Industry Brief, July 25, 2026: Humanoids on Chinese lines, steel prices climb, robot gyms

By MillBrief Editorial · July 25, 2026 · Retrospective edition, published September 23, 2026

This is a retrospective edition. The brief did not run on July 25, 2026; we wrote it on September 23, 2026 using only reporting that was published on or shortly before July 25, 2026, and it makes no forecasts.

This is a retrospective edition covering July 25, 2026, using only reporting published on or up to two days before that date. Three items for the weekend: Chinese factories testing humanoid robots on single, narrow tasks, a steelmaker expecting prices to keep rising, and a robot maker opening sites where customers can train robots before deploying them. Our own words, links to the original outlet, vendor numbers labelled as vendor numbers.

Chinese factories test humanoids one station at a time

China’s state news agency Xinhua reports on humanoid robots in production trials. Robots from Aitu, a sewing technology company in Zhejiang, are being tested at garment makers including Sunrise Group, Eifini and Chenfeng Group on narrow jobs: loading and unloading fabric at template sewing machines, positioning material for pocket setters, and placing cut pieces at fusing presses. For one product category and one fabric type, Aitu says its robots separate fabric pieces successfully 97% of the time, with more than 98% of the resulting sewing meeting quality standards, and it estimates a robot could pay for itself in about 18 months. Those are the company’s figures. Its general manager, Lou Lingtong, told Xinhua the robots can do single processes well but cannot yet run a whole production flow without people.

The article also describes PIABOT Robotics’ G2, which the company says completed 2,283 tasks over eight hours without an error in a trial at a large electronics maker, and cut a hole-tapping cycle on aluminum seat-belt parts from about 18 seconds by hand to as little as 12.9 seconds. Xinhua says a government programme aims to build deployment capacity at the 10,000-unit scale by year end. It lists the hard parts as manipulation, handling materials the robot has not seen before, and factory conditions such as reflections, dust and continuous running, and notes that a lack of common interfaces and safety standards raises integration costs.

Why it matters: Note the conditions attached to the best numbers: one product, one fabric, one station. That is the right way to start any automation project, humanoid or not. Our guide to what to automate first explains why a stable, repetitive, low-mix task is the place to begin, and our payback guide shows how to test an 18-month claim against your own costs.

Source: Xinhua, July 24, 2026.

Cleveland-Cliffs expects steel prices to keep rising

Steelmaker Cleveland-Cliffs expects its adjusted earnings before interest, taxes, depreciation and amortization to more than double in the third quarter, to $575 million, as Section 232 tariffs and tight supply push steel prices higher, Manufacturing Dive reports. Its average selling price rose to $1,124 a ton in the second quarter, up $76 from the first, and the company expects another rise of about $55 a ton in the third quarter, with higher pricing to come in the fourth. It also sees a chance in the coming months to reset many fixed-price contracts at higher levels.

Second-quarter shipments were just over 4 million tons, down 6% from the first quarter, which the company put mainly down to maintenance outages and stronger automotive demand that lengthened lead times. It expects shipments to rise by 300,000 tons in the third quarter. Shipments to automotive customers were the highest in two years. About 30% of sales each go to steel service centres, carmakers and makers of infrastructure and heavy equipment. Chief executive Lourenco Goncalves called Section 232 the most effective industrial policy in a generation.

Why it matters: For fabricators and machine shops, this is a supplier telling investors to expect higher steel prices and contract resets. Every scrapped part costs more when the metal does, so quality losses deserve a fresh look; our OEE calculator separates quality losses from downtime and slow running so you can see which to fix first.

Source: Manufacturing Dive, July 24, 2026.

NEURA opens “gyms” where robots train before deployment

The Robot Report reports that NEURA Robotics, based in Metzingen, Germany, has partnered with RWTH Aachen University on a robot training facility of about 3,000 square metres at the university’s Campus Melaten. A second German site, of more than 2,300 square metres, is planned at Munich Airport with the Technical University of Munich. NEURA says it is building 10 such facilities across Europe, the US and China, expects half to be running by the end of 2026, and is funding them from a Series C round of up to $1.4 billion.

The sites combine physical training with high-fidelity simulation to produce data for NEURA’s cloud-based Neuraverse development platform. NEURA’s pitch to buyers is that partners can train and validate robots for their own use cases before committing to a full deployment, which it says reduces adoption risk. The company says manufacturing, automotive, healthcare and industrial automation firms have expressed interest.

Why it matters: Testing a robot on your own task before you buy is sound practice, wherever it happens. What counts is that the test uses your parts and is judged against numbers agreed in advance; our notes from engineers explain why numeric acceptance criteria belong in the contract.

Source: The Robot Report, July 24, 2026.