Resilience in the Age of AI
Transforming Heavy Industry, Energy, and Infrastructure
Introduction – A Climate Investor’s Journey to Resilience
In recent months, I’ve found myself returning to a single word in my investing pursuit: resilience. As an investor focused on climate solutions, I’ve been exploring how industries can not only decarbonise but also adapt amid uncertainty – whether that’s volatile energy markets, supply chain shocks, or the physical impacts of climate change. Resilience has become a core theme in my exploration as an investor, and a curious question keeps arising: How can emerging technologies bolster the resilience of our heavy industries, energy systems, and infrastructure in the face of climate change? Lately, one technology in particular stands out as a catalyst – Artificial Intelligence (AI). The recent IEA 2024 Energy and AI report and real-world case studies have convinced me that AI is starting to transform these backbone sectors, making them more efficient, cleaner, and adaptive. In this post, I’ll share what I’ve seen on AI’s growing role in enhancing resilience across heavy industry, energy, and infrastructure, drawing on examples from the United States, Europe, and beyond.
Resilience in Heavy Industry: Efficiency Meets Adaptability
Heavy industries like steelmaking, mining, and manufacturing are often labeled “hard-to-abate” sectors due to their massive energy use and emissions. The steel industry alone contributes roughly 8% of global CO2 emissions, and mining and materials firms face similar challenges. Building resilience here means finding ways to produce more efficiently, adapt to input variability, and cut emissions – all without compromising output. This is where AI has begun to make a tangible impact.
Smarter Steelmaking (Gerdau): Consider steel recycling. Using scrap metal in electric-arc furnaces (powered by clean electricity) can dramatically reduce emissions versus traditional coal-fired blast furnaces. The catch is that scrap metal typically comes in inconsistent quality – varying impurities can disrupt production and waste energy. At a Gerdau steel plant in North America, an AI platform by startup Fero Labs is tackling this challenge. By analyzing years of production data, the system learned how different “recipes” of scrap and additives affect steel quality (link). Now, before each batch, the AI measures the scrap’s composition and recommends the minimal mix of alloys needed to meet quality specs. The result? The mill saved time, avoided excess alloy use, and in 2024 cut the greenhouse emissions for a common steel grade by 3.3% – all with no new hardware (link). This might sound like a small percentage, but in an industry as emissions-intensive as steel, it’s significant progress from just software improvements. The International Energy Agency noted in its 2024 report that if such AI process optimisations were scaled up globally in industry, they could save on the order of 8 exajoules of energy by 2035 – about as much energy as Mexico consumes in a year (link). In other words, smarter manufacturing AI could yield energy savings at a national scale.
Mining and Power Optimization (Fortescue): Mining operations are another arena where AI boosts both efficiency and resilience. Fortescue Metals Group, an Australian mining giant, deployed AI to optimize how its automated mining equipment uses power. Large mines often have to build their own power infrastructure, and the peak power capacity needed can be costly. Fortescue’s AI system intelligently allocates and schedules power for drills, trucks, and processing plants. Impressively, this helped reduce the required power system capacity by 9%, meaning the mine can run on a smaller, leaner power setup. This translated into an estimated $500 million in capital savings for the company (link). Cutting 9% of peak power demand not only saves money, but it also makes the operation more resilient to energy supply issues (for instance, if part of the power grid or generators go down). Likewise, AI-driven predictive maintenance in mining equipment has reduced unplanned downtime and extended machinery life, keeping production steady. Fewer breakdowns and a more right-sized power supply mean a mining site is less likely to be disrupted – a clear resilience boost in a tough environment.
Adaptive Manufacturing and Process Industries: Beyond these examples, many heavy industries are seeing similar benefits. AI models can analyze sensor data from factory machines to predict failures with up to 90% accuracy, allowing maintenance to be scheduled proactively (link). In one example from the wind power sector, such predictive maintenance cut turbine downtime by 25%, and similar approaches are being rolled out in oil refineries and chemical plants. AI can also help optimize energy use in real time on factory floors – from adjusting furnace temperatures to tuning chemical process parameters – to shave off waste. European firms are actually at the forefront here, holding over half of the global market share in industrial automation solutions – a critical foundation for deploying AI in factories. This gives hope that heavy industry in Europe and elsewhere can rapidly integrate AI for efficiency. Ultimately, a more energy-efficient, predictive, and flexible industrial process is a more resilient one – it can withstand energy price spikes, adapt to variable raw materials, and maintain output with less downtime.
AI’s Impact on the Energy Sector: Smarter, Greener Operations
If heavy industry is the backbone, the energy sector is the lifeblood of our economy – and making energy systems resilient underpins all climate efforts. Here too, AI is proving to be a powerful tool, from oil fields to renewable grids. I’ve seen examples of AI helping energy companies optimize production, reduce emissions, and better balance supply and demand, all of which enhance the resilience of energy infrastructure.
Upstream Oil & Gas (Shell, ADNOC): Oil and gas may seem at odds with climate goals, but reducing the carbon footprint of necessary production is still vital in the transition period. The oil & gas industry has actually been an early adopter of AI for operational efficiency. Shell, for instance, uses AI-driven analytics to improve its drilling accuracy and optimize well placements. By crunching geological data, AI can make exploration and drilling more precise, cutting down on dry holes and unnecessary effort. Shell estimates that AI-enhanced drilling optimization could save billions of dollars annually in its operations. Another major producer, ADNOC (Abu Dhabi National Oil Co.), is leveraging AI to monitor and control its field operations in real-time. One ADNOC oil field reached an industry-leading carbon intensity of just 0.1 kg CO₂ per barrel of oil, partly thanks to AI systems that fine-tune equipment settings to minimise energy use and emissions, as reported in the Journal of Petroleum Technology (link). Additionally, AI-powered leak detection is helping oil and gas firms spot methane leaks (a potent climate pollutant) much faster. Using satellite and drone imagery with AI, operators can detect and fix leaks in days instead of months. The IEA found that if continuous AI monitoring replaced today’s periodic checks, it could prevent nearly 2 million tons of methane from escaping annually (equivalent to ~60 million tons of CO₂) – a major win for both climate and operational safety. In essence, AI is making fossil fuel operations more nimble and less wasteful: wells can be managed with surgical precision, maintenance can be anticipatory, and even carbon capture projects can be planned more efficiently with AI-aided simulations. This not only cuts emissions per unit of energy, but also means energy companies can remain profitable and reliable with tighter carbon constraints – a form of business resilience during the clean energy transition.
Electric Power Grids and Renewables: On the electric power side, AI is the secret sauce turning yesterday’s grid into a smarter, more adaptive network. As the grid incorporates more renewable sources like wind and solar (which are intermittent by nature), maintaining reliable power becomes more complex. AI is rising to this challenge by forecasting, optimizing, and automating grid management in ways that were not possible before. For example, in the UK, National Grid partnered with an AI startup to deploy sensors and algorithms on transmission lines, and discovered that many high-voltage lines can safely carry 20–30% more power than their old static ratings allowed, thanks to real-time monitoring of conditions (link). By dynamically increasing throughput when conditions permit, Britain “unlocked” an extra 600 MW of capacity to transmit offshore wind power – roughly the output of two mid-sized power plants – without building a single new line. The IEA concurs that remote sensors plus AI can boost transmission capacity significantly; globally up to 175 GW of extra capacity might be gained in this way. This is a clear resiliency boost: more power can be routed where needed, easing bottlenecks and preventing overloads during peaks.
AI is also speeding up fault detection and grid repair. Machine-learning models can pinpoint anomalies in voltage or frequency that human operators might miss until it’s too late. In one case study, an AI detected a brewing grid instability in a single day that would have taken analysts weeks to diagnose manually. By catching problems early (or even predicting them), AI helps avoid outages. The IEA reports that AI-based fault detection can cut outage durations by 30–50% on average – meaning faster recovery when storms or failures do occur. Shorter and fewer blackouts = a more resilient grid, obviously.
On the supply side, renewables themselves benefit from AI. Google’s DeepMind applied AI to its wind farms, using weather forecasts and turbine data to predict power output 36 hours ahead. This allowed scheduling of energy deliveries in a way that increased the value of the wind energy by 20% and made it easier for grid operators to use that power. In Chile, an AI tool called Tapestry (from Google’s X) is helping planners anticipate grid congestion and site new renewable projects more optimally, potentially even accelerating coal plant phase-outs by a decade in one scenario. All these examples underscore a theme: AI is enhancing the flexibility, efficiency, and foresight of power systems. A more flexible grid is inherently more resilient – it can handle swings in supply or demand, recover from shocks, and integrate new sources without breaking.
Utilities and Energy Consumers (Octopus Energy’s Kraken): Resilience isn’t just about the big generators and wires; it also extends to how energy is used on the demand side. One fascinating European example comes from Octopus Energy in the UK. Their software platform “Kraken” uses AI to manage a constellation of smart homes, electric vehicles, and batteries as a coordinated resource. The AI crunches 8 billion datapoints a day from nearly 500,000 devices, turning devices on or off at just the right times to soak up cheap renewable power or relieve grid stress. By shifting loads to off-peak times, Octopus’s distributed assets (including a big chunk of Britain’s grid-scale batteries) collectively avoided over 16 million tonnes of CO₂ emissions in 2024 (~10 tons of CO2 per kW) – an eye-opening number, equivalent to taking many millions of cars off the road. This kind of AI-orchestrated demand response not only cuts emissions and costs, but also keeps the grid from overloading during peaks, improving resilience. It’s a win-win: households get lower bills, and the grid gains flexibility to handle more renewable energy smoothly.
In sum, AI is injecting much-needed adaptability into the energy sector. From oil fields operating at ultra-low emissions to self-balancing renewable grids and interactive consumer programs, these innovations strengthen the energy system’s ability to cope with change. An AI-augmented energy system is leaner (fewer losses), cleaner (lower emissions), and quicker to react to problems – all pillars of resilience as we strive for net-zero emissions.
Smart Infrastructure: AI in Transportation and Buildings
Resilience isn’t just about heavy industry and energy production – it also extends to the infrastructure that moves people and goods and the buildings we live and work in. AI technologies are increasingly being applied in transportation logistics and the built environment to improve efficiency and adaptability. This both reduces emissions and makes these systems more robust against disruptions.
AI-Optimized Shipping (Maersk & Ports): Global shipping is a key element of trade, but it’s also a major emitter and vulnerable to delays (think of port bottlenecks or weather). AI is helping maritime firms sail more efficiently. Maersk, the world’s largest container shipping line, has been using AI route optimization tools for over a decade to chart optimal voyages. These systems factor in real-time weather, sea currents, and port traffic to minimize fuel burn and avoid congestion. The payoff has been substantial: Maersk’s AI-guided routing cut fuel consumption by roughly 10–15%, translating to about $250 million in fuel savings per year. That’s not only a big cost reduction but also a big emissions cut (less fuel burned means less CO₂). It also improves resilience – ships that adjust course intelligently can steer clear of storms and adapt if a destination port is unexpectedly crowded or closed. We saw a vivid example of AI in action during port congestion issues in recent years. In Rotterdam (Europe’s busiest port), a platform called PortXchange uses AI to coordinate the arrival and departure of vessels. One chronic inefficiency in shipping is “hurry up and wait” – ships racing to port only to idle offshore awaiting a berth. By analyzing dozens of data points (vessel locations, port schedules, tides, etc.), PortXchange can advise ships to slow down en route and arrive just-in-time. This cut idle waiting time by 20% for ships using the system. Shell, the oil major, reported that using this AI platform across its barge and tanker operations significantly reduced fuel waste from idling. In essence, the port became more fluid and less prone to pile-ups. During the COVID-era supply chain crisis, such AI-driven scheduling tools were invaluable in boosting the resilience of logistics networks, helping goods keep flowing while saving fuel.
Smarter Rail and Transport Networks: It’s not just ships – AI is also optimizing other transport infrastructure. In railways, operators like Deutsche Bahn and SNCF have deployed AI “eco-driving” systems for train drivers, yielding energy savings of 10–15% by automatically adjusting speed profiles. Predictive maintenance algorithms for rail tracks and equipment similarly reduce delays from service disruptions. And in aviation, airlines using AI for flight path optimisation have trimmed fuel burn (Lufthansa saved €150 million a year through AI route planning, cutting fuel use ~5%). These efficiency gains across transport modes mean not only lower emissions, but also extra slack in the system – a flight with optimal fuel and route can better handle unexpected headwinds, and a train with well-maintained tracks is less likely to derail or be delayed. The IEA projects that applying known AI applications in transportation broadly could cut energy demand equivalent to the consumption of 120 million cars by 2035. That implies hundreds of millions of tonnes of CO₂ avoided and more reliable transit networks as a side benefit.
Intelligent Buildings: Our built environment – commercial buildings, factories, homes – is another critical piece of infrastructure where AI can drive both efficiency and resilience. HVAC is often the single largest energy consumer in large buildings and is notoriously wasteful when run on fixed schedules or manual control. Cloud-based AI connects to a building’s existing HVAC controls and continuously adjusts settings based on a barrage of data: indoor temperatures, occupancy, weather forecasts, even electricity grid conditions. The results have been impressive in pilot projects. Buildings using AI optimisation have seen up to 25% reductions in HVAC energy use, and as much as 40% reductions in HVAC-related carbon emissions (link). For instance, one shopping mall saw a 21% drop in HVAC electricity consumption after a year on the AI system. These savings not only cut costs and emissions, but they also make the building more resilient to heat waves or cold snaps. The AI can pre-cool or pre-heat spaces in anticipation of a weather event, and adjust in real-time to keep conditions stable without breaking the energy budget. Moreover, by reducing strain on equipment, the tech lowers the risk of critical failures – e.g. a chiller going down on a sweltering day. Multiply this across millions of buildings, and the aggregate effect is a more reliable, peak-shaved urban power demand profile. In fact, the IEA estimates that scaling up existing AI building optimizations globally could save around 300 TWh of electricity annually – roughly the entire yearly generation of Australia and New Zealand combined. That’s a big chunk of demand that can be shed during peak times, helping prevent grid overloads. Smart buildings essentially become active participants in grid resilience, adjusting their consumption intelligently to balance the system.
From autonomous shipping routes to self-adjusting smart homes, AI is infusing infrastructure with a new level of responsiveness. These sectors traditionally operate with thin margins and tight schedules, so any efficiency gain directly improves their robustness. A ship or truck that burns less fuel has more margin to detour when needed; a building that self-regulates climate can ride through extreme weather without failing; a train system that predicts maintenance issues will have fewer surprise breakdowns. In all cases, emissions reduction and operational resilience go hand-in-hand – by optimising for efficiency, AI is inherently making these systems better able to handle shocks.
Benefits and Challenges of AI Adoption in Climate-Critical Sectors
The examples above make a compelling case that AI can be a force-multiplier for decarbonizing heavy industry and infrastructure while also improving their performance. The benefits can be summarised as follows:
Operational Efficiency and Cost Savings: AI algorithms optimise processes to eliminate waste – whether it’s fuel, energy, time or materials. This yields lower operating costs (e.g. hundreds of millions saved in fuel or power) and often higher throughput. Efficiency gains also mean less resource consumption and emissions for the same output, contributing to climate goals. Many of these efficiency improvements (like fewer idle ships or lower peak power needs) directly enhance resilience by providing buffer capacity and reducing strain on systems.
Emissions Reduction: Nearly every case of AI deployment cited has a clear emissions benefit – 3.3% less CO₂ in steel, 9% less power capacity (hence fuel) in mining, 20% less fuel burn in shipping idle time, 40% less HVAC emissions in buildings, 16 million tons CO₂ avoided by smart grid controls, etc. While AI alone is not a silver bullet for climate change, the IEA found that broad adoption of existing AI tech in energy and industry could cumulatively cut global energy-sector emissions by 1.5 gigatons by 2030 (a material chunk, though still only a slice of what’s needed). Every bit helps, and importantly these reductions often come at low cost or even with profit – making green transitions more economically palatable.
System Adaptability and Reliability: AI brings predictive and adaptive capabilities that keep systems running optimally even as conditions change. This includes predictive maintenance (preventing breakdowns and downtime), real-time reconfiguration of networks (power rerouting, traffic management), and rapid data-driven decision making in emergencies. For example, AI that detects a pending transformer failure can reroute power and dispatch a crew before a blackout occurs. A human team might react only after the outage. That speed and adaptability is crucial for resilience in critical infrastructure. In one utility example, what used to take weeks of analysis (identifying a grid instability source) took one day with AI. Faster response = more resilient operations.
Accelerated Innovation: (A forward-looking benefit) AI is also being used to accelerate R&D and innovation in climate tech – something the energy industry sorely needs. An AI model can sift through thousands of material combinations to find better battery chemistries or more efficient solar cell designs far faster than traditional lab experiments. The IEA report noted that only 0.01% of possible novel solar materials have been tested experimentally, but AI can drastically narrow that search space. By shortening the “lab-to-market” cycle for new clean technologies (which traditionally takes decades in energy), AI could indirectly boost resilience by bringing advanced solutions online sooner. However, this is just beginning – only 2% of energy sector startup funding to date has flowed to AI-focused firms, suggesting a lot of untapped potential if investors (like us in climate tech) funnel more capital into this crossover of AI and energy innovation.
That said, adopting AI at scale in heavy industries and infrastructure is not without challenges. Through conversations with industry participants and reading reports, I’ve identified a few key hurdles:
Data Access and Quality: AI runs on data, but many industrial companies have historically poor data infrastructure. Sensors may be sparse, data may be siloed in proprietary systems, or simply not collected at all. Access to relevant, high-quality data is a significant barrier to unlocking AI’s potential in the energy/industrial sector. For example, a steel mill might not have digitized certain furnace parameters, or an older power grid might lack real-time sensors on many assets. Companies need to invest in IoT sensors, data platforms, and cloud connectivity to feed the AI beast. Moreover, data sharing between entities (e.g. between a shipping line and a port authority) can be sensitive due to competitive and privacy concerns. Without pooling data, some AI solutions (like sector-wide optimisation) may be hard to achieve. Addressing this will require both technology upgrades and new partnerships or data-governance frameworks.
Trust and Transparency: Even when the data is there and the AI model works, you need people to trust its recommendations. In safety-critical or mission-critical operations, there is understandable reluctance to hand decisions to a “black box” algorithm. Operators ask: How do we know the AI won’t screw up? Building trust in AI systems is therefore essential. This can involve using explainable AI (so engineers can understand the why behind suggestions) and a gradual approach to autonomy (keeping a human in the loop until confidence is earned). In heavy industry, a plant manager might ignore an AI optimization if it conflicts with their intuition, especially without clarity on how it works. Social trust can also be tested if AI becomes very pervasive – e.g. workers might worry about being displaced, or the public might balk at fully autonomous ships or trains due to safety fears. Ensuring AI systems are reliable, transparent, and augment (rather than replace) human expertise is key to adoption. Initiatives around ethical AI and third-party validation are interesting developments to address this challenge.
Workforce and Skills Transformation: A recurring theme I’ve heard is that implementing AI is as much a people challenge as a tech challenge. Traditional industries often lack in-house digital skills; the prevalence of AI-related skills in energy and industry is much lower than in the tech sector. This skills gap can slow adoption – companies may not even know where to start with AI, or how to integrate it into workflows. National Grid for example places a big focus on workforce transformation to increase the integration of AI in its operations (link: https://www.riiot3.nationalgrid.com/sites/g/files/atxybb411/files/documents/2024-12/A03%20NGET%20Workforce%20and%20Supply%20Chain%20Resilience%20Strategy_.pdf). Moreover, there may be resistance from workers who fear automation will make their jobs obsolete or upend established practices. Tackling this means upskilling the workforce (training engineers in data science, hiring AI talent into industry) and emphasizing that AI can empower workers, not just replace them. Managing that transition requires careful change management. Success stories often involve a champion who bridges domain expertise and AI know-how, translating between data scientists and plant operators.
Infrastructure and Investment: Some AI solutions need upfront investment in digital infrastructure (sensors, connectivity, cloud computing) or adjustments to business models. Companies must be willing to allocate capital to these long-term efficiency gains, which can be tough if they’re laser-focused on next quarter’s results. There’s also the challenge of integrating AI with legacy systems – sometimes upgrades or retrofits are needed, which can be costly or cause downtime during installation. Policy incentives might help here: governments can encourage digital retrofitting of factories or grid modernization as part of climate policy, for example. On the flip side, capital is increasingly available for climate-AI crossover projects – as evidenced by large companies like Microsoft investing in grid AI startups and funds targeting “Industry 4.0” solutions. The momentum is building, but ensuring smaller firms and developing economies are not left behind will be an important challenge to solve.
Regulatory and Security Concerns: Finally, regulation can both help and hinder. In some sectors, outdated regulations might prevent AI adoption – for instance, rules mandating human inspection of equipment that could be done by AI vision systems, or legal barriers to autonomous vehicles on public roads. Policymakers will need to update standards to safely accommodate AI (e.g. defining liability for AI decisions, setting cybersecurity requirements, etc.). Speaking of security, increased digitalization does raise cybersecurity risks. More connected devices and AI control means a larger attack surface for hackers. The energy sector has already seen cyberattacks triple in the past four years, some enabled by AI tools. At the same time, AI can also strengthen cybersecurity by detecting anomalies. This is an ongoing tug-of-war: as we digitize for efficiency, we must also invest in securing these systems, or we risk new kinds of failures. Resilience here means cyber-resilience too.
Despite these challenges, the overall trajectory feels optimistic. The fact that legacy industries are engaging with AI at all represents a sea change. The conversation has shifted from “why use AI?” to “how do we implement it responsibly and effectively?” – which is great to see. Trust-building, workforce training, and smart policy will be pivotal in the next few years to ensure AI’s promise translates into real-world resilience gains in a broad-based way.
Conclusion: A New Frontier for Climate Tech Startups and Investors
Stepping back from all these examples, it’s clear to me that we are at an inflection point. AI is moving out of the tech realm and embedding itself in the physical industries that undergird our civilization. For someone like me who invests in climate tech, this convergence of AI with heavy industry, energy, and infrastructure is one of the most intriguing (and promising) frontiers out there. It signals a shift in how we tackle climate solutions: in addition to inventing new clean energy sources, we are also making our existing systems dramatically smarter and more efficient. This two-pronged approach – build the new, optimize the old – is vital for meeting climate goals in time.
From an investor perspective, the implications are exciting. Traditionally “old school” sectors are becoming fertile ground for tech innovation and startup disruption. I interpret that as a huge opportunity for climate tech entrepreneurs: there are so many niches where AI can optimise processes and cut emissions, and many problems (from steel mill recipe optimisation to grid congestion forecasting) are just beginning to be solved. Startups that understand both the domain and the AI technology – the “hard hat and hoodie” combination – will be in high demand. And they might find eager corporate partners in incumbents looking to accelerate their digital transition.
For investors, backing these kinds of solutions can deliver both climate impact and solid returns. The ROI for industrial AI can be very attractive because it often unlocks cost savings (energy, downtime, etc.) and emissions reductions as a co-benefit. That means there’s a natural business case driving adoption, not just a moral case. We could soon reach a point where not using AI is the competitive disadvantage. Companies that ignore these tools risk being left behind – less efficient, more polluting, and more prone to disruption.
Of course, we must remain thoughtful and guard against overhyping. Not every factory needs a fancy AI, and not every AI will succeed. There will be pilot projects that flop, algorithms that don’t generalize well from one plant to another, and instances where the energy to run an AI might outweigh its savings (the energy footprint of AI itself isn’t trivial, especially for large models). But the direction is clear. The key for climate tech investors is to identify real problems that AI is uniquely suited to solve (like complex optimisation or prediction tasks) and to support teams that have a plan to overcome the adoption barriers we discussed (data, trust, etc.). Those will be the game-changers.
In closing, AI is not a magic wand – we still need policy, human ingenuity, and many other innovations. Yet, it’s a powerful new tool in our toolkit to strengthen the very foundations of our clean economy. The heavy industries of the 20th century are getting a digital facelift, and the early results show not just marginal improvements, but step-changes in efficiency and adaptability. It’s a reminder that sometimes the greenest energy is the energy we don’t waste. By using AI to squeeze out waste and anticipate issues, we are effectively creating virtual capacity and reducing emissions that would otherwise require building new assets or burning more fuel. That makes our whole society more resilient to the challenges ahead.
As a curious learner and investor, I’m excited to watch this space evolve. I believe the coming years will see AI move from pilot projects to a ubiquitous part of industrial operations, much like automation did in decades past. And as that happens, resilience – in the form of robust, low-carbon, efficient systems – will no longer be an afterthought; it will be built into the very algorithms that run our world. For those of us supporting climate tech, that means now is the time to lean into AI’s potential, fund the hard work of deployment, and ensure these innovations scale in a responsible way. The payoff could be enormous: a cleaner planet and industries that can weather the storms (literal and figurative) of the 21st century. That’s a future worth investing in.
If you're building or investing in this space, I’d love to connect.