineos mothballs three Hull chemical plants over energy costs
Ineos said September 22 that it is suspending operations at three chemical plants in Hull, blaming European gas costs that it says make production uncompetitive with the U.S. and China. The facilities employ almost 4,000 people and make inputs used in medicines, clothing, cosmetics, detergents, construction materials and explosives.
Why it matters: Energy cost can idle technically sound assets and then propagate shortages far downstream. Repair and operations teams should know which critical consumables depend on energy-intensive plants, model shutdown and restart risks, and preserve equipment carefully if a line enters an extended mothball state.
Read the September 22 Reuters report
8. BYD seeks more than 8,000 workers as Xi’an output ramps
BYD is recruiting more than 8,000 people across multiple Xi’an plants and business units, Reuters reported September 21, following production-line adjustments and a rapid output increase. Local government and industry organizations are advertising incentives of up to 6,000 yuan for successful hires.
Why it matters: Automated factories still need large human ramps when volume changes quickly. Maintenance leaders should translate the hiring wave into instructor capacity, qualification matrices, spare-parts demand and shift coverage so new throughput does not arrive faster than safe technical competence.
Read the September 21 Reuters report
9. EU proposes energy-and-water labels for data centers
The European Commission proposed a September 21 labeling system requiring data centers rated at 500 kW or more to report energy and water efficiency, local water-stress context and options such as waste-heat reuse. The scheme would not cap consumption or disclose total power use, but could precede mandatory minimum standards as EU data-center capacity grows toward a projected 28 GW in 2030.
Why it matters: Resource efficiency is moving from an internal KPI toward a comparable operating disclosure. Facilities teams need auditable meters, defensible boundaries and ownership of reporting data—plus maintenance practices that keep cooling and heat-recovery performance from drifting after commissioning.
Read the September 21 Reuters report
10. Danube cargo reveals Iron Age bulk production and transport
An Antiquity study published September 21 analyzes at least 500 bipyramidal iron bars recovered during Danube gravel extraction near Deinham, Austria. The roughly one-metric-ton cargo dates to the third or second century B.C.; researchers consider transport loss more likely than ritual deposition, though excavation destroyed the original context.
Why it matters: For Automation History, the standardized semifinished bars document production capacity, inventory form and heavy river logistics before modern factories. The loss also carries a modern lesson: material can be recovered while process context is destroyed, so traceability and disciplined incident documentation are part of preserving operational knowledge.