Data is an asset and, like all assets, it comes with overhead. Storage, compliance, handling and control - it all adds up. Our current centralized data paradigm doesn’t just cost a lot of money, but drains the lifestream of the planet itself. Computers, the paragons of supposed efficiency, are hungry for energy, water, and resources. Data is a fossil fuel. It’s not just the raw compute, but the whole data management architecture we operate in that vastly contributes to the global carbon output that's ruining the rock we call home.

We’ve spoken at length about how local-edge distributed data management systems can revolutionize sectors like agritech, energy, and industry and, in doing so, vastly decrease the environmental burden of these sectors and their productivity output. Having edge devices operate independently of the cloud results in efficiency and runtime gains that reduce the environmental overhead of these core global systems. Moreover, by delivering their keystone operations better, it has knock-on effects for society as a whole.

The larger problem, though, is one of how we manage all data - everywhere. How we gather, send and especially save data creates emissions and harms our environment. We are producing data at a logarithmic rate, and it will only increase. Centralized data storage and cloud-efficient operations are absolutely not carbon efficient. Giant centralized compute and storage mainframes ripped out of some dystopian 90s cyberpunk are just as scarring to the environment as the coal furnaces, tanneries, and lime kilns of the Industrial Revolution.

The new Stargate is just the most egregious example. The fact is that every cloud server everywhere contributes to carbon overhead due to the way they fundamentally manage data. By reversing that principle, we can create not just more secure, sovereign data, but more efficient, cleaner data too. Happy data, healthier planet.

Hungry Elephants: The Cost of Data Centers

Data centers in the European Union are predicted to account for 3.21% of total electricity demand by 2030. The internet itself was estimated for 2% of global energy consumption in 2020, a number that has only increased. The cooling systems, raw power draw and over-provisioned redundancy systems in place at these centers are drastically inefficient, their expense only covered by the value of the data itself. 

There’s also the surfeit of always-on backend services that constantly draw on resources even when idle. We all know switching the TV off and not leaving it on standby saves us money. In an average data center, there are thousands of devices switched on just-in-case. These systems are built to scale even if demand is sporadic so that their services can compute, but we all know elasticity in operations is not the equivalent of efficiency. Server mirroring, backup infrastructure, even diesel and gas generators. All of this is excess waste due to the nature of the centralized model itself. Transferring data isn’t free - it’s an under-discussed energy drain. Every query and every API call is an arrow that travels across the globe through a faroff data center to stick in the heart of the earth itself. 

That’s just storage and serving data, but what about analysing or using it? The computational load of sifting through all this unstructured data which has been acquired in the ‘hoover’ strategy of the last decades, leaves us with a situation where the environmental burdens caused by the centralized server strategy is becoming overwhelming. The volume of “dark data” is growing, and estimates suggest that up to a staggering 88% of data stored, processed and operated on by corporations could be ROT (Redundant, Obsolete or Trivial). When you figure that ROT data is just as carbon expensive as useful data to maintain, the problem becomes obvious. More painful still, you have compliance and logging overhead (a whole new set of compute and storage demands) for all that junk data due to the reactive systems we have sleepwalked into building.

The Ecological Edge

The current cloud model is an ecological disaster. Data centers can sign all the self-flagellating treaties they like, the fact is no amount of renewable energy or freshwater will stop the environmental scarring caused by rampantly escalating data centers - especially in an era of AI. If we’re only ever going to produce more data, we need to figure out how to manage it better.

Edge-native systems are the key to drastically reducing data waste. By edge, we mean true edge, not just more local data centers. We mean on-device processing of data, local storage, reduced data proliferation, and necessity-based access. To stop our data from constantly being shipped to and hoarded in technotropolis data mines which guzzle on our privacy and on our planet. If data lives near where it needs to be used, and contains within its own permissions for use and decrees for syncing and so on, we can reduce the waste associated with data processing not by 10%, but by 90%.

If data has its own access rules, you don’t need constant external queries to perform on it. Edge devices can sleep and spin up only when needed, rather than being a constant idle draw. Systems can be smaller and task-specific (i.e cameras on a highway) with optimized processes for its utility, not reliant on tapping into a general-purpose server. It also dramatically reduces the proliferation of ROT data. If storage is local and with intent, data hoarding is mitigated - with edge networks efficient by design. 

Lower latency, lower bandwidth demands, lower infrastructure footprint. All contribute to better environmental outcomes for our societal systems. The efficiency gains in those sectors adopting a local edge infrastructure further compounds this benefit. If we are dead set on building out AI, we can at least do it at the edge, reduce their carbon footprint, and bring sustainability back from the brink.

Many Devices, One Planet

Source Network helps developers create edge-native software that store, process and coordinate data locally without necessary recourse to an environmentally problematic data center (but our stack is interoperable with current environments). DefraDB manages distributed data, LensVM for schema transformation, SourceHub for trust - Source empowers applications to work at the edge and reduce their reliance on the cloud, and remove the absolute need for it entirely. We are empowering the ecological edge. We are championing the ethical management of data. We are creating better ways to build - for people, for systems and, most crucially, for the planet.