AI's Dirty Little Secret

Every time we go through a crypto boom (or should we say renaissance?), we are quickly reminded of the energy it consumes. At last count, Bitcoin consumes an estimated 150 terawatt-hours of electricity annually — more than the entire country of Argentina, with a population of 45 million. Producing that energy emits approximately 65 megatons of carbon dioxide into the atmosphere annually — comparable to the emissions of Greece — making crypto a significant contributor to global air pollution and climate change.

Crypto has a new environmental counterpart – AI. The AI industry is poised to rival an unexpected contender in energy consumption by 2027: the Netherlands. The disruptive debut of ChatGPT late last year marked a turning point, driving tech giants to rapidly adopt AI-driven services. But there's a catch – these high-tech applications are voracious consumers of power, amplifying our digital footprints with each interaction. And if the recent advancements with Dalle3 and voice-optimized ChatGPT are any indications, we are just getting started – we can anticipate AI becoming as integrated into people's lives as smartphones. This is the crux of a groundbreaking study by Alex De Vries from the VU Amsterdam School of Business and Economics.

However, offering a silver lining, the study suggests that the environmental impact of AI might not be as catastrophic as feared, provided its rapid growth moderates. The caveat? Many in the tech realm, including the report's author, believe these forecasts are sprinkled with conjecture, primarily because tech behemoths remain reserved about sharing comprehensive data.

The study hinges on several constants: the current AI growth rate, chip availability, and servers working relentlessly at full throttle. Delving into the research unveils a striking estimate: tech titan Nvidia is expected to meet nearly 95 per cent of the AI sector's demands. Parsing this data, De Vries anticipates AI's energy consumption could range between 85-134 terawatt-hours (TWh) annually, akin to powering an entire country.

"You're essentially gauging the Netherlands' energy consumption. This equates to a staggering 0.5 per cent of global electricity consumption," he remarked. Market favourite, Nvidia, however, opted to remain silent on the matter.

AI’s Thirst for Power (and Water)

Scratching beneath the surface of prominent AI-driven chatbots like ChatGPT and Google's Bard, as well as image generation engines and the myriad of developers tapping into the Application Programming Interfaces (APIs) of Generative AIs, reveals colossal data centres at the heart. The catch? This cutting-edge machinery isn’t just electricity-hungry; it demands vast quantities of water, depending on intensive cooling systems. Notably, the study overlooked the energy required for cooling, and major tech companies remain ambiguous about these specifics, catalyzing demands for greater transparency. Contrary to initial beliefs, big tech, once hailed as the vanguard of the post-fossil fuel era, isn't as eco-friendly as once thought.

Demand for AI-centric hardware is surging, and concurrently, so is the energy to keep them chilled. Danny Quinn of DataVita, a Scottish data centre magnate, painted a picture: from receiving a mere handful of AI kit inquiries at the outset of 2023, they're now swamped with hundreds. He underlined the pronounced energy difference between standard and AI-driven server racks, accentuating AI's prodigious energy consumption.

Intriguingly, Microsoft's most recent sustainability report unveiled a revelation: a 34 per cent spike in water consumption between 2021-2022, amassing a staggering 6.4 million cubic meters.

Reverse Engineering

Yet, among these environmental concerns, AI emerges as a glimmer of hope, potentially unlocking solutions to some of the planet's urgent ecological dilemmas. Technology exhibits a recurrent pattern: solving problems, spawning new challenges, and then, in turn, introducing more tech solutions for its recent set of issues.

This trend mirrors structural unemployment in the job market. As technology boosts productivity and eliminates roles, it reduces the cost of goods and services, freeing up capital to be reallocated elsewhere.

Recent innovations, such as Google and American Airlines' AI-powered contrail reduction tool, showcase promising eco-friendly advancements. Likewise, the US government's nuclear fusion initiatives, bolstered by AI, might just unveil an infinite green energy reservoir. AI stands poised to be a linchpin for pioneering energy experiments. The most captivating aspect? Big Tech, in addressing its challenges, could potentially carve out an entirely new market niche – energy.

The narrative seems clear: The insights driven by Generative AI will persistently refine. As Big Tech's hunger for high-energy-consuming GPUs and data centres escalates, AI itself might unearth novel methods to generate and store energy more efficiently, capitalizing on renewable resources' low marginal costs. Being the early birds, these tech giants will harness this knowledge to curtail their expenses and subsequently vend the surplus to the masses.

In past commentaries, I've emphasized this: Big Tech transcends mere corporate entities; they embody modern infrastructure. And it appears they're on the cusp of infiltrating the original industry from the Gilded Age – Electricity. Barring any effective antitrust measures, they're poised to keep propelling forward, at least into the foreseeable future.