Industry Brief, August 25, 2026: Automation loans, readiness gaps, humanoid tests

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

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

This is a retrospective edition covering August 25, 2026, using only reporting published on or up to two days before that date. Today: a state loan programme built for exactly the kind of equipment upgrade a small manufacturer puts off, a survey suggesting robot plans are running ahead of the management needed to run them, what BMW is actually testing a humanoid on, and China’s move to write the rulebook for humanoids. Our own words, links to the original outlet, vendor numbers labelled as vendor numbers.

Montana relaunches its automation loans: up to $100,000 direct, plus collateral help

The Montana Department of Commerce announced a revised Montana Automation Financing Program for manufacturers upgrading, replacing or expanding equipment. It now has two parts. Commerce itself lends up to $100,000 directly to eligible manufacturers. Separately, a collateral support option lets partner lenders serve manufacturers who are creditworthy but lack enough collateral for a conventional loan. Funding is $1.4 million for direct lending and $1 million for collateral support, and both pools are meant to revolve.

The dates matter. The first direct-loan window opens September 1 and closes October 23, 2026, with all materials due by November 6 and the review committee expected to meet the week of November 9. After that, direct loans reopen on January 1, 2027, first come, first served while funds last. Collateral support opens December 1, 2026 on a rolling basis, and the application has to be completed by a participating lender. The release cites Red Oxx, a Billings bag and gear maker, which used the direct loan for an AI-assisted laser-cutting system; its CEO says the low rate gave the crew time to learn the machine before it had to pay its way.

Why it matters: Cheap, patient money changes the payback arithmetic more than most buyers expect, because a lower financing cost buys you a longer ramp-up without the project looking like a failure. The collateral option is the more unusual piece: small shops often have the cash flow to service a loan but not the assets to secure one. If you are in Montana, the calendar is tight, so get a firm quote early. Elsewhere, it is worth asking your state commerce department whether anything similar exists.

Source: Montana Department of Commerce, August 25, 2026.

Robot plans are outrunning robot strategy, says an Intel-commissioned survey

Manufacturing Dive covers a survey commissioned by Intel of 800 business and IT leaders, robotics specialists and government and healthcare officials worldwide, run by Man Bites Dog and Coleman Parkes Research. Nearly a quarter of respondents were from manufacturing. The headline finding: seven in 10 senior manufacturing leaders expect their organisation to run a robot fleet within five years, but only about four in 10 have a formal strategy for managing people and robots together. The report puts the manufacturing “preparedness gap” at 26 percent, larger than in retail and healthcare, smaller than in smart cities and defence (defence was widest, at 48 percent).

The report also flags gaps in workforce skills, safety, form factor and scaling, and suggests unclear ownership of robotics strategy inside companies may be part of the problem. Intel’s John Healy is quoted saying most organisations are unprepared and that many are building on systems and teams from before robots could adapt.

Why it matters: Treat the numbers with care: this is a vendor-commissioned survey, and Intel sells into robotics. But the gap it describes is familiar. A common reason a first cell disappoints is not the hardware, it is that nobody owns it after commissioning. Before you buy, name the person responsible for the cell’s uptime, training and next steps. We cover this in why automation projects fail.

Source: Manufacturing Dive, August 25, 2026.

BMW moves a humanoid from the body shop to parts sequencing, and declines to give volumes

Automotive Logistics reports that BMW is testing Figure AI’s Figure 03 humanoid at its Spartanburg plant in the US on a sequencing job. Parts arrive unsorted in large containers, and the robot picks them into a sequencing trolley for just-in-sequence delivery to assembly. A tugger train or transport robot then takes the trolley lineside. BMW says it is the first humanoid in a logistics workflow at any of its sites. The earlier Figure 02 worked in the body shop loading sheet-metal parts into fixtures and, over just under a year, supported production of more than 30,000 X3s.

According to the report, Figure 03 adds soft components, wireless charging, tactile hands with palm cameras, and a run time of 4 to 5 hours, with two robots working in tandem so one charges while the other works. Figure also says its San Jose plant can build 12,000 robots a year. BMW’s own framing is more cautious: the test is about validating the application and gaining experience, not announcing target quantities, and occupational safety experts are involved early on clearances and spacing.

Why it matters: Note the task. Picking unsorted parts into a sequence is a job where fixed automation struggles with variety, which is where a flexible robot has a real argument. Note too that a 4 to 5 hour run time needs a second robot to cover a shift, which doubles the hardware in any cost comparison. For most mid-size plants, a conventional arm or a cobot with a vision system is likely to stay the cheaper, proven option for some time.

Source: Automotive Logistics, August 25, 2026.

China drafts a plan for at least 100 humanoid robot standards by 2028

Macau Business reports that China’s Ministry of Industry and Information Technology has published a draft guideline proposing at least 100 key standards for the humanoid robot industry by 2028, open for public comment from August 25 to September 23. The proposed standards cover capability testing and evaluation, key technologies, platforms and systems, application scenarios and safety governance, with a target of more than 200 companies implementing them. The draft also calls for common methods for grading robot capability and for unique robot identifiers, plus industrial requirements on operating capability and environmental adaptability in structured and semi-structured settings. Ethics and social-impact standards are included too.

The report notes that more than 100 humanoid-related standards are already being developed in China and abroad, and that the ministry said in July that China had developed more than 400 humanoid models. It also lists practical industrial problems: reflections and dust can confuse perception, and continuous running stresses durability, heat management and component life.

Why it matters: Standards are how a buyer compares machines without taking the vendor’s word for it. A graded capability scale and common test methods would make humanoid spec sheets checkable. They are still a draft, and they will reflect Chinese priorities, but anyone evaluating an imported humanoid in the next few years may meet them. Until then, ask any vendor which safety standards the robot is certified to today, as part of your integrator questions.

Source: Macau Business, August 25, 2026.