The use of AI across every sector, public and private, is now undeniable – and the CleanTech sector is no exception. However, AI’s growth comes with a big issue. As the International Energy Agency (IEA) put it in its 2025 Energy and AI report, “there is no AI without energy.” Data centre electricity demand rose 17% in 2025 alone, far outpacing the 3% growth in global electricity demand overall, and AI-focused demand specifically is projected to triple by 2030 (IEA, 2026).
Whilst this is a real cost, it is not the whole story. The IEA’s 2025 Energy and AI report is equally clear that, if used well, AI can meaningfully accelerate the search for climate solutions through faster R&D, cheaper experimentation, and more efficient data analysis across energy, industry, and the built environment. The question for the CleanTech sector is not whether to use AI, but how to use it for applications with genuine, measurable climate impact.
This Climate Action Week, we take a look at where AI is already being used for good by spotlighting some of the startups doing exactly that with the support of The Greenhouse, Undaunted’s 12-month CleanTech accelerator, which has supported 180 startups since 2012 and helped its alumni raise over $1.33bn in investment. We’ll also look at why, as AI-driven CleanTech innovation accelerates, a clear IP strategy is becoming essential to turning a good idea into a defensible, fundable business.
AI Innovation in Practice
Utilities and the built environment
In the UK, the energy required to maintain buildings accounts for almost a quarter of the country’s carbon footprint. At the same time, demand for sustainable office space is rising as ESG credentials become a bigger driver of commercial property value. This leaves building owners and facilities teams with a difficult issue: how to retrofit and manage existing stock sustainably, cost-effectively, and at scale.
Carbon Shift, a Greenhouse graduate, is tackling this with AI software that improves decision-making around sustainable retrofits to help building owners identify the most cost-effective and environmentally impactful interventions before committing capital. Cosy Sense, another graduate, has developed a management system (GB2701612A) that gives facilities teams in retail, office, and hospitality settings a single platform for monitoring energy use, with automated controls that let managers act on that data directly to reduce emissions.
Both are examples of AI applied to a hard, high-impact problem. AI isn’t being used as an add on feature but as the mechanism that makes sustainability decisions faster, cheaper, and more confidently taken.
Manufacturing, engineering, and maritime
The maritime sector faces a similar challenge at a larger scale. Shipping is responsible for 3-4% of the EU’s overall carbon dioxide emissions, and although the International Maritime Organisation considers it the least environmentally damaging mode of transport, its sulphur, nitrogen oxide, and carbon dioxide emissions remain firmly in regulators’ sights. The UK’s own Maritime Decarbonisation Strategy targets net zero for the domestic maritime sector by 2050, adding commercial pressure to an already complex engineering problem.
BlueNose, another Greenhouse alumnus, addresses this with AI-driven software that models the cost and emissions impact of retrofitting existing cargo ships, then designs aerodynamic retrofit structures (US2025382031A1) that can be fitted to vessels already in service. BlueNose estimates that, if rolled out fleet-wide across active container ships, its retrofits could cut emissions by 11 million tonnes of CO₂ a year.
What connects Carbon Shift, Cosysense, and BlueNose is that each uses AI as the engine behind a specific, well-defined climate outcome – lower retrofit costs, lower energy waste, lower fuel burn. That specificity matters, both for genuine climate impact and, as we explore below, for what can actually be protected as IP.
Why This Matters for IP Strategy
Despite the potential, AI adoption in the energy sector remains surprisingly low. The IEA’s 2025 Energy and AI report found that only 2.3% of energy start-ups have an AI-related value proposition, compared with 7% in life sciences and 4.3% in agriculture, and roughly only 1% of energy-related patents reference AI as part of the claimed innovation. The Greenhouse alumni therefore seem to be the exception, not the rule.
That gap is an opportunity but it is also exactly the situation in which IP strategy matters most. When a sector is under-exploited, the startups that move first have the clearest run at building a defensible position. As more capital and attention flow into AI-driven CleanTech, that window narrows, and clear, well-drafted protection becomes the difference between a startup that can defend its market position and one that cannot.
AI-driven inventions also raise distinct patentability questions that founders should consider early, including:
- What, exactly, is being protected? A patent claim built around “using AI” in the abstract is unlikely to be allowed. Instead, applications should focus on claiming the specific technical system, method, or structure that the AI enables rather than the algorithm in isolation.
- Timing of disclosure. Early conversations with investors, pilot customers, or conference audiences can constitute a public disclosure that destroys novelty if a patent application filing hasn’t happened first. This is a particularly easy trap for AI-driven startups, who often want to demonstrate a working model before they’ve locked down what, precisely, they intend to claim.
- Patents vs. trade secrets. Not all aspects of an AI system should be protected using patents. Training data, model weights, and certain algorithmic refinements may be better protected as trade secrets, while the system or output they produce may be the more suitable subject of a patent application. Getting this split right from the outset avoids having to unpick it at a later, less convenient moment.
None of the above considerations need to slow a startup down. If done early, a review of your IP considerations can be a relatively light-touch process that runs alongside fundraising and product development rather than competing with it. It is also the kind of groundwork that investors expect to see in place before they commit capital.
Climate Action Week is a good moment to look at how far AI-driven CleanTech has come – and Carbon Shift, Cosysense, and BlueNose are a small sample of what’s possible when AI is pointed at a specific, well-defined climate problem. As more startups follow their lead, the firms that protect their innovation early will be best placed to turn that progress into a lasting commercial advantage.
At Mathys & Squire, our team has deep expertise in AI, machine learning, and CleanTech and can assist you with your IP Portfolio. For advice or any questions related to your UK and European patent or design rights, please contact Charlotte Penney, Andrew White, or your usual Mathys & Squire patent advisor.
