Zum Inhalt springen

The History of Green IT

Zusammenfassung

Computing feels weightless — bits, clouds, streams — but it runs on electricity, rare metals, water, and a relentless churn of physical hardware. Green IT (also “green computing” or “sustainable IT”) is the effort to measure and reduce that footprint: the energy data centers consume, the carbon they emit, the mountains of e-waste discarded devices become, and the resources mined to build them. Once a niche concern of facilities managers worried about electricity bills, it became a strategic and ethical issue as data centers grew into industrial-scale infrastructure and, most recently, as the AI boom sent computing’s energy demand sharply upward. This article traces four intertwined threads — the energy cost of computing, e-waste and the hardware lifecycle, energy-efficient hardware, and the move toward renewables — and the gap between green marketing and physical reality.

The Energy Cost of Computing

For decades, computing’s energy appetite was hidden inside the rounding error of national grids. That changed as the internet industrialized. By the late 2010s, data centers consumed roughly 1–1.5% of global electricity — and that figure held remarkably flat through the 2010s even as compute demand exploded, thanks to efficiency gains and the shift from small inefficient server rooms to hyperscale facilities.

The key metric is PUE — Power Usage Effectiveness (introduced by the industry consortium The Green Grid in 2007): the ratio of total facility energy to the energy actually delivered to computing equipment. A PUE of 2.0 means every watt of computing costs another watt of overhead (mostly cooling); the best hyperscale data centers drove PUE down toward 1.1, a dramatic efficiency gain. Google famously applied DeepMind machine learning to its cooling systems to shave further percentage points.

Cooling is the central engineering challenge: servers turn electricity into heat, and removing that heat consumes energy and, increasingly, water. Some facilities evaporate millions of liters annually, putting data centers in conflict with drought-stricken communities — a sustainability cost that PUE alone doesn’t capture.

Two demand shocks reshaped the picture. Cryptocurrency mining (especially Bitcoin’s proof-of-work) grew to consume electricity on the scale of a mid-sized country, generating fierce debate about waste. And from 2023, AI training and inference — with its hunger for power-dense GPU clusters — broke the decade of flat data-center energy growth, prompting hyperscalers to sign deals for dedicated power including, strikingly, nuclear capacity (Microsoft’s 2024 agreement to restart a Three Mile Island reactor being the emblematic example).

E-Waste and the Hardware Lifecycle

Every device eventually dies, and computing produces an enormous waste stream. The UN’s Global E-Waste Monitor reported a record 62 million tonnes of electronic waste generated in 2022, of which under a quarter was formally documented as collected and recycled. E-waste is both hazardous (lead, mercury, cadmium, flame retardants) and valuable (gold, copper, rare earths) — and much of it flows, often illegally, to informal recycling hubs in the Global South, such as Agbogbloshie in Ghana, where workers burn cables to recover metal at grave health cost.

The policy response built up over two decades. The EU’s WEEE Directive (Waste Electrical and Electronic Equipment, 2003) made producers responsible for collection and recycling, while RoHS (Restriction of Hazardous Substances, 2003) banned lead and other toxins from new electronics. These embody Extended Producer Responsibility and the ideal of a circular economy — designing for repair, reuse, and recovery rather than the linear “take-make-dispose” model. The grassroots right-to-repair movement, plus the EU’s 2022 decision to mandate USB-C as a common charger, attack the planned-obsolescence and accessory waste that inflate the hardware churn.

Efficient Hardware and the Renewable Push

Much of computing’s sustainability story is, paradoxically, written by the relentless pursuit of performance per watt. The defining example is ARM: the low-power RISC architecture (see Acorn/ARM) that conquered mobile precisely because energy efficiency was its design goal, and which by the 2020s was displacing power-hungry x86 chips even in data centers (Amazon’s Graviton, Apple Silicon). The broader RISC philosophy, dynamic voltage/frequency scaling, and specialized accelerators (TPUs, efficient inference chips) all push the same lever.

On the consumer side, the US EPA’s voluntary Energy Star program (launched 1992, with computers among its first product categories) standardized efficiency labeling and drove the now-ubiquitous sleep mode. On the supply side, hyperscalers became among the world’s largest corporate buyers of renewable energy, signing power-purchase agreements and siting data centers near hydro, wind, and solar — and near the cold climates (Nordic countries, see the Nordic tech industry) that cut cooling costs. Google claimed to match 100% of its annual electricity use with renewables from 2017, and set a far harder goal of 24/7 carbon-free energy by 2030 — matching consumption hour by hour, not just on annual average.

Dead End: Carbon Accounting Theater

The recurring failure of Green IT is greenwashing through accounting — claiming carbon neutrality on paper while physical emissions continue.

The mechanism is the gap between annual matching and real-time decarbonization. A company can buy enough renewable energy certificates or carbon offsets over a year to claim “100% renewable” or “net zero,” even though its data centers actually draw fossil-fueled grid power at night when the sun isn’t shining and the wind is calm. The electrons powering the servers at 2 a.m. are often coal or gas; the green claim is a financial instrument layered on top. Carbon offsets drew especially harsh scrutiny when investigations revealed many forest-protection credits represented reductions that would have happened anyway or didn’t survive.

The industry’s own leaders effectively conceded the point: Google’s pivot from “annual renewable matching” to 24/7 carbon-free energy was an admission that the original claim, while not false, papered over the physical reality of a still-carbonized grid. Meanwhile the AI surge of the mid-2020s blew holes in several big-tech climate targets — Microsoft’s and Google’s reported emissions rose as they built out AI capacity, despite their pledges. The honest lesson of Green IT’s history is that efficiency gains are real and substantial, but they are repeatedly outrun by growth in demand (a manifestation of the Jevons paradox: making computing cheaper and more efficient makes us use vastly more of it). Sustainability claims that ignore the rebound of demand are accounting theater, not decarbonization.

📚 Sources