In Search of a Little Downtime in Asia

Uptime=cold, downtime=hot. It’s a binary that anthropologists have identified in research on data centers and the people who work in them. The idea is that cloud infrastructure workers associate cool temps with technology that works and overheating with equipment, and centers, that don’t. It’s also a meaningful framework to view the digital moment we’re in.

Since OpenAI unleashed ChatGPT on the wider public in 2022, artificial intelligence, specifically generative AI, has taken the world by storm. A Reuters Institute report from this year estimates that 34% of people use an “AI system” at least weekly, compared to 18% surveyed the year before. In the Asia Pacific, the numbers are perhaps even more astounding, and the region outpaces the rest of the world in terms of adoption. According to a 2025 KPMG study, 92% of Indians surveyed reported using genAI at least semi-regularly (tied with Nigeria). This relentless uptake demands more uptime.

Given its ubiquity as a tool for knowledge workers and the supposed efficiencies derived from it, many view the upside of genAI as limitless. This is especially the case for techno-optimist “zoomers,” as described by LinkedIn co-founder Reid Hoffman, or less endearingly referred to as TESCREALists (adherents of “transhumanism, extropianism, singularitarianism, cosmism, rationalism, effective altruism and longtermism”). GenAI is also frequently touted as a democratizing force for knowledge—making information more accessible to more people.

Yet this knowledge isn’t truly free: it comes with externalized costs, including environmental impacts, labor exploitation, and the commodification of data.

Those who aren’t taken by the genAI moment see downsides from preventing downtime. Many, in fact. As journalist Karen Hao outlines in her recently published Empire of AI, from the psychological harm posed by training models to increased social inequality perpetuated by elites profiting from them to the insidiousness of algorithmic bias to the erosion of both government and Big Tech accountability, genAI’s ugly underbelly isn’t so hidden.

The numbers behind data centers’ water use and energy dependence are not unfamiliar to followers of tech news, but they are nonetheless mind-boggling. The carbon emissions for training a large language model and the water consumption of a data center are referenced in the introduction to this special report, so I won’t belabor the point. But the regional context in Asia may be more troubling still. In Batam, Indonesia, many of the island’s poor lack potable water to drink, and will likely suffer further deprivation with the expansion of a data center park there. Meanwhile, at a time when countries must redouble their efforts to not overshoot the 1.5C warming threshold, Malaysia, India and Taiwan are considering a reversal of their net-zero emissions commitments and approach to a renewable energy transition. Instead, they acknowledge, tacitly or otherwise, the need to power data centers with polluting methane gas and coal – in some cases, a middle-of-the-road approach powered by nuclear plants. In these countries, reintroducing fossil fuels and nuclear were quite literally pipedreams for their proponents just a few years ago.

AI-yay-yay

Over the past year, Western media has devoted significant coverage to the natural resource use of data centers and genAI; Asian outlets have begun to catch up. While this may be welcome, in a peculiar twist, old(er)-school technologies, such as cryptocurrency “mining” and cyber scamming (especially the evocative dark art of “pig butchering”), have increased in scope, yet are largely disregarded as environmental threats. This is a major oversight. 

In Asia, Bitcoin mining—the process of using powerful computers to solve complex mathematical problems to validate transactions and add them to a public ledger known as a blockchain—is continuing apace. Laos is an object lesson: The country’s hydroelectric wealth is being redirected in vast swaths to bolster uptime in crypto mining these days. This has implications for the amount of energy and water necessary to maintain Rube Goldberg mining contraptions that collectively run quintillions of calculations or “guesses” per second.

As economist Alex de Vries has noted in his research on Bitcoin, the machinery used for the guessing quickly becomes  obsolete, creating an enormous e-waste problem to boot. Said equipment utilizes large amounts of copper, silicon, and rarer elements still. And as it turns out, the same precious critical minerals that appear to be driving geopolitical conflict between China and the West are also getting diverted toward the ends of investors and a not-insignificant number of crypto bros.

Cyber scam operations, meanwhile, have made Asia —particularly Southeast Asia—their preferred headquarters. From rebel factions in Myanmar to the underground economy in Cambodia, pig butchering and associated scamming yield billions of dollars for their beneficiaries. And unlike one’s local abattoir, these facilities run day and night, employing people in dreadful conditions. Despite the extensive coverage of these scams, there is little scrutiny of their environmental costs and the strain they place on already energy-starved local communities. These operations rarely run on renewable energy grids; instead, they are often powered by dirty fossil fuels, compounding the social and ecological toll.

The Greener Side of Slate

It’s easy for the public to become focused on the alleged human rights abuses of this new digital infrastructure and what it portends for people desperate to escape energy poverty, take a sip of clean water, or appreciate humane labor conditions. Those angles alone would be valuable for any reportage, but it would be folly to ignore the resource-intensive nature of this tech. Impacts that deepen energy scarcity, infringe on land rights or worsen access to potable water make vulnerable communities even more so.

Many of the environmental impacts of Big Tech often escape media scrutiny, not because they’re insignificant, but because they lack the obvious visual cues of destruction. Belching smokestacks and soul-crushing bulldozers are historically recognizable signifiers of devastation. But while new(er) digital technologies seem innocuous, they’re quite the opposite. A couple of lonely engineers manning a faceless, sanitized data center as it whirs along, emptying a bathtub’s worth of water every second; a Bitcoin mine engaged in an infinite guessing game sucking up electricity and leaving communities in the dark; a scam center resembling a bland corporate office park, which somehow reeks of diesel smoke from generators that never stop — these are hardly the image technologists sold us on when they spoke of a glorious “TechnoGarden” in the UN Millennium Assessment Report 20 years ago. (Do they serve high tea at that garden party?)

Despite these impacts, the AI resistance movement is not likely to make significant headway for the foreseeable future. So, what options remain to lighten the dark side of the boom?

In a more timely UN publication, UNESCO recently pointed out that fine-tuned, small language models running on your desktop, rather than their grandiose large-language cousins, could be the solution to our reliance on water-guzzling and renewable energy-draining servers. Similarly, there are cryptocurrencies on the market not nearly as rapacious as Bitcoin when mined. And even marginally cracking down on the illegal provision of supplies to scam centers could do wonders to curb the industry’s growth. 

All of which is to say: if we’re serious about the planet, then downtime shouldn’t just be tolerated—it should be prized. Dare I say a hot item, but in a good way.