AI changes organisations’ understanding of risk, making decisions, and acting on environmental threats in real-time. From predictive modelling to emissions transparency, the tools for climate resilience are here, and the cost of inaction is rising fast.
There was a time when weather forecasts were little more than hopeful guesswork, and emissions data arrived in government spreadsheets years after. That world no longer exists. AI is helping to redraw the map of climate resilience, not through hyperbole or moonshots, but by solving real problems at speed and scale. From predictive fire monitoring and city-level flood modelling to emissions tracking and operational carbon reduction, climate AI is moving rapidly from possibility to necessity.
The technologies involved, including generative AI, satellite-based sensing, and agentic systems, are powerful, but the shift is not just technical. It is organisational. Whether helping governments mitigate disaster risk or enabling businesses to track their emissions with surgical precision, AI is changing how decisions are made and who can make them. The emerging picture is one of agency: tools that put meaningful action in more people’s hands more quickly and affordably.
“We are combining proprietary datasets with AI models to create new ways of predicting wildfires and planning for them,” Karthik Kashinath, Principal Engineer and Scientist, AI-HPC at NVIDIA. says. “We are operationalising this blueprint in real-world scenarios, like extreme winter storms or urban flooding. The entire prediction pipeline is being transformed from data to actionable insight. This is no longer about the weather. It is about what you do with it.”
From observation to action
Earth-2, NVIDIA’s climate and weather modelling platform, is at the heart of this transformation. It merges high-performance computing, simulation and generative AI to produce accurate, localised and probabilistic forecasts. Partnering with organisations like Tomorrow.io is becoming the connective tissue for operational weather intelligence.
“Accuracy depends on data,” Shimon Elkabetz, CEO and Co-Founder of Tomorrow.io, says. “But 90 per cent of the world has no radar coverage. That pushed us to build and launch our own satellite payloads; we now have eight proprietary satellites in orbit, delivering real-time, high-resolution data into our models. That capability allows us to serve global customers with the same quality we provide to US major airlines or sports leagues.”
The company provides automated, AI-powered decision support to organisations ranging from NASA and the US Air Force to Uber and Delta. Their use of Earth-2’s generative capabilities means models are faster, more accurate and easier to integrate with real-world sensor data. Elkabetz is clear on the business case. “We helped the largest US railway reduce weather-related accidents from eight per year to zero,” he adds. “Each incident cost around $35 million. In mining, we help prevent environmental fines by optimising blast schedules. The savings are both environmental and financial.”
These systems do not replace human judgement; they sharpen it. When a city needs to decide whether to shut a metro line or a mining operation needs to reschedule an activity to avoid rainfall-triggered landslides, real-time weather AI becomes a critical operational tool.
From platform to ecosystem
But platforms alone do not make decisions. Data must be made usable, contextual, actionable, and available when it matters. That is where the agentic layer comes in. At Salesforce, AI systems are trained as assistants and collaborators, not black boxes.
“We are helping both for-profits and non-profits build conversational interfaces that make climate data actionable,” Suzanne DiBianca, EVP and Chief Impact Officer at Salesforce, says. “For example, in the LA fires, we supported Good360 in using WhatsApp to let people ask what items were needed for disaster relief. They received precise guidance, which reduced waste, optimised logistics, cut emissions, and improved the speed and impact of relief.”
Salesforce also practices what it preaches. It built Net Zero Cloud to manage emissions across its operations, introduced a ‘carbon per serve’ metric to evaluate infrastructure efficiency, and locates data centres based on grid cleanliness and hardware optimisation. AI is being embedded not just in customer-facing systems but also in internal operations, engineering metrics, and procurement decisions.
These are not theoretical tools. “I can ask Net Zero Cloud for my worst-performing building, get why, identify other tenants, and receive suggestions for actions,” DiBianca explains. “It even provides ROI analysis so I can make a case to the landlord.”
Agent-based systems are now being layered onto emissions data from sources such as Climate TRACE, a global open-source emissions tracking platform aggregating satellite, lidar and ground-based data to monitor over 600 million emission points in near real-time. “We’re working with Climate TRACE and Project Drawdown to build an interface where a new county supervisor in Denver can identify their jurisdiction’s largest sources of emissions and receive a menu of evidence-based actions to reduce them,” DiBianca continues. “That combination of TRACE’s data, Drawdown’s solutions and a usable interface makes it actionable for anyone, anywhere.”
From commitment to accountability
Unfulfilled pledges and shallow commitments have often derailed the corporate climate conversation. But as weather events intensify, that posture is becoming untenable. Colin le Duc, Founding Partner at Generation Investment Management, framed the problem starkly. “We are in an energy addition phase, not an energy transition.” He says. “AI is accelerating energy demand, and nobody is willing to compromise on performance or latency. So, AI systems are powered by any available source, whether clean or not. Emissions are still rising, and we are now at record highs for coal, oil, gas, and deforestation.”
Investors and regulators alike must stop treating emissions risk as an abstraction. “Many are backtracking on climate commitments, but ignoring risk does not eliminate it,” le Duc adds. “At some point, the weather or regulators will intervene forcefully.”
Generation only invests in companies with science-based decarbonisation targets and measurable outcomes to counter this. More radically, it has linked its own performance fees to impact. “We only get paid if we deliver both financial returns and a megaton-scale reduction in emissions per year,” le Duc says. “That was impossible three years ago, but new tools now allow us to track decarbonisation precisely.”
Climate TRACE, which Generation helped to initiate alongside Al Gore, is critical in enabling this precision. “Until now, greenhouse gas inventories were based on self-reported spreadsheets, which were often wildly inaccurate or incomplete,” he continues. “TRACE reveals sources previously overlooked, like piles of coal that emit more than the airline industry or ghost ships emitting beyond jurisdictional reach.”
The platform is publicly available and intentionally not monetised. “We refused to commercialise it because this data must remain universally accessible,” le Duc explains. “In the hands of policymakers, citizens, and investors, that transparency becomes a mechanism for accountability.”
From latency to leadership
It is tempting to treat AI as a future technology, something to be piloted cautiously and deployed later. However, the story unfolding across weather systems, infrastructure planning, emissions tracking, and disaster response suggests otherwise. The technology exists, the use cases are compelling, and the barriers are no longer technical.
What matters now is latency: how long organisations take to recognise climate AI not as an option but as a critical part of resilience and performance. “Every company has systems for cyber risk, but almost none have IT solutions for weather risk,” Elkabetz says. “That is no longer acceptable.”
There is also a shift in how expertise is distributed. The most potent AI models are not centralised but adapted and embedded across systems, from satellites to supply chains. “Generative AI enables us to generate high-resolution predictions at finer scales, potentially down to street level,” Kashinath says. “It supports probabilistic forecasting and excels at fusing diverse data types. This opens the door for entirely new applications, from city-scale early warnings to ultra-local impact analysis.”
Executives responsible for digital transformation and sustainability must now consider climate AI a core infrastructure layer. It spans everything from training and deploying models responsibly to selecting energy sources for computing, measuring emissions in real-time, and putting insights in the hands of operational decision-makers.
Tenika Versey Walker, Global Head of Inception Sustainable Futures at NVIDIA, sees this convergence as the defining task of the decade. “As we explore how climate science and AI are working together to build a more sustainable future, I am excited to see the advances being made,” she says. “We are building initiatives to help solve some of the world’s greatest challenges.”
Model performance alone will not decide whether those challenges are met. It will depend on who acts, how soon, and how seriously. Climate AI is not a single technology or a standalone solution. It is a lens through which the future can be made more knowable and more manageable, one decision at a time.
