AI Forecasting for Near-Zero Steel Procurement

Steel is getting a new kind of pressure test: not “can you make it strong?” but “can you make it clean—and prove it?” This week, a North America-focused collaborative request for proposal (RFP) for near-zero emission steel drew 13 commercial-scale project proposals aiming for delivery in the early 2030s, with buyers collectively seeking over 1 million tons of near-zero steel.

That’s the headline. The more useful story for energy and utility leaders is what sits underneath it: near-zero steel isn’t blocked by chemistry alone—it’s blocked by bankability. And bankability, in 2025, is increasingly a data problem.

In our AI in Energy & Utilities series, we usually talk about demand forecasting, grid optimization, renewable integration, and predictive maintenance. This steel procurement milestone touches all four. Near-zero ironmaking (especially hydrogen-based pathways) will be as constrained by clean power availability, transmission, and price volatility as it is by metallurgy. The organizations that can forecast, schedule, and contract around those constraints are the ones that will actually get projects financed.

Near-zero steel is a power-and-data problem, not just a steel problem

Near-zero steel scales when clean electricity is predictable enough to underwrite long-term contracts. Most proposed pathways—particularly hydrogen-based direct reduced iron (DRI)—depend on large volumes of reliable, low-cost renewable electricity. If you can’t show that power will be there (hour by hour, season by season), project finance gets expensive fast.

Here’s what’s changing: collaborative procurement mechanisms, like aggregated buyer RFPs, are starting to provide credible demand signals. But demand signals alone don’t keep an electrolyzer or an electric arc furnace (EAF) running economically. You also need confidence in:

• Hourly/seasonal power availability (renewables + firm capacity)
• Congestion and curtailment risk at the interconnection point
• Hydrogen production cost variability (electricity is the dominant input)
• Operational flexibility (what can ramp, when, and at what cost)

This is where utilities and energy providers can stop being “infrastructure in the background” and become active enablers of industrial decarbonization.

Snippet-worthy reality: If the steel sector is asking for near-zero steel, it’s also asking for near-zero power—delivered with near-zero surprises.

What the first North American near-zero steel RFP tells us

The market just demonstrated it will show up—on both sides—if the ask is credible. The inaugural Sustainable Steel Buyers Platform (SSBP) RFP aggregated demand from nine buyers targeting more than 1 million tons of near-zero emission steel. It attracted 13 first-of-a-kind proposals for early-2030s delivery.

That combination matters because it addresses the chicken-and-egg trap:

• Producers won’t build multi-billion-dollar assets without offtake certainty.
• Buyers won’t commit to premiums without credible supply and verification.

A notable shift: ironmaking is the main battlefield

Most emissions are locked upstream in ironmaking, so the proposals skewed toward iron-focused investments, not just “cleaner furnaces.” Expect more:

• Standalone near-zero iron production
• Cross-border “green iron” trade patterns
• Hybrid supply chains where iron is decarbonized before it reaches steel plants

For energy and utilities, this is a major load-shape signal: near-zero ironmaking can behave like a new category of industrial demand, with large flexible and inflexible components (electrolyzers, compression, EAFs, heat treatment, downstream processing).

Book-and-claim is quietly doing a lot of work

The RFP accommodated both physical delivery and book-and-claim transactions—where emissions attributes are decoupled from the physical steel.

This matters because supply chains for steel are messy. A data center developer, an OEM, and a construction contractor may not be able to trace “this beam from that mill” cleanly. Book-and-claim can still direct money to near-zero projects if the accounting is rigorous.

The AI tie-in: book-and-claim systems are only as trustworthy as their measurement, reporting, and verification (MRV). Scaling them requires clean data pipelines, anomaly detection, and auditable calculations.

Where AI fits: turning demand signals into bankable contracts

AI’s job here isn’t marketing—it’s reducing uncertainty. Uncertainty is what drives higher financing costs, conservative operating assumptions, and contract clauses that nobody likes.

Below are four practical ways AI in energy & utilities can directly support near-zero steel production—especially hydrogen-based DRI and upgraded EAF pathways.

1) AI-driven demand forecasting for “green commodity” procurement

Aggregated RFPs work best when buyers can commit to volume with confidence. Many steel buyers aren’t steel companies; they’re data center operators, renewable developers, manufacturers, and industrial equipment firms.

AI demand forecasting helps them move from aspirational targets to contractable demand by improving:

• Forecasts of construction starts, expansions, and retrofit cycles
• Materials demand planning by project portfolio
• Emissions accounting that ties procurement to Scope 3 reporting schedules

For utilities, there’s a parallel: you already forecast load growth. The new opportunity is collaborating with industrial customers to forecast decarbonized load growth—the portion of demand that must be served with clean attributes to satisfy buyer requirements.

2) Grid optimization for hydrogen-based DRI and electrolyzer load

Hydrogen DRI is only “near-zero” if the electricity feeding electrolyzers is low-carbon and available at scale. That creates a tight coupling between steel procurement and grid operations.

AI-powered grid optimization (and more broadly, advanced energy management systems) can:

• Predict congestion and recommend dispatch or siting changes
• Schedule flexible loads (electrolyzers) around renewable peaks
• Reduce curtailment by turning excess renewable output into hydrogen
• Improve interconnection planning using probabilistic generation models

A concrete example of what works in practice: treating electrolyzers as grid-responsive industrial demand, with automated setpoints that respond to price, carbon intensity, and constraint signals. This isn’t theoretical—utilities are already doing versions of this with large demand response. The difference is the scale and the stakes.

3) Predictive maintenance for high-availability, capital-intensive assets

Near-zero steel routes are capital intensive, and downtime kills economics. Electrolyzers, EAF retrofits, compression systems, and high-temperature downstream equipment all require high availability to meet take-or-pay obligations.

Predictive maintenance helps by:

• Catching degradation trends early (voltage drift, heat signatures, vibration)
• Reducing unplanned outages during high-price periods
• Improving spare parts planning for constrained components

Utilities can contribute here too: for industrial customers, power quality events, flicker, and voltage dips can be operationally expensive. Asset health monitoring paired with grid telemetry can reduce finger-pointing and shorten incident resolution.

4) MRV analytics: making “near-zero” measurable and comparable

The contract will only be trusted if emissions performance is computed consistently. The steel market is crowded with standards and claims; buyers need comparability.

AI (plus strong data engineering) can support MRV by:

• Automating data validation across meter feeds, production records, and certificates
• Flagging anomalies (sudden intensity drops, inconsistent batch attribution)
• Producing audit-ready emissions calculations with version control

Clear stance: The fastest way to stall the near-zero steel market is to let “near-zero” mean five different things in five different contracts.

A practical playbook for energy & utility leaders supporting near-zero steel

If you want near-zero steel to scale, treat it like a coordinated energy transition project, not a one-off industrial customer request. Here’s a pragmatic checklist I’d use when talking with steel producers, hydrogen developers, and large buyers.

Step 1: Model the hourly load shape—then stress-test it

Don’t accept annual MWh estimates. Ask for (or help produce) an hourly profile that includes:

• Electrolyzer operating strategy (baseload vs flexible)
• Storage assumptions (hydrogen, thermal, inventory buffers)
• Ramp rates and minimum stable loads
• Seasonal production constraints

Then run scenarios:

• Transmission upgrade delays
• Low-wind/low-solar weeks
• Price spikes and negative price periods
• Capacity shortfalls during extreme weather

Step 2: Offer contracts that match industrial decarbonization reality

Steel and hydrogen projects need confidence over 10–20 years. Utilities can help by structuring:

• Clean power products with hourly matching options
• Hybrid supply arrangements (renewables + firm clean + storage)
• Curtailment-aware tariffs for flexible electrolyzer loads
• Interconnection roadmaps with milestone-based commitments

Step 3: Build a shared data layer for procurement + operations

Near-zero steel procurement will increasingly sit at the intersection of ERP systems, energy market data, and carbon accounting. A shared data layer should cover:

• Metered electricity and delivered energy attributes
• Production volumes (iron/steel) and batch metadata
• Emissions intensity calculations and audit logs
• Forecasts (demand, prices, renewables output, congestion)

This is where AI becomes a force multiplier: once the data layer exists, optimization and forecasting become deployable, not hypothetical.

Step 4: Treat book-and-claim like financial infrastructure

Book-and-claim can help scale early markets, but only if it’s built with the seriousness of a payments system:

• Clear rules for issuance, retirement, and double-counting prevention
• Transparent methodologies for emissions attributes
• Controls that stand up to buyer scrutiny and regulator attention

Utilities and energy retailers can play a role by aligning clean electricity attributes with industrial product claims—especially when electrolyzers are the bridge between electrons and molecules.

What happens next: a flywheel, if the grid keeps up

The near-zero steel pipeline is real, but the grid is the pacing item. The RFP results show credible interest from legacy producers, start-ups, and industrial operators, spanning multiple pathways: hydrogen DRI, novel electrochemical or bio-based routes, sodium-based concepts, and EAF upgrades.

All of them share one dependency: reliable, affordable, low-carbon energy at scale. If transmission queues and interconnection delays remain the norm, early-2030s delivery targets will slip, premiums will rise, and buyers will hesitate.

From the energy & utilities side, I think the opportunity is straightforward: be the partner that makes clean industrial power predictable. Do that with AI-enabled forecasting, grid optimization, and asset reliability, and you’re no longer reacting to industrial decarbonization—you’re shaping it.

If you’re planning for 2026 budgets right now, here’s a useful internal question to ask: Which industrial loads in our territory will become “clean-attribute critical” by 2030, and what data do we need today to serve them confidently?