Sustainability was once a highly touted advantage in public cloud computing. Enterprises and cloud providers flaunted their green initiatives, promoting data centers powered by renewable energy that would reduce carbon footprints. This topic has quietly slipped off the radar in recent months. The culprit? The insatiable demand for artificial intelligence capabilities that is pushing the significant expansion of cloud data center footprints. It’s a reality many would prefer not to highlight.
Does a desire for AI technology trump the health of the planet? Right now, you can’t have both. A stellar sustainability record will not offset massive data center construction no matter how many solar panels and electric car chargers you have at HQ.
The AI revolution and its demands
The rise of AI has been nothing short of meteoric. Enterprises are vigorously adopting generative AI to automate processes, glean insights from big data, and offer personalized customer experiences. Such advancements come at a cost. AI applications require immense processing power, increasing demand for data center capacity. Complex AI models are voracious energy consumers. This type of consumption is at odds with the sustainability narrative that the cloud industry once touted. Today, enterprises and cloud providers actively avoid conversations about sustainability.
Don’t get me wrong. I’m sure some companies still do a good job curbing their carbon footprint and should be congratulated. I’m talking about industrywide generalities, and the larger discussions about sustainability have suddenly faded into the background. The press rarely asks about it and most enterprises have dropped sustainability from their outreach programs.
A rush to build data centers
The need for rapid scaling has created a paradox. The larger these cloud data centers’ footprint and processing power, the further the providers drift from the ideals of sustainability they once espoused. Many enterprises find themselves in a difficult position: They need to harness AI capabilities to remain competitive, yet they are unable to reconcile this reality with their sustainability commitments.
On the one hand, enterprises want to show sustainability progress, perhaps even a solid environmental, social, and governance (ESG) score. Many enterprises still view AI as optional, but AI will come to every enterprise in some form, and there is no way around the amount of power AI will consume.
You only need to look at the differences in power consumption between GPUs and CPUs. CPUs were fine for most traditional processing, on the cloud and on-prem. However, the shift to AI has made GPUs the preferred processors. GPUs draw more power due to their design for parallel processing, which is necessary for graphics rendering and intensive computations in AI and gaming.
Top GPUs use 200 to more than 450 watts, while high-end desktop CPUs range from 65 to 150 watts. Midrange GPUs use about 100 to 250 watts, whereas mid-range CPUs usually consume around 65 watts. Replacing CPUs with GPUs will require much more power, cooling, and data centers. Enterprises are telegraphing a high demand for AI. If you do the math, you’ll quickly figure out that there is not enough power being generated on the planet to meet those future demands.
The sound of silence
Discussing the environmental impact of massive data center expansion could tarnish the allure of AI and its seemingly endless potential. Thus, many companies focus public discussions on AI’s innovative and transformative potential rather than its environmental challenges.
The rapid pace of technological advancement will outstrip the development of sustainable practices and technologies. Current renewable energy sources and improved energy efficiency measures need help to keep pace with the exponential growth in data processing needs. Given the speed and investment in AI, its growth will soon become unsustainable.
What troubles me is how fast this shift away from sustainability occurred. It seems like just yesterday that I presented a sustainable cloud computing strategy at a conference, and there was even a conference that focused exclusively on sustainability. The industry must revisit the sustainability of cloud computing and more traditional enterprise data centers. We need innovative solutions that balance technical capabilities with ecological responsibility to successfully address AI’s environmental impact. These include more efficient chip designs, advancements in cooling technologies, and increased investment in sustainable energy sources to power data centers effectively.
Just because we’ve stopped discussing sustainability doesn’t mean our sustainability problems have disappeared. Put aside your worries about AI-driven cyborgs taking over the world. The immediately frightening news about AI is its very real threat to our power grids. It’s time to reignite the sustainability conversations.