The tech world has been abuzz this week with news about DeepSeek, a Chinese AI startup that has developed a chatbot at a fraction of the cost and energy consumption compared to its American counterparts. While the it’s capabilities are impressive, what truly captures my attention is DeepSeek’s significant reduction in energy usage. Early estimates suggest up to 90% lower energy consumption and around 92% lower carbon footprint compared to equivalent infrastructures, alongside greatly reduced cooling needs.
DeepSeek’s AI model required approximately 836,400 kWh of energy for training, utilising 2,788,000 GPU hours. In contrast, leading AI models from U.S. tech giants consume substantially more energy, often necessitating billions of dollars in infrastructure and operational costs. This disparity not only highlights DeepSeek’s efficiency but also prompts a reevaluation of the anticipated energy demands associated with AI technologies.
This conversation around energy efficiency isn’t new. Back in 2017, I was heavily involved in a blockchain project. At the time, blockchain was an emerging concept, and one of the biggest concerns was its energy consumption. For example, Bitcoin’s network was estimated to consume around 6.6 terawatt-hours (TWh) annually, comparable to the power usage of entire countries. Despite its promise, blockchain faced significant scrutiny for its environmental impact.
Since then, there has been a noticeable shift in how emerging technologies address energy consumption. AI and blockchain are no longer solely focused on raw computational power—there’s now a growing emphasis on optimising efficiency. Innovations like DeepSeek’s AI model exemplify this shift, proving that high-performance computing doesn’t have to come at the cost of unsustainable energy usage. This not only reduces operational costs for companies but also aligns with global sustainability goals.
However, recent political developments could shape the direction of these efforts. With Donald Trump now back in office, energy policy in the U.S. is set to take a different course. On his first day back, he declared a national energy emergency, prioritising the expansion of fossil fuel production and rolling back regulations designed to push renewable energy. This marks a shift away from the previous administration’s push toward net-zero carbon targets.
For the tech industry, which is increasingly conscious of its environmental footprint, this shift presents both challenges and opportunities. While a stronger emphasis on fossil fuel energy could provide cheaper and more available power, it also raises concerns about rising carbon emissions, potentially making it harder for companies to meet their sustainability targets.
The AI boom has already sparked concerns about electricity demand, with some predicting that future AI models could consume as much energy as entire cities. This makes it more critical than ever for companies to invest in energy-efficient infrastructure and, where possible, transition towards renewable sources.
Looking ahead, the tech industry must continue prioritising energy efficiency—not just for economic reasons, but for long-term sustainability. As AI and other emerging technologies become more integrated into everyday life and business, their energy consumption must be managed responsibly. Sourcing power from renewables, designing more efficient hardware, and optimising AI models will be key to balancing innovation with environmental responsibility.
The shift we’ve seen in recent years—from blockchain debates in 2017 to AI efficiency breakthroughs today, shows that the industry is moving in the right direction. The next step? Ensuring that these advancements don’t just reduce energy consumption but actively support a clean energy future.
By making sustainability a non-negotiable part of AI and tech development, we can harness the benefits of cutting-edge technology while protecting the planet for future generations.
