Data Processing Resource Usage

If all data processing is environmentally unsustainable – and it is – from a resource consumption viewpoint, GenAI processing goes far beyond rational.

The largest competitor for the water and electricity needed to run GenAI data centers is human beings. Three years ago Google released the amount of water it needed to cool its data centers in 2021: 4.3 billion gallons! One company, for one year…three years ago, before GenAI became a common corporate goal…used as much water as a city of 120,000 people.

And then AI’s energy consumption. The CEO of ARM Holdings, stated in April 2024 that GenAI processing, already consuming between 3% and 4% of the US electricity supply, could consume as much as 20% to 25% of the US electricity supply by 2030 if current consumption rates prevail.

What happens when resource providers need to decide which resource is prioritized:
data centers or people?


The Greenest Possible Processing We Could Imagine

We know that Augmetrics® is not a universal solution to sustainability problems that we face, but we also know it is a start; one that took over 10 years to develop.

Cost control and environmentally sound processing were designed into Augmetrics® from its inception, with the goal of making Augmetrics® the greenest possible processing we could imagine; a mission that drives us every day. Information on demand is the central concept of its processing architecture using an NLP-supported system to democratize data access, capture expert context and thus to create and proliferate the demand.

How do we do that?

First we limit the amount of “fixed” processing overhead to basic tables and what it takes to update them according to the user’s needs. Then we apply “information on demand” principles, processing only information requested by the users and doing so according to the following rules:

  • Augmetrics® does not duplicate information processing. When one person creates an information object (think of IOs as broadly relating to a single fact, like a KPI), it is processed once, is instantly available to all users for use in any form of communication (i.e. reports, presentations, personal dashboards, emails, etc.)
  • There are no joined or merged data sets when building IO values from multiple data sources. The ‘joining or combining action’ happens at the UI level.
  • Processing control and efficiency are dramatically improved when peer to peer (P2P) processing combined with Time to Process (TTP) budgeting minimize spin-up of unnecessary container(s) and network resources.
  • All Augmetrics® network resources to date have used lower cost, lower resource-consumption non-GPU chips and we can see no reason to change that in the future.

“In a world that increasingly boils oceans of data for minimal gain, the risk of boiling the planet exponentially accelerates as those models fail to deliver on informational speed to value.”