Warren Buffett defined a moat as a business’s ability to maintain competitive advantages over its rivals to protect long-term profits and market share. “A truly great business must have an enduring ‘moat’ that protects excellent returns on invested capital,” he said. But what is a moat in the AI era? Does it even exist?
It’s tempting to believe that building with AI guarantees a defensible business. Powerful models, automation, and the ability to learn and improve feel like the ultimate advantage. But for startups trying to build durable moats around AI — and especially energy and climate tech startups — the reality is far more complex.
The democratization of AI tools has lowered the barrier to entry. Many of the traditional startup moats are being redefined or eliminated altogether. The challenge isn’t building AI software anymore; it’s leveraging AI strategically to create something tangible and enduring.
When OpenAI, Google, Meta, and others made large language models widely available, they flattened the playing field. Virtually anyone can now use APIs like GPT4 to create products, so the model is no longer the differentiator; instead, it’s the infrastructure wrapped around it that defines a startup’s success.
One of the most durable ways to create a moat is by incorporating a hardware component. Not long ago, cleantech investors were fixated on software-only companies that could offer promises of recurring SaaS revenue. Hardware such as batteries and solar had been commoditized, yielding slim margins. But the rapid rise of LLMs has shifted that thinking. Today more investors are acknowledging that hardware, as a vehicle for software deployment, brings both recurring revenue and customer stickiness.
A good example of this is in the home energy automation space. Companies that deploy a hardware hub to monitor and manage devices like thermostats, smart appliances, batteries, and solar, gain a distinct advantage. These local controllers learn household patterns over time, enabling smarter optimization for cost savings and resilience. Because they operate within the home ecosystem, they create a defensible position that’s hard for competitors to disrupt. Just as importantly, they offer a platform for utilities to engage directly with homes. This provides insights that support optimization for grid operators and added revenue streams for the company.
It’s a challenging approach given that hardware introduces higher costs and more operational complexity but in today’s rapidly moving software landscape, it can serve as a stake in the ground.
Another powerful moat is access to proprietary, high-quality data, especially about user engagement. Sensors that collect site specific data or systems built around user preferences, can be trained and refined in ways others can’t duplicate. The more users interact, the smarter and more tailored the product becomes. That feedback loop can produce better outcomes, attract more users, and lead to a defensible advantage.
But just having the data isn’t enough. Startups must secure ownership or at least access rights, and invest in making the data clean, structured, and useful for training and fine-tuning models in ways that materially improve performance.
This is particularly important in the utility sector, where vast datasets already exist. Gaining access is difficult and making sense of them is even harder. But the payoff can be transformative.
Another path to defensibility is specificity. AI built for a defined use case often outperforms general purpose tools. My advice to startups is the same I once gave students: learn something deeply, then apply business principles to it. A startup that understands the workflows of grid operators, architects, lawyers, or zookeepers can embed AI in ways that feel natural and necessary. That kind of domain depth and user empathy is hard to copy.
A counterintuitive moat is the human layer. In many domains such as legal advice, infrastructure design, and finance, full autonomy for AI still feels risky. Having a human in the loop to review, edit, or assure quality adds confidence and builds trust. Yes, it’s harder to scale and margins may be lower but it’s also harder to copy. A curated team of experts becomes part of the product DNA. Over time, that can translate to stronger retention and a better user experience.
And at the end of the day, customer acquisition and product distribution still reigns. History is full of cases where the best product didn’t win. AI will be no different. Startups that secure early distribution advantages through partnerships with utilities, legacy energy providers, or enterprise platforms can lock in access to users, which can make it harder for competitors to incentivize a switch.
The reality is that most startups won’t build a true moat around AI. They’re moving fast and lean, often without a unique wedge. Many will be acquired, copied, or outpaced by incumbents with broader reach and faster scale. The path to durability starts with strategic clarity. Who is your customer? Why does AI make your solution uniquely valuable? What problem are you solving in a way no one else can?
It’s worth remembering: It is still early days, and the pace of change is relentless. The rules for defensibility are being rewritten in real time. The challenge is not building with AI; it’s building moats that protect against its commoditization. The startups that succeed won’t just be the most technical. They’ll be the most thoughtful, the most strategic, the ones who recognize that in a world where AI is everywhere, context matters most.
Anna Demeo is the managing partner at Clean Tech Strategy Advisors. The opinions represented in this contributed article are solely those of the author, and do not reflect the views of Latitude Media or any of its staff.
