Artificial intelligence is the foundation of next-generation energy systems, and the global shift to green hydrogen and synthetic fuels is changing industrial decarbonization. The combination of AI with advanced manufacturing, digital twins, predictive analytics, and intelligent process optimization is emerging as a key competitive advantage as the industry looks for scalable, profitable ways to reach net zero goals. However, the problem goes beyond using complex algorithms. To turn renewable energy into bankable industrial solutions that can support long-term economic growth, a systems-level strategy integrating engineering, economics, sustainability, and governance is needed.
Oliver-Andreas Leszczynski, Deputy Chairman of the INER Institute of Northern-European Economic Research, is one of the few people who truly understands this confluence. He promotes a workable and auditable framework that views AI as a strategic operational model rather than a stand-alone technology and is well-versed in the economics of artificial intelligence-driven industrial transformation, maritime manufacturing, green hydrogen, and synthetic fuels.
In an exclusive interview with Insights Success, he discusses sustainable fuel production, AI-driven hydrogen innovation, the Cognitive Power-to-X Optimization Methodology, and the future of the global clean energy ecosystem.
Your career has taken you from AI-led transformation in complex maritime manufacturing to the economics of green hydrogen and synthetic fuels. What was the decisive insight that convinced you that artificial intelligence could become not merely a supporting tool, but a core industrial lever for making green hydrogen and e-kerosene commercially viable?
The decisive insight came from complex industrial reality rather than from a technology demonstration. My work in maritime manufacturing taught me that performance is rarely lost because one individual component is fundamentally inadequate. Value is usually lost at the interfaces: between engineering disciplines, production stages, suppliers, data systems and human decisions. A ship may contain excellent components and still be delivered inefficiently if the industrial system surrounding them is fragmented. My work in industrial AI, design transformation and predictive maintenance has therefore always focused on the orchestration of the entire value chain rather than on isolated algorithms.
Green hydrogen and e-kerosene present the same systems problem in an even more concentrated form. A plant may use first-class electrolyzers, storage systems and synthesis technologies and still fail economically if renewable-power availability, electricity prices, stack degradation, hydrogen and CO₂ inventories, maintenance requirements, fuel quality and certification obligations are optimized separately.
That realization led me to formalize what I call the Cognitive Power-to-X Optimization Methodology, or C-PtX. It is an auditable industrial method with six essential stages: establishing a measurable counterfactual baseline; mapping energy, material, carbon and cash flows; constructing a hybrid digital twin; optimizing the integrated system under hard safety, quality, degradation and certification constraints; introducing AI-supported control in carefully governed stages; and independently verifying the resulting economic and environmental performance.
My definition of success is therefore not the existence of an AI model. Success is a reproducible improvement in plant economics, reliability and lifecycle integrity. The strategic shift is from AI-enabled individual assets to AI-orchestrated industrial value chains. That is where artificial intelligence can help transform green molecules from a policy aspiration into bankable industrial output.
The hydrogen sector is rich in ambition, yet often vague about where artificial intelligence creates genuine economic value. Across the full power-to-liquid chain, where can AI change the economics most dramatically, and where is the industry confusing digital theatre with real optimization?
The greatest value does not usually lie inside a single process unit. It lies in coordinating the interfaces between units and making better decisions across different time horizons.
The first major value pool is the joint optimization of renewable generation, power procurement, storage and electrolyzer dispatch. The cheapest hour of electricity is not necessarily the most economical hour of hydrogen production if operating at that moment accelerates degradation, creates an inventory imbalance or forces a downstream process into an inefficient regime.
The second value pool is degradation-aware asset management. AI can help estimate states that are difficult to measure directly, identify emerging performance losses and determine whether additional production today justifies its effect on future stack or catalyst life. The third is integrated plant scheduling: coordinating hydrogen production, CO₂ availability, intermediate storage, synthesis campaigns, maintenance windows and product delivery. Further value comes from predictive maintenance, anomaly detection, yield optimization, off-spec prevention, supply-chain planning and automated carbon-accounting evidence.
The industry enters the realm of digital theatre when it mistakes visibility for optimization. A dashboard is not an optimization system. A chatbot connected to plant documents is not an industrial control strategy. A forecasting model with impressive statistical accuracy has little economic value unless it changes a dispatch decision, maintenance intervention, process setpoint or commercial commitment.
I apply a simple test. What decision does the model change? Which physical and safety constraints does it respect? Against which counterfactual is its contribution measured? Who remains accountable for the decision? Unless those questions can be answered precisely, one is probably looking at a digital presentation layer rather than industrial intelligence.
AI should not merely explain what happened yesterday. It should improve the quality of a consequential decision today and demonstrate the value of that decision tomorrow.
You have consistently approached AI as an industrial operating model rather than as a stand-alone application. What would your ideal cognitive production architecture for a green-hydrogen and e-SAF facility look like, and how would it combine physics-informed digital twins, machine learning, real-time control and human engineering judgment?
My ideal architecture consists of six connected layers. The first is the physical and engineering layer: renewable-generation assets, electrolyzers, water treatment, compression, storage, CO₂ conditioning, synthesis, upgrading and utilities. Every optimization must begin with the thermodynamics, electrochemistry, equipment limits and safety requirements of these assets.
The second layer is a trusted industrial data fabric. It must connect operational technology and enterprise systems while preserving time synchronization, data provenance, quality controls, cybersecurity and clearly defined ownership. Without reliable data lineage, an apparently sophisticated model can become an industrial liability.
The third layer is the hybrid digital twin. First-principles models define what is physically possible. Machine-learning models estimate unobservable states, represent complex degradation behavior and correct residual errors where purely mechanistic models are insufficient. AI should complement physics, not compete with it.
The fourth layer is hierarchical decision intelligence. A planning optimizer operates across hours, days and weeks, considering weather forecasts, energy contracts, storage, maintenance and delivery obligations. Beneath it, model-predictive control manages process variables over seconds and minutes. This division is essential because strategic scheduling and real-time control are different decision problems.
The fifth layer is an assurance system covering functional safety, cybersecurity, model drift, uncertainty, authorization and human override. No black-box model should possess unconstrained authority over a safety-critical asset. AI may recommend, simulate or control within a validated envelope, but accountability must remain explicit.
The sixth layer is an integrated economic, carbon and certification ledger. Every operational decision should be evaluated not only in terms of throughput, but also in terms of cost, asset life, emissions intensity and regulatory eligibility.
C-PtX is therefore not a single algorithm or vendor platform. It is a modular, vendor-neutral and auditable operating architecture. Human engineering judgment remains central because the objective is not autonomous production at any price. The objective is better industrial judgment, exercised continuously and at scale.
Renewable-power variability, electrolyzer degradation, storage constraints and downstream fuel specifications can pull an integrated production plant in conflicting directions. How would you design a degradation-aware, multi-objective control system that lowers production costs without sacrificing asset life, plant availability, safety or fuel quality?
The first step is to reject the misleading objective of minimizing instantaneous electricity consumption or maximizing current hydrogen output. The correct objective is to maximize the plant’s risk-adjusted value over its operating life.
I would begin with a state-of-health estimator that combines voltage behavior, temperatures, pressures, gas purity, load history, start-stop cycles, maintenance records and other diagnostic data. The purpose is not merely to predict failure. It is to estimate the marginal cost of degradation associated with a particular operating decision.
Above this estimator, a stochastic planning system would evaluate renewable forecasts, electricity prices, hydrogen and CO₂ inventories, delivery obligations, maintenance requirements and uncertainty. It would determine the preferred operating corridor for the coming hours and days. A lower-level model-predictive controller would then manage variables such as current density, temperature, pressure, ramp rates, purity and storage flows inside that corridor.
The objective function would include electricity, water, maintenance, replacement, curtailment, off-spec production, lost availability and carbon-compliance costs. Safety, gas purity, pressure limits, thermal limits and final-fuel specifications would be hard constraints, not negotiable terms in a cost function.
A particularly useful concept is the shadow price of degradation. It expresses the expected future economic consequence of consuming additional asset life today. The operator can therefore compare the immediate revenue from higher production with its effect on replacement expenditure, availability and future output.
The system would initially run in simulation and shadow mode, followed by advisory operation and only then by supervised closed-loop control. AI must never be allowed to learn safety boundaries through uncontrolled experimentation. Dynamic operation, durability and cost are recognized as interdependent electrolyzer-development challenges; that is precisely why operating strategy must incorporate degradation rather than treating it as an afterthought.
Investors do not finance algorithms; they finance auditable risk reduction and credible cash flows. Which operational and financial metrics would you require before claiming that an AI intervention has materially improved a hydrogen or e-SAF project?
The first discipline is to distinguish a model metric from an investment metric. A lower forecasting error or higher machine-learning accuracy has no balance-sheet value unless it improves a consequential industrial decision.
For green hydrogen, I would examine the levelized cost of hydrogen, electricity consumption, availability, water consumption, purity, curtailment utilization, degradation rates, maintenance expenditure and expected stack-replacement intervals. For e-SAF, I would add hydrogen and carbon utilization, conversion yield, product selectivity, off-spec production, catalyst performance, lifecycle carbon intensity and the minimum sustainable selling price.
At plant level, I would measure unplanned downtime, inventory stability, schedule adherence, production variance and the frequency with which downstream assets are forced outside their preferred operating range. At financial level, the decisive measures are not only net present value and internal rate of return, but also downside resilience, debt-service capacity, sensitivity to electricity prices and the distribution of expected cash flows.
However, the metrics alone are insufficient. The verification design matters just as much. Before implementation, the system boundary, baseline, counterfactual, measurement period and relevant external variables must be defined. The model should be tested in shadow operation before a controlled rollout. Improvements must survive different weather conditions, feedstock regimes and production campaigns. Wherever possible, results should be reviewed by an independent technical party.
I would require an intervention to pass three tests. The effect must be statistically credible, operationally material and financially bankable. A result that is statistically detectable but commercially irrelevant is not an industrial success. Nor is a short-lived improvement that disappears after conditions change.
My principle is straightforward: no baseline, no benefit; no audit trail, no bankability. C-PtX is designed to make that discipline part of the technology itself.
An e-kerosene plant may be highly efficient at the level of individual process units and still fail the wider climate-integrity test because of its electricity sourcing, the origin of its CO₂, conversion losses or inadequate traceability. How can AI optimize and verify the entire lifecycle so that economic efficiency, regulatory compliance and genuine emissions reduction reinforce rather than undermine one another?
A molecule is only as green as its full chain of evidence. Efficiency inside the plant cannot compensate for carbon-intensive electricity, an unsuitable CO₂ source, unrecognized transport emissions or double counting.
I would therefore create a digital carbon and compliance twin alongside the physical process twin. It would record the origin and timing of electricity, electrolyzer operation, water and treatment requirements, CO₂ provenance, capture energy, intermediate storage, conversion losses, transport, co-products and final fuel output. Every batch would receive an auditable digital product record connecting physical production data with lifecycle and certification evidence.
The optimizer would treat lifecycle carbon intensity and regulatory eligibility as operational constraints rather than retrospective reporting indicators. It could decide, for example, that production during a particular period is economically attractive but would jeopardize the product’s renewable-fuel status or exceed a carbon-intensity threshold. In that case, the correct decision may be to reduce production, use storage or change the production schedule.
This is particularly important for products intended for Europe. European rules for renewable fuels of non-biological origin include requirements concerning renewable electricity, additionality and temporal and geographical correlation. These requirements can also be relevant to fuel produced outside the European Union when it is intended for the European market. ReFuelEU Aviation has furthermore created specific demand for synthetic aviation fuels and requires eligible fuels to meet sustainability and lifecycle criteria.
AI can automate evidence collection, identify inconsistencies, forecast compliance risk and optimize production around certification constraints. But it cannot convert a structurally unsustainable process into a sustainable one by relabeling the data.
Climate integrity and economic performance ultimately reinforce one another. A product that cannot withstand regulatory or scientific scrutiny faces discounting, rejected certification and stranded-market risk. Traceability is therefore not administrative overhead. It is part of the product’s economic value.
Europe is creating durable demand for sustainable and synthetic aviation fuels, while Chile and Argentina are seeking export-oriented industrial growth. What should a new Europe–South America energy compact look like if it is to combine European demand and technology with South American renewable resources, local value creation, skills transfer, credible certification and sovereign control of industrial data? Where do you see your own role in building that bridge?
Such a compact should not be another memorandum of understanding. It should be a market-making institution.
First, it requires predictable demand. European airlines, fuel suppliers and public institutions should aggregate long-term e-SAF demand through credible offtake structures, auctions or risk-sharing instruments. Producers cannot finance first-of-a-kind plants on the basis of broad political ambition alone.
Second, the compact should coordinate shared infrastructure. Renewable generation, water systems, transmission, CO₂ supply, storage, ports and certification facilities must be planned as industrial corridors rather than as disconnected private projects.
Third, local value creation must be contractual. Chile and Argentina should not be limited to exporting renewable potential in molecular form. Local engineering, operations, control centers, maintenance, supplier qualification, applied research and professional education should be integral parts of each project.
Fourth, the partnership needs a federated industrial-data architecture. Raw operational data should remain under the control of the asset owner and the relevant national jurisdiction. Shared models, performance indicators and certification evidence can be exchanged through governed interfaces without requiring countries to surrender industrial sovereignty.
Europe has established legally anchored demand for synthetic aviation fuels. Chile’s 2026–2030 strategy emphasizes domestic demand, export development and local value creation, while Argentina’s current investment agenda explicitly includes green and low-emission hydrogen among the industries targeted for expansion
My role would be that of an independent industrial systems architect and initiator. I would propose a Europe–Southern Cone Clean Molecules Demonstration Alliance, bringing together governments, universities, technology providers, energy companies, airlines, ports, certification bodies and financiers.
I would contribute the C-PtX reference architecture, a common verification protocol and a portfolio approach to demonstration projects. I am not interested in exporting a technological black box. I want to help create a transferable industrial capability that Chile, Argentina and their European partners can jointly govern, challenge and improve.
Chile’s updated 2026 hydrogen strategy places renewed emphasis on domestic demand, export capacity, local value creation and green derivatives, while Magallanes and Antofagasta present very different renewable-resource, water, logistics and territorial conditions. If the Chilean government invited you to design one flagship AI-enabled hydrogen-to-e-SAF demonstrator, what would you build, where would you locate it, and which milestones would have to be achieved within twenty-four months to justify industrial-scale investment?
I would locate the first flagship in Magallanes, while designing its digital and control architecture for later replication in Antofagasta.
The project would be a modular, highly instrumented power-to-liquids demonstration platform. Its purpose would not be to maximize headline production volume. Its purpose would be to demonstrate bankability: reliable operation under variable renewable conditions, degradation-aware electrolysis, coordinated hydrogen and CO₂ management, stable synthesis, certifiable product quality and complete lifecycle traceability.
A 10-to-20-megawatt-class electrolyzer would be sufficiently substantial to generate industrially relevant operating data while remaining compatible with a staged demonstration approach. The plant would include hydrogen storage, an eligible and traceable CO₂ pathway, a modular synthesis train, advanced sensing and a C-PtX control environment. The design should remain technology-pathway neutral until a rigorous comparison of synthesis routes, carbon availability, water requirements, certification and scale-up economics has been completed.
Within the first three months, the project would require a public-private consortium, a data-governance charter and a clearly defined baseline. By month six, it should possess an integrated engineering and economic model, a permitting roadmap, a lifecycle-certification concept and a local skills program. By month twelve, the hybrid digital twin and hardware-in-the-loop control environment should be validated. By month eighteen, procurement, site integration and supervised shadow operation should be substantially advanced. By month twenty-four, the project should have completed a representative operating campaign or reached an equivalently robust commissioning stage, accompanied by an independently reviewed scale-up case.
The decisive outputs would be verified production economics, degradation behavior, availability, fuel quality, lifecycle carbon intensity, water performance and a bankable industrial-scale design. Chile’s updated strategy already treats e-fuels, local capability, shared infrastructure, certification and regional hubs as central elements of the next development phase.
Magallanes would host the first demonstration, but the architecture should be transferable to an Antofagasta configuration combining solar resources, mining demand and industrial decarbonization. The true national asset would not only be the pilot plant. It would be the validated method for replicating it.
Argentina has announced an expanded large-investment incentive framework that explicitly includes green and low-emission hydrogen, while simultaneously advancing export infrastructure, synthetic-fuel applications and international work on hydrogen origin and quality standards. If you were advising the federal government and the Patagonian provinces, which priority should come first: national resource and site optimization, an interoperable industrial data space, port and logistics planning, certification infrastructure, domestic offtake or a bankable e-SAF pilot; and how would you sequence the remaining priorities?
Argentina should begin with national resource-to-market optimization implemented through a sovereign, federated industrial data space. These are not two separate priorities. The optimization cannot be credible without the data architecture, and the data architecture has little value without a defined investment decision.
I would establish a Federal Hydrogen and E-Fuels Intelligence Platform. It would integrate renewable-resource profiles, grid conditions, water availability, potential CO₂ sources, environmental sensitivities, ports, pipelines, roads, workforce capabilities, domestic demand, permitting conditions and international certification requirements. The information would remain under appropriate federal, provincial and corporate control while supporting transparent comparison of potential industrial corridors.
The second step would be to identify a small number of priority corridors rather than allowing dozens of disconnected projects to compete for infrastructure. The third would be to establish certification and measurement rules at the design stage. The fourth would align ports, transmission, water and logistics with the selected corridors. Domestic offtake should then be developed around applications capable of providing early demand and operational learning.
Only after this process would I select the location and configuration of a Patagonian e-SAF pilot. A pilot chosen before completing the resource, infrastructure and market analysis risks becoming politically visible but economically isolated.
Argentina’s 2026 Super RIGI announcement explicitly includes green and low-emission hydrogen and offers enhanced fiscal incentives. Such incentives can improve investment conditions, but they cannot identify the optimal site, create an offtake market or produce certifiable molecules. Argentina’s official hydrogen planning has also recognized the importance of synthetic fuels, domestic and export markets, infrastructure, pilots and industrial capabilities.
Looking toward 2030 and 2035, what single scientific and industrial achievement do you want the name Oliver-Andreas Leszczynski to stand for in the hydrogen economy, what measurable result would prove that the mission had succeeded, and what coalition of governments, universities, technology companies, airlines and energy producers must be assembled now to make it real?
I want my name to stand for transforming industrial AI from a promising concept into a verified operating discipline for the green-molecule economy.
The specific achievement would be the establishment of C-PtX as an internationally recognized and independently testable methodology for optimizing green hydrogen and synthetic-fuel production. It should enable different technologies, countries and industrial partners to work within a common framework while retaining their technical autonomy and data sovereignty.
By 2030, I would want to see independently audited demonstrations in three contrasting industrial environments: one in Europe, one in Chile and one in Argentina. Together, they should demonstrate at least a ten-percent improvement in risk-adjusted production economics relative to pre-registered baselines, achieved through lower controllable costs, higher availability, reduced off-spec production or better asset utilization. That improvement must not be purchased at the expense of safety, asset life, fuel quality or lifecycle carbon integrity.
By 2035, success would mean that the methodology is no longer dependent on me. It should be incorporated into public demonstration programs, financing assessments, university research and industrial procurement. Other experts should be able to challenge it, reproduce it and improve it.
The necessary coalition must unite energy and economic ministries, regional governments, universities, electrolyzer and synthesis-technology providers, renewable developers, engineering companies, airlines, airports, ports, certification bodies, financial institutions and the communities in which these projects are built. Each participant possesses only part of the solution. Leadership consists in creating an architecture within which those parts become an effective industrial system.
I do not want to be remembered for having described the hydrogen transition eloquently. I want to have helped give it an operating system and to have demonstrated, with evidence, that it works.
