Your phone got “smarter” over the last two years. It also got hungrier — for electricity, for bandwidth, for data centre storage, for your battery, and for your money when the “unsupported device” notice finally arrives. Almost nobody asked for this. The bill is landing on all of us.
Google promised net zero. Then it built the AI boom.
Google has committed to net-zero emissions across its operations by 2030. It is moving in the opposite direction: Its 2026 environmental report shows a carbon footprint that rose significantly each year. In 2024 the gap was 48%, and has widened considerably since, and now sits 81% above the 2019 baseline it measures its halving target against. Google’s own sustainability chief has conceded the cause: the AI buildout is accelerating faster than the grid is decarbonising.
Nobody voted for this
The AI features driving that buildout are neither opt-in nor popular. In North America, Pew’s June 2026 survey found 52% of adults more concerned than excited about AI’s growing role in daily life, against just 9% the reverse. Among under-30s, concern hit 55%, up from 31% in 2021.
Now look at what’s shipping anyway. Chrome has been silently downloading a roughly 4GB AI model onto users’ browsers in the background, no prompt, no opting out. Delete it and Chrome re-downloads it. One privacy researcher estimated the climate cost of a single model push at scale between 6,000 and 60,000 tonnes of CO₂e. Meanwhile Gemini AI is being woven into Android at the system level, and Apple Intelligence AI is built in rather than offered as an app you choose. Technologies most people are wary of, installed without consent, with the off switch progressively harder to find.
The AI that doesn’t wait for you to prompt
Here’s the part that gets missed: the information flow to data centres doesn’t start when you type a query to an AI. It never stops. Research from Trinity College Dublin found that iPhones and Android handsets transmit data to Apple and Google on average every four and a half minutes – while sitting idle in a pocket. Hardware serial numbers, IMEI, SIM details, Wi-Fi MAC addresses, phone numbers, location data assembled from GPS, Wi-Fi and Bluetooth; motion and ambient sensors. Both platforms kept transmitting telemetry even when users explicitly opted out, even before any programs were installed. The device in use also sends keyboard input; health and fitness metrics; app usage patterns. Each of those streams is a connection to a server, processed and stored somewhere, and few into predictive models that draw power around the clock.
This is where the privacy problem and the climate crisis stop being separate stories. The surveillance is the workload. Every intimate data point harvested is a data point transmitted, processed, indexed and retained – and the appetite for AI training and inference has made that raw material more valuable, not less. Adding AI to the phone adds both energy demand and a reason to collect more data.
On-device privacy?
Chipmakers now offload AI work to a Neural Processing Unit, marketed as the privacy-respecting option. Take that claim carefully. These are companies whose revenue depends on knowing what you do; a local model doesn’t change the business model, and “processed on-device” is a statement about one step in a pipeline, not a promise about the whole of it.
The energy cost hasn’t vanished either – it’s been externalised to your pocket: a sealed glass-and-metal sandwich with no fan. Researchers testing sustained on-device AI showed devices heating under load, and a phone that runs hot more often is a phone that feels old sooner, not to mention that heat degrades lithium-ion batteries. Both can contribute significantly to a working device ending up sooner in a landfill.
AI as the new planned obsolescence
Apple Intelligence requires 8GB of RAM, which permanently excludes, by hardware, all but the most recent iphones. Google’s next-generation on-device AI reportedly demands 12GB, a bar that excludes even the Pixel 9 – a criteria that critics have openly questioned as commercial rather than technical.
This matters because of where a phone’s emissions actually come from. Life-cycle assessments consistently put roughly 80% of a smartphone’s total carbon footprint before you ever switch it on – mining, chip fabrication, assembly, shipping. The use phase is around 13%; academic studies put manufacturing at 70–90% of lifetime emissions. So charging with green electricity barely moves the needle. Keeping the phone two more years does. Extending a device’s life from 2.5 to 5 years cuts lifetime emissions by roughly 35%, and each second-hand transaction extends use by 40%, also cutting the annual footprint.
No AI feature justifies discarding a working phone.
Degoogled Android is different
The AI spying is gone. A degoogled system doesn’t ship Gemini, and the OS doesn’t quietly decide to ship your interactions to a data centre. Nothing is sent off-device for processing until you choose. On top of that, iodé’s blocker maps every recipient your apps try to reach and blocks trackers and prevent malware-serving domains, on-device. Blocked trackers aren’t only a privacy win: they’re requests that never leave your phone, never cross a network, ads that never get served, and code that doesn’t run on your device or on a server. Privacy and energy efficiency are interconnected.
Old hardware stays useful. iodéOS runs on more than 60 devices from a variety of manufacturers: those focused on sustainability such as Fairphone and Shift, but also from Google, Sony, Motorola, Volla, Braxtech, Xiaomi, Samsung and OnePlus – and continues to support phones whose makers stopped shipping updates. A phone that still runs is a phone you can justify keeping, which is the single highest-impact environmental decision available to you.
While we also sell new devices, we encourage you to consider refurbished with iodéOS pre-installed. When selling new devices, we focus on manufacturers like Fairphone and Shift that are committed to sustainability. Buying anything has a footprint, but everyone eventually replaces their phone. The real question is whether you buy into a cycle designed to obsolete your device every two or three years, making room for features you didn’t request – or a device built to last, running software that answers to you.
Your next phone should run iodéOS.
