Industry Brief, July 24, 2026: Forced-labor tariffs, Samsung robot unit, physical AI poll

By MillBrief Editorial · July 24, 2026 · Retrospective edition, published September 23, 2026

This is a retrospective edition. The brief did not run on July 24, 2026; we wrote it on September 23, 2026 using only reporting that was published on or shortly before July 24, 2026, and it makes no forecasts.

This is a retrospective edition covering July 24, 2026, using only reporting published on or up to two days before that date. Four items: new US tariffs tied to forced labor that take over from an expiring global duty, Samsung’s new robotics unit, a services firm’s survey of manufacturers on physical AI, and a cobot customer story from ABB. Our own words, links to the original outlet, vendor numbers labelled as vendor numbers.

New forced-labor tariffs on 60 trading partners replace the expiring global duty

Manufacturing Dive reports that the United States began charging new Section 301 tariffs on imports from 60 trading partners on Friday, July 24, the same day the temporary global tariff imposed under Section 122 expired. Goods from China, the European Union and Mexico are among those covered, at either 10% or 12.5% depending on the country. The tariffs follow an investigation launched in March, and US Trade Representative Jamieson Greer proposed them in June after finding that the 60 countries had failed to impose and effectively enforce bans on goods made with forced labor. A USTR fact sheet says the tariffs would cover 99.4% of US imports.

The details matter for import costs. For partners with Most-Favored-Nation rates, such as the EU, Japan and South Korea, the new tariff is charged net of the MFN duty, so the combined rate does not exceed 10% or 12.5% unless the MFN duty alone is already higher. Many agricultural goods are exempt, as are goods already subject to Section 232 tariffs such as steel and aluminum. Goods loaded on a vessel before July 24 and entered before July 28 are not affected. Canada and Brazil, which both received separate new tariffs in the previous week, are on the list, and the filing did not say how the measures would interact. A separate Section 301 investigation into global manufacturing capacity is still under way.

Why it matters: Imported robots, controls and spare parts may now carry a different duty than they did a week ago, and the rules may change again. When you collect quotes, ask each bidder to list its assumptions and exclusions, as the bid format in our RFQ guide requires, so you can see whether duties are included, and rebuild the delivered cost in the TCO calculator.

Source: Manufacturing Dive, July 24, 2026.

Samsung sets up a robotics unit under its device chief

Tech Times reports that Samsung Electronics launched a Robotics eXperience (RX) Business Promotion Office on July 21, reporting directly to TM Roh, chief executive of its Device eXperience division. The unit’s strategy is led by Executive Vice President Lee Dongkun, who previously directed robotics strategy at Hyundai Motor Group and joined Samsung in May. Samsung plans robotics research hubs in the United States, China and Japan, a main research base in Seoul, and a “Robot Data Factory” at its Gumi plant, where robots doing transport and assembly on real production lines will generate data to train its robot AI models.

According to the article, Samsung’s stated order is industrial first: humanoid robots will be used in its own factories, then offered to other manufacturers, with consumer robots later. Its goal is to turn all of its manufacturing sites into “AI autonomous factories” by 2030. Samsung became the largest shareholder of Rainbow Robotics, with about 35%, by the end of 2024, and the article says Rainbow’s RB-Y1 humanoid gives the new unit a hardware base. Samsung Research’s own figures for its robot AI model include a 95% success rate on a precise hose-insertion task; those are the company’s results, not an independent test.

Why it matters: Large manufacturers testing humanoids in their own plants first is a sensible order, and the one to copy at smaller scale: prove a robot on your own parts before you depend on it. For most small plants today, a proven arm is still the practical option; our cobot vs industrial robot comparison covers that choice.

Source: Tech Times, July 24, 2026.

TCS survey: manufacturers keep spending on physical AI, but ownership is unclear

Tata Consultancy Services (TCS) has published a physical AI readiness report based on a survey of 300 executives and vice presidents at manufacturers in North America and Europe, Dataquest reports. None of the surveyed manufacturers plans to cut spending on physical AI, meaning AI-driven machines and robots, and 26% plan to increase it. Warehouse operations were the area most expected to benefit, named by 77% of respondents as facing a significant or transformational impact, followed by assembly and manufacturing operations at 75%.

The same survey shows how early most companies are: 68% remain in experimental or non-deployment phases. 44% said accountability for physical AI failures is unclear or undefined in their organisation, and 40% said they are unprepared for emerging regulation. TCS names legacy system integration, data infrastructure and workforce skills as the biggest hurdles. TCS sells AI and technology services, so treat this as a vendor survey.

Why it matters: The 44% figure is the useful one. A robot or AI system that nobody in the plant is accountable for tends to stall after the installer leaves, whatever technology it uses; lack of in-house ownership is one of the causes in why automation projects fail. Name the owner before you sign the order.

Source: Dataquest, July 24, 2026.

ABB customer story: a Slovenian cell maker’s software-only speed gain

In a customer story published by ABB, Sunrise Robotics, based in Ljubljana, Slovenia, describes building its autonomous cells for small and medium-sized manufacturers around ABB’s GoFa cobot. The cells combine on-board AI, built-in perception and radar-based safety and handle high-mix pick and place, positioning and machine tending. Sunrise says it chose GoFa for a working envelope large enough for euro pallets at floor level, speed of up to 2 metres per second and positional accuracy, and a software interface that can run its AI models in real time.

The headline claim is that Sunrise raised throughput by 20%, from 72 to 90 parts per hour, by changing the arm’s motion parameters in software, with no change to hardware or AI models. Sunrise also says its engineers can control the arms remotely to recover cells from unexpected states without travelling to the customer. These are vendor and customer claims published by ABB, which sells the arm.

Why it matters: A 20% gain from a parameter change is a reminder that the arm is rarely the whole story; motion tuning, grippers and integration decide what a cell delivers, and our cobot cost guide explains why the arm is only part of the budget. Remote recovery is convenient, but it is also a network connection into your plant: agree in writing who can connect, how, and when.

Source: ABB, July 24, 2026.