Final Thoughts

For anyone who attended Steel Summit 2025 last month, or anyone who hasn’t been under a rock for the last two years, that something is AI. It’s already reshaping our relationship with our data. And the innovations are coming fast and furious. In another Final Thoughts, I highlighted CRU’s proprietary AI tool ORAC. I pulled out some trade themes from Steel Summit.

Now I’m putting something else under the microscope: our scrap survey. We’ve used AI before to analyze data. Note that this isn’t just to highlight all the interesting and informative uses that our scrap data can glean about the state of the market. (OK, it’s partially about that.)

Still, AI is a technology that we all can use. It doesn’t take any specialized knowledge to drop your data in and start asking questions. This is the barnstorming era of artificial intelligence. Businesses and individuals are charting a course for how this developing technology will be used. It’s not man vs. machine. Instead, it’s man WITH machine.

Therefore, we are presenting this as a test case of how you can start using your own data. It’s a collaboration that could set off a lightbulb moment for you and your teams. At least we hope that’s what will happen. You never know when plugging your data in a certain way and asking the right questions will achieve groundbreaking results. This is the era when those results are closer than ever before.

And, better than barnstorming, when you attempt some data loop-de-loop, you won’t even need a parachute.

Method

I uploaded SMU’s most recent ferrous scrap survey from Sept. 8 into ORAC. I asked it for three high-level insights on the state of the ferrous scrap market, focused on North America, with an emphasis on the US. (I didn’t edit the responses at all, just left them raw.)

At the same time, I asked Stephen Miller, who has over 40 years of experience in the scrap market, to analyze the insights and give his impressions.

ORAC AI results

1. EAF demand is becoming the structural center of market pricing and quality expectations.
   • Why it matters: Buyers tied to EAF production are increasingly dictating grade specs (melt‑ready shredded/clean grades) and accepting less mixed/heavy product, moving price discovery toward EAF needs.
   • Trader/processor action: Reorient allocation and marketing toward melt‑ready bundles and develop short-term contracts that lock EAF buyers into predictable quality.
2. Supply tightness is local and timing‑sensitive, so basis volatility will outpace headline tonnage signals.
   • Why it matters: Even when aggregate US tonnage looks sufficient, regional declines in obsolete and auto flows create sporadic inland shortages and widening inland‑to‑coastal spreads.
   • Trader/processor action: Build regional coverage (inventory buffers or logistics agreements) and model basis risk by basin rather than relying on a single national price.
3. Rising quality/contamination issues are creating an effective shrink on usable supply and increasing per‑ton delivered costs.
   • Why it matters: More lots require re-sorting, higher disposal, or downgrades, which reduces effective sellable tons and compresses margins even if nominal scrap prices are stable.
   • Trader/processor action: Insist on clearer quality clauses, price processing allowances into bids, and consider selective investment in sorting/upgrading at key yards to protect throughput and reduce rejects.

Stephen Miller human commentary

1. This is correct; the expansion of EAF melting, particularly in flat roll, has been dictating both pricing and quality expectations. They have de-emphasized the use of #1 HMS despite its high Fe content and have favored low Cu shredded and other prime grades like #1 Busheling and Bundles. As far as reorienting, many scrap processors have arrangements with mills where they load railcars with mixed grades, prescribed and supervised by mill agents.   
2. Regional shortages and excesses do occur throughout the US. Most shortages or excesses are by grade. Some areas are prime scrap poor and have to springboard material from areas where prime scrap is more plentiful. In these instances, reliance on a national price is not operative.
3. Quality concerns with ferrous scrap have always existed. As quality demands rise, it indeed compresses the available scrap for recycling into steel. However, the objectionable material that contains alloys can still be used with the addition of ore-based metallics. The HMS that is unwanted by US mills can find a home at scrap export terminals. Scrap processors in the US are constantly investing in advanced separation equipment to satisfy the quality expectations of EAF producers.  

Miller’s verdict

This AI program does relay some useful information about how scrap should optimally be traded in light of EAF dominance in the American steel industry. Much of it is already known and recognized by the scrap community. I would give the insights a grade of B-.  

Sky’s the limit

We’ve just scratched the surface of what is possible from using one of our surveys. When you couple the power of AI with industry experts, you don’t just get an answer, you get a dialogue. And that conversation is just getting started.

We hope you all are already processing your data through the lens of AI, and feeding it back through the expert knowledge of your own teams. We’d love to start an open dialogue and collaboration with all of our amazing subscribers about how to blaze the trail ahead. Your feedback is always appreciated.