Data’s carbon footprint can’t be kept a dirty secret
The concept of data as the ‘new oil’ has served as a crude metaphor for this intangible commodity’s increasing value to business. It doesn’t hold much weight under scrutiny, though.
Data’s value is in its effective use, not hoarding to pump the price. Data can be a self-perpetuating resource and any business with the right server and analytics software can tap it and start drilling for insights. Insights that may indeed be valuable to that business but not necessarily marketable to others, either because of their proprietary nature or because it would be a substantial risk to reveal them to competitors.
The oil metaphor does chime pretty well with data’s environmental impact, though. The bits themselves are invisible to end users and our imagining of data as a non-physical entity is helped along by terms such as ‘the cloud’, from which data streams back and forth.
But these ‘clouds’ are more hard-edged and hulking than the vision they inspire. In the world of data, clouds are massive warehouses of servers, operating 24/7 and using so much power they need constant cooling so they don’t completely burn out.
‘What use is this resource-intensive data economy if it’s just to collate an eternal backlog of all our internet likes?’
Many major data centre announcements these days come with a commitment to 100pc renewable or so-called ‘clean’ energy, which at least shows that businesses are accounting for the significant environmental impact of these data powerhouses. Less shouted about in press releases are the ‘dirty’ data centres that are widely operational, burning through fossil fuels at an alarming rate.
Until very recently, Amazon was noticeably reluctant to disclose the carbon footprint of its business, which amounts to that of a small country. As well as a widespread logistics business to factor into this equation, there’s Amazon Web Services – an infrastructure that keeps much of the internet as we know it running. AWS is powered by a network of 50 data centres around the world and almost half of these run on fossil fuels.
We often hear of emerging technologies made possible by the availability of more and more computing power, but not about where this power comes from and its environmental impact. Cryptocurrency is yet to make it mainstream but if it were to scale, the energy requirements would be colossal. In 2017, the fact that the verification of bitcoin transactions consumed more energy than all of Ireland made headlines. Today, that annual energy use has more than doubled.
However, crypto aversion is not enough to take financial services off the hook. This industry, which relies heavily on data and analysis at breakneck speed, has significant energy needs, not to mention a long, polluted history of investment in the oil and gas industries.
After blockchain, artificial intelligence is being ushered in as the inevitable next phase of computing. But turning computer architecture into something resembling our neural pathways is a resource-intensive process. If we’re thinking in terms of carbon (and we should be) the average human has an annual carbon footprint of about 5,000kg. In its lifetime, a car will emit around 57,000kg. Taking a huge step up, training a large AI model can emit more than 280,000kg of carbon.
The University of Massachusetts study that revealed these figures also warned of the increasing cost of building and training AI, which means well-funded private enterprises will likely take the lead on its research and development, motivated solely by commercial interests. This could mean the reward for all that energy input will be less novel drug discovery and more digital surveillance devices disguised as assistants.
What use is this resource-intensive data economy if it’s just to collate an eternal backlog of all our internet likes? We need to weigh up the cost of data mining against the results.