1) The digital world emits as much as some top-emitting countries

Depending on how you count it, the information and communications technology (ICT) sector—devices, networks, and data centers—accounts for roughly 1.5% to 4% of global greenhouse gas emissions. That range overlaps the annual emissions of some of the world’s largest emitting countries. The exact number is debated because the sector evolves quickly and boundaries (e.g., whether to include crypto or embodied manufacturing emissions) differ across studies.

What’s certain: data centers alone consume around 1% to 2% of global electricity, and the broader ICT sector consumes several percent. As more of life moves online and AI workloads grow, demand will keep rising unless tempered by efficiency and clean power.

2) The energy per gigabyte has crashed—yet total demand still soars

Over the past decade, the energy required to move a gigabyte across the internet has plummeted—improvements of 10x to 50x are commonly reported thanks to better hardware, networking, and data center design. But there’s a catch: traffic growth has outpaced efficiency gains. We stream more, game more, back up more, and now run AI everywhere. The result is a classic rebound effect: more efficiency can lead to more total consumption.

3) AI is powerful—and can be power hungry

Training and serving large AI models can draw significant electricity and water for cooling. The environmental impact varies widely by:

• Data center location and grid mix (coal vs. renewables)
• Hardware efficiency (newer chips can do more with less)
• Cooling strategy (air, water, or advanced systems)
• Model size and how often it’s retrained or queried

One widely discussed finding: training a state-of-the-art model can directly consume hundreds of thousands to millions of liters of water for cooling, depending on geography and timing. On the flip side, AI also helps optimize grids, forecast renewables, cut food waste, and detect methane leaks—reducing emissions elsewhere. Whether AI is a net positive depends on what we use it for and how cleanly we power it.

4) E-waste is the world’s fastest-growing waste stream

The world generated on the order of 60+ million metric tons of electronic waste in 2022—more than the weight of the Great Wall of China. Only about one-fifth was formally collected and recycled. The rest was landfilled, informally processed, or lost—squandering tens of billions of dollars in recoverable materials like gold, copper, and rare elements, and exposing workers to toxic substances when handled unsafely.

Most of a gadget’s lifetime emissions occur before you ever open the box. Extending device lifetimes by just one or two years is often the single most impactful consumer action.

5) Your smartphone is a tiny mine

A typical phone contains dozens of elements—including cobalt, nickel, copper, gold, and rare earths—sourced from complex global supply chains. Mining can drive deforestation, water contamination, and human rights risks when poorly managed. The good news: design changes (modular phones, fewer material types), certification programs, recycled content, and “right to repair” laws are starting to bend the curve. But the cleanest kilogram of mined metal is the one you never need because you kept your device longer.

6) Data centers keep getting cooler—literally and figuratively

Power Usage Effectiveness (PUE), a metric of data center efficiency, has fallen dramatically over the last two decades. Many hyperscale facilities operate near a PUE of 1.1–1.2, versus ~2.0 in the 2000s. That means far fewer watts are wasted on overhead like cooling for each watt used by servers. But PUE doesn’t capture everything—it ignores the carbon intensity of electricity and the water impact of cooling. More operators now report water metrics and seasonal carbon intensity as well.

7) Water is a quiet cost of the cloud

Beyond electricity, large facilities use significant water for evaporative cooling and for manufacturing chips (which require ultra-pure water). Water use varies by design and local climate; some sites use air or liquid-cooled loops with minimal water. The key lever is siting: choosing regions with resilient water supplies, investing in recycling, and aligning heavy workloads with cooler hours or seasons can slash local water stress.

8) Crypto can rival countries in power demand—and in hardware waste

Depending on price and technology, Bitcoin alone has consumed on the order of 100–200 terawatt-hours of electricity per year—comparable to a medium-sized country. Proof-of-stake systems (like Ethereum after its 2022 transition) cut energy use by orders of magnitude, but proof-of-work chains still dominate crypto’s footprint. There’s also the hardware churn: specialized mining rigs quickly become obsolete, contributing to substantial streams of electronic waste.

9) Streaming isn’t “free”—but context matters a lot

An hour of HD video can use gigabytes of data, and the energy behind each gigabyte depends on network efficiency and grid mix. Watching a stream on a large TV draws more device power than on a phone; downloading once and watching offline can be more efficient than streaming repeatedly. The biggest lever isn’t individual streaming choices—it’s decarbonizing the electricity powering networks and data centers.

10) The night sky is changing

Large constellations of low-Earth-orbit satellites improve global connectivity and Earth observation, but they also contribute to light pollution and streak astronomical images, complicating research and affecting cultural heritage of dark skies. Mitigations—darker coatings, sunshades, lower-reflectivity designs, and altitude choices—are being tested, but the environmental and scientific trade-offs are still evolving.

11) Undersea internet cables can become accidental marine sanctuaries

Subsea fiber-optic cables carry most international data. Installing them disturbs seafloor habitats locally, but the cable corridors often become de facto no-trawl zones, allowing marine life to recolonize. While not a substitute for protected areas, it’s a rare case where digital infrastructure can provide incidental ecological benefits when responsibly deployed.

12) Software bloat has a carbon shadow

The average webpage has ballooned over the past decade, and inefficient code wastes energy across billions of devices. Leaner apps and sites reduce network traffic and device power draw, improving battery life and slashing aggregate emissions. “Green coding” practices—efficient algorithms, adaptive refresh rates, and smarter caching—scale surprisingly well because they multiply across users.

13) Smart doesn’t always mean greener—unless it’s designed that way

5G networks, smart thermostats, and IoT sensors can cut energy use by making systems responsive. For example, smart thermostats often reduce household heating and cooling energy by around 10% or more. Yet billions of “always-on” gadgets also add up: standby power (“vampire energy”) quietly consumes electricity even when devices seem idle. The net impact depends on efficient default settings, clear power management, and durable design.

14) Digital twins and 3D printing can turn bits into big material savings

By simulating factories, buildings, and cities before construction, digital twins can spot inefficiencies that would be expensive to fix later. Additive manufacturing (3D printing) often uses less material than subtractive methods and enables lightweight designs that save fuel in cars and planes. Pairing these tools with low-carbon materials compounds the benefit.

15) Clean power turns tech from problem to powerhouse

When data centers and factories run on renewable electricity, their operational emissions plummet. Large tech firms have helped scale wind and solar by signing long-term power purchase agreements and, increasingly, by investing in 24/7 carbon-free energy portfolios that match clean power to consumption every hour. The next frontier: ensuring new demand adds new clean capacity and improves local grids rather than just buying credits elsewhere.

16) Batteries enable renewables—but their supply chains need care

Lithium-ion batteries help stabilize grids and electrify transport, cutting tailpipe emissions. Yet mining lithium, nickel, and cobalt has environmental and social impacts. Progress is rapid on:

• Recycling technologies to recover lithium, nickel, and cobalt at high yields
• Chemistries that reduce or eliminate cobalt
• Better traceability and certification to address labor and biodiversity risks

17) Precision agriculture and satellites cut emissions you never see

Technologies like soil sensors, drones, and satellite analytics help farmers apply fertilizer and water only where needed—often trimming inputs by double-digit percentages while maintaining yields. Methane-detection satellites and AI analysis can find leaks from oil and gas infrastructure quickly, enabling repairs that avoid large, short-term climate impacts.

18) The greenest device is the one you already own

Manufacturing typically dominates the lifecycle footprint of phones and laptops—often 70% to 90% of their total emissions. Keeping a smartphone or laptop for an extra year or two usually beats any marginal efficiency gains from a new model. Repairs, battery replacements, and refurbished markets meaningfully reduce emissions and e-waste.

19) Cloud can be greener than on‑prem—but only if you use it wisely

Because of higher utilization and modern cooling, cloud workloads can have a lower footprint than servers in small, inefficient data rooms. But moving to the cloud doesn’t automatically reduce emissions. The big wins come from:

• Choosing regions with clean electricity and low water stress
• Right-sizing instances and turning them off when idle
• Using serverless and autoscaling to avoid overprovisioning
• Measuring and optimizing embodied emissions in storage and compute choices

20) We can measure what matters—and that changes behavior

New tools can estimate the carbon and water impacts of software, track supply-chain emissions, and forecast the hourly carbon intensity of grids. When procurement and engineering teams see those numbers, they make different choices—shifting workloads to cleaner hours, picking lower-footprint materials, or redesigning features for efficiency.

Nuance matters: technology is a tool, not destiny

Technology is neither inherently green nor dirty. Its impact depends on design choices, energy sources, and what problems we point it at. The same compute that trains a frivolous model could instead stabilize a wind-heavy grid. The same phone that replaces a camera and a GPS device might also fuel upgrade culture. The decisions we make—policy, engineering, business, personal—determine which future we get.