How do we scale and sustain AI adoption without burning the planet?

Every AI inference call comes with an energy cost. However, these costs remain invisible and neglected. As AI access and adoption accelerates, the current trajectory is likely to accelerate climate change. Additionally, the compute burdens of AI falls disproportionately on regions least equipped to bear it. However, without any labelling (and measurement) mechanisms, development trajectories are unlikely to change: what gets measured, gets built.  As the SDG and Paris Agreement deadline looms closer, this BoF is structured around a simple, uncomfortable question: Can we build an eco-cost standard for AI inference APIs and who would actually enforce it?

We'll unpack few forces in tension:

• Eco-labelling AI APIs & Standards: What would a meaningful standard look like? How can we coordinate between developers, hyperscalers, governments, and civil society? What is technically feasible today vs. what requires new infrastructure.

• Stakeholder Landscape of Who Cares: Who are the prominent players, bodies and funders interested in enabling climate-conscious AI development and reducing carbon footprint?

• Other sub-topics brought by the room:

  • Eg. Geopolitics of Compute & Jevon’s Paradox: US export controls on advanced chips are reshaping who can build AI and where. China's domestic ecosystem workarounds (Huawei Ascend, domestic fabs) are accelerating for more energy-efficient compute. Yet, as models get cheaper and more efficient, usage explodes and net energy consumption rises, not falls. Efficiency is not always a solution. How to work around this?

• The distinction between AI-driven tools for climate change vs reducing AI’s own carbon footprint. Our focus will be on the latter

• Feasibility of  eco-cost standard for interfaces  

• Importance of emission disclosure and energy labelling on AI inference APIs for consumers and businesses to make climate-conscious decisions (like Fairtrade standards for food industries)

• Enabling architecture for consumers to judge tradeoffs between cost, latency, accuracy and environment impact of their AI purchases

• Geopolitical implications of AI and energy-use amidst middle east crisis and Jevon’s Paradox