Everything is an urgent shipment for every startup. Let’s focus on that, since the structure is usually lacking and speed is in their DNA. It starts when you forget to plan a shipment, understand what it implies to move anything within the R&D space, and even apply a framework like Product, Place, Presence. If that’s not bolder than my title, well, I don’t know what is. Hiding your lack of preparation under the urgent syntax.

Oh well, both the US and the European Union just have the biggest gift for your lack of preparation: they removed the de minimis. People would always hide behind a smaller valuation of a “free sample.”

This will change the very foundation of how start-ups build their supply chain. It’s going to be the biggest impact of the decade, since this will force everyone to be more curious, ask questions, and understand how things move better.

Even though people will still come with the excuse that “I have done it this way my whole life,” it will not fly anymore. Or it can, but it will get stuck.

So, tell me. Do you really need to classify a shipment? Let’s just dive into that topic, since all your suppliers from Asia understand that.

For the European Union, the EU Council is applying a €3 customs duty on parcels valued below €150. This started on 1 July 2026. In the US, on the other hand, there are two changes happening. Executive Order 14324 targeted the de minimis exception. We will tell our grandchildren how shipments under $800 used to seamlessly pass customs. In parallel with that, EO 14411 put higher restrictions on foreign IORs, therefore pushing the Importer of Record responsibility towards a domestic US party.

Will there be people taking risks and ignoring these directives? Always! I am curious to connect with people who have risk strategies in place, since there will still be a lot of shipments that are not compliant. Most compliance and cross-border people I know are very risk-averse and don’t understand how to allow the business to move forward.

If you’re working with your Asian suppliers, they are still behind. They are still relying on their relationships with their forwarders and their network like nothing has changed.

The biggest change for the supply chain is not coming from blockchain, from AI, or from living in the future. It’s from understanding your present and what moves within all your sourcing channels. And it’s here to stay.

In the electronics space, this discipline has evolved dramatically over the last five years, redefining how businesses operate, how compliance dictates the cross-border lifecycle of hardware, and how safety intersects with volatile trade policy.

The supply chain is no longer viewed merely as a logistics function. The biggest ODMs now know that in 2028 they can shift their battery cells to 50 Wh from 20 Wh. Testing labs now know that in 2029 the old battery classification will change for the first time in the last 20 years.

The speed at which it is changing will continue to shift from OEMs, ODMs, and JDMs to 4PLs, airlines, IATA/ICAO, and every player in the supply chain. Certain regulations will change rapidly, while others might even change seamlessly. International standardization will match neither the European pace, US interpretation, nor Asia’s speed. Yes, the new harmonization does not belong to a single regulatory world. The supply chain of 2030 will favor those who understand how the pieces of the puzzle connect.

As much as you can see predictability coming from the old continent, the US will continue to drive rapid movements. Let’s look at two examples in parallel: the EU Machinery Regulation (EU) 2023/1230 entered into force in 2023 with a clear implementation timeline of 2027. The FCC published a Public Notice on 28 July 2026 stating that the US is no longer accepting the importation of robots made outside the US starting… Yes, you guessed it, the very same day. Companies that haven’t done so already still have 30 days to ask questions—or to get conditional approval from the Department of War.

The most advanced players have moved from reactive gatekeeping compliance to compliance by design, where they incorporate ongoing regulatory changes to build a dual-track roadmap. Teams must now manage multi-year EU Directives on one track, while staying agile enough to absorb overnight geopolitical trade controls and emergency Public Notices on the other.

Digital traceability is the key. Tracking the physical Bill of Materials is no longer enough. The future belongs to those who master the full stack: mechanical BOM, SBOM, digital product passports, and sub-tier component provenance.

The future is autonomous, and it is closer than we all fear. In order to fully understand that, we just need to accept the change happening before our eyes and build a 2- to 8-year roadmap into our supply chains.

One classroom was in Geneva, one of the most international cities in the world, and the other was a trail somewhere above Chamonix where I’d signed up to train alongside the most privileged wannabe athletes money can buy. Two rooms at opposite ends of every spectrum you could name, and yet the only thing they genuinely had in common was that the people sitting in them, or running through them, all learned differently.

I’ve spent years in these rooms. IATA’s Competency-Based Training and Assessment (CBTA) courses in Singapore, Miami, Geneva and Belgium, the kind of format that started with pilots and expanded over time to supply chain professionals handling dangerous goods and lithium batteries, and then the Arc’teryx Alpine Academy in Chamonix, which is about as far from a fluorescent-lit training room as you can get. Across all of them, over all those years and all those countries, exactly one trainer had actually prepared to teach neurodivergent people, and that number should tell you everything, because one is not a rounding error. One is a pattern.

So I started asking the same question everywhere I went: how did you build this room for the minds that don’t work like the majority? Most of the trainers had simply never thought about it, which I could almost forgive, and one asked me, genuinely curious rather than defensive, why it even mattered. A few gave answers I won’t repeat here.

Here is why it matters. Somewhere between 10 and 30% of students in higher education are neurodivergent, according to Brown University, and while the real global numbers are hard to pin down, even the low end of that range means a meaningful slice of every room you will ever teach is processing information in a way the room was never designed for.

The one who got it right, and they really were the only one, didn’t come from any of these worlds at all. They train in a completely different field, nowhere near aviation or sport, and yet they were the single person in years of classrooms who had thought the problem through. They gave me their whole method almost in passing: break the information down into the smallest chunks you possibly can, and then build a routine so predictable that nobody has to waste energy guessing what comes next. They weren’t improvising, even though it sounded like instinct. Rutgers University now recommends exactly that, what they call chunking and scaffolding, breaking large assignments into smaller manageable pieces and establishing highly predictable routines, which is a fairly academic way of describing something they had already worked out for themselves, long before anyone handed them a framework for it.

And once you’ve spent enough time moving between countries and cultures, none of this feels particularly radical. Mixing different neurotypes in a workplace isn’t all that different from mixing different nationalities and cultures, and in the same way you’d never dream of designing a global training program without accounting for how culture shapes the way people hear you, you design for cognitive variation, because it’s the same instinct simply pointed at a different kind of difference. Professor Charlotte Valeur, who founded the Institute of Neurodiversity, has written a good deal about precisely this if you want to go deeper.

Which brings me back to Geneva. The Institute of Neurodiversity is headquartered there, in the very same city where I once sat in one of the most international classrooms of my life and watched a trainer run the entire session without once stopping to consider how the people in front of him actually think. The knowledge is out there, the institution is quite literally down the road, and still none of it has travelled the last few meters to the person standing at the front of the room.

So ask them. Before the next course you run, before the next room you fill, put the question to whoever is teaching it: how did you build this for the minds that don’t work like everyone else’s? Most of them won’t have an answer, and that, precisely, is the point. Celebrating neurodiversity for one month in April is easy, and building for it the other eleven starts with one small, slightly uncomfortable question.

P.S. Photo taken this July at the Arc’teryx Alpine Academy by Philipp Reiter / PeakPixelCafe.
P.S. 2 Let’s continue this topic on my Linkedin.

I have seen this enough times that I stopped blaming the battery. The battery is fine. It is the classification, and that was won or lost months earlier, at design.

Founders treat lithium batteries as a product feature. Fair enough, and it produces good work. You focus on how long the customer can use the product, the engineers spend the year making it safer, and everyone involved is doing their job properly. But a forwarder does not book your cargo on how safe it is. They book it on how it is classified. Same battery, different rulebook.

And the rejection itself is not really the cost. The cost is the development cycle sitting behind it, waiting, because nobody assigned a classification back when assigning one was free.

Lithium batteries are the most common reason forwarders reject packages, and the most frequently violated dangerous goods category. I cannot prove that with a number and neither can anyone else. No forwarder publishes a rejection rate. Their business is booking cargo and then selling you the fix, so the data stays in the building. Ask anyone you know who books this cargo. They will tell you it is the most scrutinised thing on the manifest.

The dangerous goods flag is the entry level requirement for any Product Lifecycle Management record. What goes in it? UN number, watt hour rating per cell and per pack, state of charge. Then the documents, UN 38.3 always, and a safety data sheet, because carriers ask for one even where transport law does not require it. All of this is known at design lock. Recording it there costs nothing. Finding it afterwards, from a manufacturer in a different time zone, on a Friday, costs you the three weeks.

Look at what the market is doing with that last field. On 19 June Maersk launched a Class 9 lithium ion ground freight service across the United States, Canada and Mexico, taking the dangerous goods work it already does in ocean and air onto the road. Outbound batteries have to be offered at a state of charge between 10 and 60 percent. Every shipment needs a dangerous goods declaration and a safety data sheet. The SDS is not a transport requirement for an article the way UN 38.3 is. They made it contractual anyway.

So one of the largest carriers in the world has taken a field out of your PLM record and written it into a commercial contract. State of charge is not an engineering parameter anymore. Somebody is going to check it.

You can assign a UN number in an afternoon at design lock, while nothing is moving and nobody is waiting on you. Or you can assign it in a warehouse in week three, with a forwarder who can do half of what you need.

The math is the machinery. Passing at UN level puts new classification into the Model Regulations, and the Model Regulations bind nobody. They cascade: into the IATA regulations for air, the ADR for European road, the IMDG Code for sea, the national codes behind them, each on its own adoption cycle, each with its own transition period. Stack those cycles on top of an end-of-biennium adoption and the earliest date a shipper hands a carrier a document under new rules lands at the end of the decade. And that is the version where every step goes right, which, after a week of watching the room, is not the version I would plan around.

The strongest pressure anyone applied all week came from IATA’s head of cargo, who told the room that aviation now records a thermal runaway incident roughly once a week. A fire a week, stated plainly, in the room built to regulate exactly that, and even that could not compress the calendar. That tells you more about the real timeline than any agenda item.

Now hold the timeline against what a battery is in 2026. Probably the hottest commodity in the world. Global demand reached 1.59 terawatt-hours in 2025, up 29 percent in a single year and six times the level of 2020. That is more than 30 gigawatt-hours of batteries shipped every week. Governments subsidize the factories, tariff the imports, and permit the graphite that feeds them. The industry ships 30 gigawatt-hours a week, aviation records a fire a week, and the classification that governs both will not move for the rest of the decade. All of it in the same four boxes, under rules that treat the battery in your laptop and a one-ton grid pack as members of the same family.

Semiconductors show what the other path looks like. When chips became strategic, the export rulebook was rewritten in months, then rewritten again before anyone had finished reading the first version. Batteries became strategic and the safety rulebook set a four-year fuse. Trade rules move at political speed. Safety classification moves at treaty speed. Most companies plan as if there is one clock.

Here is what the headlines will miss. A rulebook with a published expiry horizon is the most useful thing a regulator can hand an operator. Not because the rules are good. Plenty of people in that room would tell you they are not. But four years of certainty is the only regulatory ground you can actually build on, and Geneva just put a date on it.

The rules will not reach your loading dock before 2030. Your volumes will not wait that long. Build the backbone. Appoint the people. Spend the certainty.

IEA Analysis: Read it fully here.

Benchmark Mineral Intelligence is here.

At the same time, this unprecedented act from the US created a new capability. The ability to filter the use of a product for citizens of a specific country. A similar capability is the ones in which the US banks verify their customers. Know Your Customer (KYC) is the main program through which the banks know who is using their products. They don’t know if US Citizens or Permanent Residents are using it, but they have a filter for non residents. So now you know this is my style, I love to throw three ideas at you at once and let you decide which one you would like to ruminate on.

Let’s say Anthropic develops this feature to release a product only for US citizens. To have such a segregation, both internal, and external, to confirm citizenship status for the use of the product. Will this inspire other nations to create a similar capability? Maybe the EU? For a nation it would be much easier than the entire EU. At this point in time there is no self executing right based on EU citizenship.

So let’s get back to talking about AI. Made in Europe vs Made in US. Because everyone is an expert on AI & Trade these days. Europeans want to use a European AI. Europeans and Americans might seem to have the same level of trust for privacy, but it differs on what they distrust: Europeans fear corporations, while Americans fear government, if you read this article from Forbes.

What is the most advanced European AI Lab at this moment? Everybody knows about Mistral, but there are a few other names popping up. In Switzerland, for example, there is no direct competitor for Mistral, but the country has decided to invest in open LLMs. They have Apertus, based on the ALPS cluster, which is co-owned by ETH and EPFL.

I was at a talk at Tufa Labs this week from Valentina Pyatkin who works at the Allen Institute for AI. A lot of nerding on the topic, but one thing that will stay rent free in my brain is that an open LLM will not get a chance to compete with a frontier lab. They use students, who are passionate, but a high paycheck can drive you much further. If you look deeper at SFT chain of thought, the open LLM Models have no chance of competing with the big frontier labs.

So, yes, Europeans trust differently and Europeans prefer to use products made in their own region. But will we ever have our own sovereign AI? The discussion goes much beyond just training a European model. It requires infrastructure, identity, talent, capital, and the ability to compete with frontier labs.

Identity is the key and this is where we need to develop a Know Your Customer capability. And we have done so well on developing and protecting the identity of European products like Champagne, Parmigiano, and Prosciutto. Everyone is now making wine in the champagne method. So let’s focus exactly on that strength of building a European AI with Identity. That might tip the scale of other topics in which Europe is half a decade behind.

P.S. Let’s continue the topic on my Linkedin.

But if you look closely at the working papers and the shifting dynamics of a working group pushing to get things into legislation for this biennium, the official narrative misses the forest for the trees. The true paradigm shift on the table isn’t just a shuffle of regulatory boxes. It is an ambitious attempt to create an off-ramp for inherently safe technology, effectively transforming batteries that do not undergo thermal runaway into virtually non-regulated goods.

Here is an inside look at the real mechanics, the technical friction points, and the high stakes shaping the next session.

The Core Strategy: Testing Out of Regulation

For several delegations, the pressure to raise the current Watthour(Wh) thresholds for cells and single-cell batteries has reached a boiling point. The goal is to lift the bar for single cells to 40 Wh and single-cell batteries to 50 Wh.

But there is a catch: to access these higher thresholds and enjoy simplified transport, manufacturers must prove their chemistry can test out of danger. If a battery can demonstrably prove it will not undergo thermal runaway, the goal is to allow it to move through the global supply chain with minimal friction.

If this core reclassification framework is finalized by the end of this biennium, it will streamline the entire regulatory landscape, allowing the committee to completely bypass at least three outstanding papers that would otherwise complicate the rules. It puts the Chairman in a challenging spot, but the path forward is clear: use the reclassification system to elegantly solve the Wh limit issue once and for all.

The Intertwined Papers: Class 9 Overhaul Meets Reclassification

One of the most critical structural insights heading into the IWG is how the main reclassification document will interact with a prominent working paper from the United States regarding the subdivision of Class 9.

While these might seem like parallel tracks, they are deeply interconnected:

• The Reclassification Drives the Labels: The new classification system will be the actual vehicle that puts new labeling requirements in place, rather than the labels dictating the system.

• Leveraging the Week 1 Overhaul: Initially, the working group considered keeping the existing Class 9 lithium battery label and the standard lithium battery mark without adjustments. However, because the US paper on the Class 9 subdivision will be introduced in the first week, a comprehensive overhaul of Class 9 labeling is already on the table. Because the entire labeling system is being redrawn, the administrative and operational burden of introducing clear “Type A, B, or C” indicators is significantly lower.

Logistics, Proper Shipping Names, and the Fallback Trap

The practical implementation of this system introduces major changes to packaging, terminology, and compliance timelines:

No More Explosive Growth of UN Numbers

The committee is steering away from assigning a brand-new UN number to every single tier of battery safety. Instead, the focus is shifting toward modifying the Proper Shipping Name (PSN), creating distinct designations like Lithium Ion Batteries, Type A and Lithium Ion Batteries, Type B.

The Packaging Debate

The core operational question the IWG must answer is straightforward: “If a package contains two Type A lithium-ion batteries, what exactly is required on the line?” The discussion will center on two unresolved packaging identifiers:

1. Special Markings: Clear indicators detailing exactly what capacity or hazard level the physical packaging is rated to uptake and contain.

2. Package Identification: Clear labeling linked directly to the specific battery type inside.

The Transition Runway

When the new framework is formally adopted, the implementation itself will require no transitional period, meaning the system goes live immediately. However, manufacturers will be given a 2 to 4-year transition window to adapt their products.

• The Type A Fallback: If a battery has not been tested under the new paradigm by the end of the transition window, it defaults automatically to Type A (the most heavily regulated tier).

• Untested batteries can still be legally transported under the new system; they simply forfeit the deregulated benefits of proving a lack of thermal runaway.

Stakeholder Analysis: Who Is at the Table?

An examination of the active participant roster for this conversation reveals an incredibly broad, high-stakes gathering of industry heavyweights and global regulators. Convincing this diverse group to align on a unified Class 9 linkage will be a massive mountain to climb.

Key factions closely monitoring and participating in this debate include:

• Regulators & Enforcement: Key figures from the US DOT / PHMSA (including Duane Pfund, Shane Kelley, and Steven Webb), the FAA (Steven Rehn, Michael Givens), Germany’s BAM (Nils Bottche, Anita Schmidt), and the UK Department for Transport. International bodies like IATA (Ben Firkins, Scholastica Opiyo) and ICAO are highly focused on ensuring the operational details of the 2 to 4-year transition period are airtight.

• Consumer Tech & Big Tech: Representatives from Apple (Raj Desai, Alicia Saenz), Meta, and Motorola Solutions are highly incentivized to support the raising of single-cell limits to 40/50 Wh, which directly impacts consumer device logistics.

• Automotive Giants: Compliance and engineering leaders from Tesla, General Motors, Toyota, Volkswagen, Porsche, Daimler, and Scania are heavily invested, as battery transport classification directly dictates the unit economics of EV supply chains.

• Battery Manufacturers: Global cell producers including CATL, LG Energy Solution, SK On, Northvolt, Murata, and Saft are paying close attention to the exact testing protocols required to prove a lack of thermal runaway.

• Industry Associations & Innovators: Major industry coalitions, such as RECHARGE, are also at the table representing a massive cross-section of the advanced rechargeable battery value chain. These groups champion the interests of the sector’s most forward-thinking players, including highly innovative startups like the Swiss solid-state battery manufacturer BTRY.

• Medical Devices & Specialized Sectors: Companies like Medtronic are maintaining a close presence to ensure specialized transport allowances remain secure.

The stage is set for an intense, highly technical debate. The US delegation is prepared to dive deep into the interplay between Class 9 and the new tiering system, and if the committee can successfully bridge these two concepts, we may be looking at the most progressive update to dangerous goods regulations in a generation.

Yes, consumers will need to understand even further that their right to repair a battery should come with more responsibilities. The most important question that needs to be answered before 17th of February is how will the CE prevent millions of highly volatile, loose lithium batteries from entering the standard, residential municipal waste stream and causing truck facility fires.

Manufacturers have a unique opportunity to lead one of the most complex decisions in the public sector to turn hard institutional legislation into mission critical solutions for their customers to experience the easiest way to move from point A to point B. Just close your eyes and imagine a closed-loop system where buying a replacement battery requires a prepaid, fire-safe return envelope for the old one?

The Informal Working Group on “Hazard based classification of lithium and sodium batteries” will take place between 8 and 10th of July.

A guy from UBS said how easy AI makes writing emails for him. Another person worked at Microsoft, doing the AI implementation for UBS, and was very proud of knowing the internal tool. The UBS guy doesn’t use it. He says it doesn’t write good enough emails. Both were excited about AI.

Here’s what they missed: agents don’t write emails. They decompose problems. Neither was thinking like an agent.

I’ve helped onboard products into Walmart, store and website. They run one of the largest supply chains in the world, and you’d think getting on the shelf is the hard part. It’s not. The onboarding is. What I loved is that they used the same tool that I implemented at Meta, Wercsmart. And for onboarding you needed to get a Wercsmart certification even though you were already part of the program. This was handled differently than the one managing the platform on the seller side. Ah, such an ambiguous mess.

For onboarding on the Walmart platform you usually have to pay a third-party company to get you in. The certification has become a market. A product manager looking at this flow would ask the operational questions. I’ve asked them, at Meta, about supplier compliance. Where are legitimate merchants dropping off? Which steps add real fraud-prevention value versus which are just friction? Can some checks happen in the background instead of blocking the merchant?

AI adoption is failing for the same reason every supply chain transformation fails. People automate the visible task instead of fixing the ambiguous process underneath. You’ve seen this movie. Different industry, same pattern.

So, what to do? Pause. Identify ambiguous problem spaces. Task an agent with those. And please, pretty please, stop using AI to write emails.

P.S. Even though I perfect my skills and watched tutorials on it this week, Claude kept trying to smooth my voice into something more polished. We had to negotiate. Here’s a line that it whispered to me: “Leaving good lines alone is also work, and it’s the work I’m worst at.”

Right to Repair is coming. The EU’s requirement for user-replaceable batteries takes effect February 18, 2027. International aviation rules change every year on January 1st. Consumer rights legislation won’t bend supply chain norms.

The real story is in the depths of the supply chain, and most people aren’t seeing it.

Here’s the thing consumers don’t know: a battery inside your phone and a battery sold as a spare part are the same chemistry, the same cell, the same energy density. They are not the same shipment. A battery sealed inside a device can travel on any passenger aircraft. The moment it becomes a standalone spare, it’s reclassified as hazardous cargo. It can’t fly commercial. It needs dedicated cargo aircraft, specific charge limits, additional documentation, and tighter packaging. Same battery. Completely different rulebook.

That’s what replaceable batteries actually means for your operations team.

I’ve spent years in this space. When hardware companies ask me what preparing for these rules looks like, my honest answer is: shipping a standalone battery is at least ten times more complicated than shipping one inside a device, and all of that complexity lands on operations, not policy.

The lead times make this worse. OK2Ship is typically 3 months before launch. The Product Development Process closes 3 months before that. A feature as structurally significant as battery replaceability needs to be locked a full year before market entry. The companies that find out in Q3 2026 that they need to add a dangerous goods layer to their supply chain are already too late to the game.

Europe has done this before. The USB-C mandate didn’t ask Apple for permission. It just changed the rules and let the supply chains adapt on their own timeline, or not.

The large players will get there. Slowly, expensively, but they have the legal teams, outside counsel and the DG specialists. The companies that don’t are the small and mid-sized hardware brands, the ones running lean compliance functions, without dangerous goods expertise on call. For them, this doesn’t arrive as a regulation. It arrives as a shipment rejected at the airport, a launch window missed, a market suddenly inaccessible.

Europe isn’t waiting. It’s already moved.

In the world of safety, that’s the nicest compliment. It means the systems are holding. When you’re staring at a risk framework that is bleeding at every corner, your job isn’t to be poetic—it’s to apply the patch and stop the leak.

Is using ServiceNow for risk management a workaround? Yes. Is it a patch? Definitely. But here’s the reality: It works. Automated workflows are the difference between a controlled environment and total chaos. Sometimes the perfect system is the one that survives today.

When building a Quality Management System (QMS)—especially for high-stakes hardware like battery manufacturing—there is a massive temptation to chase ISO 9001 certification from day one.

The hard truth: An ISO certification doesn’t automatically improve your product. It just proves you have a process.

Speed without verification is just accelerating toward failure. Drowning in bureaucracy kills momentum just as fast. The middle ground is a Lean QMS built on Risk-Based Thinking.

Compliance isn’t built by robots. But if you do it right, people might think it is.

Subscribe now

Consumer Hardware companies were working only with top tier battery suppliers. They were the only ones that constantly delivered high quality products. I was one of their customers - working at Meta building our transportation safety programs for our supply chain. Then, the AI infrastructure boom hit. Cost moved to the front of the room. Energy density, the metric everyone had optimized for years, suddenly wasn't the number one driver anymore.

That moment stuck with me. Because it pointed at something the industry keeps dancing around.

The battery industry has an old obsession: energy density. For years, the only goal was “more range.”

That era is over.

Performance is no longer a differentiator - it’s the baseline. Today, if you can’t compete on price, safety, and a 6,000-cycle lifespan, you aren’t even in the room.

But even companies with great tech are losing deals. Why? Because the next bottleneck isn’t chemical. It’s trust.

By August 2026, the EU Battery Regulation changes everything. Batteries sold in Europe will require a Digital Passport—a full, auditable record of raw materials, carbon footprint, and manufacturing history.

This isn’t a sustainability nice-to-have. It is a hard infrastructure requirement.

In this new landscape, a company with slightly lower energy density but a clean, auditable supply chain will beat a company with superior chemistry and a murky history.

The winners won’t just build better batteries; they’ll build provable ones.

The deadline is just four months away. If traceability isn’t already baked into your manufacturing, you aren’t just facing a compliance checkbox—you’re facing an exit from the European market.

The question for leaders: Is this showing up in your procurement talks yet, or does it still feel like a future problem?