Discover Pandipedia

Decarbonizing heavy industry: green steel and cement

Steel and cement are difficult to decarbonize because their emissions come not only from energy use but also from how iron and cement ingredients are made. This brief compares emerging production routes, what pilot and scale-up evidence is available, the cost picture, and policy tools that can help early projects reach commercial scale.

Bottom line: Hydrogen-based direct reduced iron (H₂-DRI) has progressed from pilots toward planned industrial demonstrations, while molten oxide electrolysis (MOE) has a company-reported industrial-cell milestone but limited public detail on output. For cement, the evidence is stronger for reducing clinker use or emissions than for proven clinker-free alternatives: performance and deployment information on alkali-activated binders and geopolymers remains sparse.

Steel: hydrogen reduction and electrolysis

In H₂-DRI, hydrogen is used to reduce iron ore to sponge iron, which can then be made into steel in an electric-arc furnace (EAF). HYBRIT’s Luleå pilot began operating in 2020 and produced about 100 tonnes of hydrogen-reduced sponge iron. HYBRIT also reports that its 2018–2024 pilot programme tested the production chain through crude-steel production in an EAF, reporting 0.0 tonnes CO₂e per tonne for Scope 1 and 2 in the pilot; that figure is the company’s reported result, not independently verified in the available evidence.^[1][2][3][4]

The next step is industrial scale, but planned capacity should not be confused with operating capacity. HYBRIT describes a planned Gällivare demonstration using 100% hydrogen, a 500 MW electrolyser and a design output of about 1.2 million tonnes of crude steel per year. Its hydrogen-storage pilot offers supporting evidence on a key system cost: HYBRIT reports practical savings of 26–31% from storage and estimates 25–40% savings for future commercial plants, while further testing is planned through 2026.^{[5][6][7][8][9][10]}

Molten oxide electrolysis is a different steelmaking route that uses electrolysis rather than hydrogen reduction. Boston Metal says it commissioned an operational, multi-inert-anode industrial MOE cell at its Woburn facility in 2025 and produced “tonnage steel”; it does not report the output tonnage or cell capacity. A separate demonstration plant is planned, but the source gives no specific capacity or commissioning date.^{[11][12][13][14]}

The IEA’s 2025 assessment places hydrogen-based steelmaking closer to early commercial deployment, but does not provide a formal technology-readiness level. It estimates early commercial plants using 100% hydrogen blends could cost 50–140% more than conventional blast-furnace/basic-oxygen-furnace (BF-BOF) plants, depending on region. The IEA also reports about 10 Mt of near-zero-emissions iron capacity for 2030, while warning that much announced capacity still lacks a clear timeline for operating fully near-zero-emissions.^{[15][16][17][18]}

Cement: clinker reduction versus clinker-free binders

Ordinary Portland cement relies on clinker, the kiln-made intermediate that contributes substantially to process emissions. The available evidence calls for care with labels: belite calcium sulphoaluminate (BCSA) is an alternative clinker, not a clinker-free chemistry. The IEA reports 20–30% lower process emissions than ordinary Portland cement clinker and places BCSA at TRL 7, but reports production limited to small batches rather than commercial production.^[19][20][21]

BCSA scale-up is constrained by limited standards coverage, the need for more cost-competitive formulations, and uncertainty about long-term durability and verified strength development. Blended cements can also cut clinker use by substituting materials such as calcined clay, limestone, volcanic ash, rice-husk ash or silica fume. The IEA estimates alternative constituents may account for roughly 15–35% by mass, potentially up to 50%, but availability, transport costs, prescriptive standards and market acceptance can limit adoption.^{[22][23][24][25]}

For genuinely clinker-free binders, the retrieved evidence is much thinner. CDP identifies alkali-activated cements as an example of alternative binding systems that avoid the process emissions inherent in Portland clinker production, but its reporting guidance describes possible development stages rather than confirming specific commercial deployments. A 2025 study reports 165.3 MPa compressive strength and 7.7 MPa tensile strength for one alkali-activated ultra-high-performance concrete formulation containing 40% slag powder and 28% fly ash as cement replacements; this is a particular laboratory result, not proof of field performance or commercial scale. The available sources do not establish comparable deployment evidence for geopolymers.^{[26][27][28][29]}

A separate cost benchmark illustrates the commercial hurdle but should not be treated as a cost estimate for every cement chemistry: the IEA estimates early commercial near-zero-emissions cement plants using carbon capture and storage (CCS) have production costs 75–150% above conventional plants, with regional variation. The estimate uses regional averages and 2024 energy inputs, excludes policy support such as carbon pricing or subsidies, and does not account for regional differences in capital and non-energy operating costs.^{[30][31][32][33]}

The available evidence does not provide a like-for-like cost curve comparing hydrogen steel, MOE steel, clinker-free binders and conventional production. For steel, energy prices, technology costs, raw-material availability and regional policy affect route economics; the IEA notes H₂-DRI can be competitive where renewable electricity is cheap. Scrap-based steelmaking uses around one-eighth the energy of production from iron ore, but scrap availability and quality are not quantified in the supplied material.^[34][35][36]

Comparison: readiness, costs and evidence gaps

The comparison below separates reported cost estimates from pilot or development evidence. The figures use different baselines and technologies, so they are not a direct ranking of the least-cost decarbonization options.

Policy levers and practical implications

The IEA identifies a combination of demand creation, risk-sharing and common rules as important because first-of-a-kind projects face high costs and risks, while thin margins and international competition make it difficult for producers to absorb a green premium.^[48][49]

Takeaway: The near-term opportunity is to scale routes with substantial pilot evidence while building demand and reducing first-mover risk. For steel, H₂-DRI has the more detailed pilot chain in the available material, while MOE has a promising company-reported industrial-cell milestone but little public cost or capacity information. For cement, clinker reduction and CCS have more developed evidence than clinker-free alternatives; claims about clinker-free performance or commercial readiness should therefore be treated cautiously until supported by comparable durability, standards and field-deployment data.^{[57][58][59][60][61]}

Ethical concerns around brain-computer interfaces

A brain-computer interface (BCI) records brain activity and translates it into commands or other outputs; some systems can also stimulate the nervous system. Their ethical stakes therefore include both what neural data may reveal and how a device may affect its user. The main concerns are meaningful consent, privacy and cybersecurity, unequal access, and rules that have not kept pace with different types of neurotechnology.

Consent, autonomy, and privacy

Consent can be difficult when users cannot readily understand what a device records or what later analysis might infer from it. Neural signals may be used to infer attention, emotions, reactions, or intentions, including information beyond the original purpose a person agreed to. For long-term or adaptive systems, researchers recommend consent practices suited to changing systems, clear reporting of data safeguards, and involving patients in research design.^[1][2][3]

The privacy concern extends beyond disclosure of raw recordings: analysis may enable intimate profiling for advertising, education, workplace monitoring, or other decisions. Users may also lose control if neural data or inferences are shared or used for purposes they did not agree to, putting mental privacy and cognitive autonomy at risk.^[4][5][6][7]

Security and potential harm

Neural data can be exposed through insecure devices or systems, intercepted during wireless transmission, or accessed through weak authentication and access controls. A breach could reveal sensitive cognitive patterns. For systems that can act on the nervous system, tampering or malicious inputs may also trigger unintended actions, raising health and safety concerns for people using therapeutic implants.^{[8][9][10][11]}

Researchers have proposed safeguards such as minimizing collected data, encryption, access logging, secure software updates, and independent testing. These are proposed measures, not uniform requirements across all BCI products.^{[12][13][14][15]}

Socioeconomic divides

If BCIs for treatment or cognitive enhancement are expensive, access may be concentrated among wealthier people and regions. In competitive settings such as schools or workplaces, unequal access could widen existing educational and economic gaps; people may also feel pressured to use enhancement technology to keep up. Researchers debate fair access and who should pay, while noting that universal access commitments could strain underfunded health systems if they ignore differences in resources between countries.^{[16][17][18][19][20][21]}

Policy gaps and current guidance

Current oversight is fragmented. Medical BCIs may be regulated as medical devices, but enhancement and consumer neurotechnology may not fit those pathways. In the United States, HIPAA protections depend on neural data being handled by covered health-care entities or their business associates, and medical-device cybersecurity guidance does not cover products classified as consumer electronics. State neural-data protections are emerging, but their definitions and approaches differ.^{[22][23][24][25]}

There are relevant guidelines and initiatives, but they do not amount to one comprehensive ethics code for every BCI use. FDA guidance addresses testing and clinical studies of implanted BCIs in a specific medical context. The OECD and UNESCO have issued neurotechnology ethics guidance or recommendations, while a review of closed-loop neurotechnology calls for tailored consent, transparent reporting of data safeguards and AI use, patient participation, and training for researchers, clinicians, and review boards.^{[26][27][28][29][30][31][32]}

The central takeaway is that BCI governance needs to address the whole lifecycle: meaningful consent, limits on data collection and reuse, strong security for both information and device functions, and fair access. Existing laws and guidance cover parts of that challenge, but protections remain uneven across clinical, consumer, and enhancement uses.^{[33][34][35][36]}

Adaptive AI copilots vs. text-only chatbots

An AI copilot helps with work by using relevant information from the user’s tools and task, while a text-only chatbot mainly responds to what the user types. The key difference is context: a copilot may draw on connected apps, work history, live operational data, or sensor inputs, then adapt its response or take a task-related action. The sources describe these capabilities, but do not establish that every copilot uses sensors or that one product combines all these signals.^[1][2]

Examples across work and daily tasks

Takeaway

A chatbot is mainly prompt-driven; a copilot can be context-driven and connected to the work itself. More context can make assistance better fitted to a situation, but “copilot” does not automatically mean sensor-equipped: the examples range from operational data and connected apps to work-pattern context.^[14][15][16]