IS Research Has an Environmental Cost Problem It Mostly Ignores

The paradox that keeps nagging at me is this: organizations pursue digital transformation on the premise that it makes them more efficient. The IS literature on IT business value, from Brynjolfsson (1993) through Melville et al. (2004), makes a case for IT as a driver of productivity and competitive advantage. The efficiency argument is real. But the infrastructure that enables digital efficiency has its own environmental cost, and that cost is large, growing, and almost entirely absent from the IS business value conversation.

Data centers consume a significant and growing share of global electricity. Cryptocurrency mining is extraordinarily energy-intensive, requiring computational work that produces nothing except the proof-of-work verification that secures the currency. Training large AI models consumes substantial electricity: my understanding from publicly reported figures is that training GPT-3 required energy on the order of thousands of megawatt-hours, though I would hedge any specific number I have not verified from a primary source. Bitcoin's global energy consumption has been compared by researchers to that of mid-sized countries, though those estimates vary and I would not treat any specific figure as settled. What I am confident about is the order of magnitude: these are not rounding errors. They are material contributions to global electricity demand, which is still largely met by fossil fuels in most regions.

The Green IT literature within IS research has existed for over a decade. It examines how organizations can reduce the environmental impact of their IT operations, how IS can be used to improve environmental outcomes in other industries, and what governance mechanisms encourage or require environmental accounting for IT. My reading of that literature suggests it has remained a specialty subfield rather than a central concern of IS research. Papers on digital transformation, cloud adoption, and AI deployment routinely omit any discussion of the energy and resource implications of the systems they study.

The governance question is the one I find most tractable for IS research. Should organizations measure and report their IT carbon footprint? Many do not, or do so only voluntarily and inconsistently. The major cloud providers, Amazon Web Services, Microsoft Azure, and Google Cloud, have made public sustainability commitments. Microsoft has pledged to be carbon negative by 2030 and to remove its historical emissions by 2050. Google has claimed to match its electricity consumption with renewable energy purchases since 2017. Amazon has committed to net zero carbon by 2040 under its Climate Pledge. These claims are self-reported and use accounting methodologies that are contested. Renewable energy certificates, for example, do not necessarily mean that the electricity powering a specific data center at a specific moment is renewable. They mean the company purchased credits equivalent to that energy from renewable sources somewhere on the grid. Whether that is the right accounting methodology is a governance question, not a technical one.

IS governance researchers have developed frameworks for understanding IT accountability, compliance, and reporting. Those frameworks apply directly here. What measurement standards should apply to IT carbon accounting? Who should audit the reports? What disclosures should be mandatory versus voluntary? What organizational roles should be responsible for IT environmental performance? These are not engineering questions. They are organizational and governance questions that IS researchers are equipped to study.

The complementarity problem runs in both directions. The same IS literature that argues for IT's efficiency benefits has to reckon with the fact that digital efficiency at the task level does not automatically translate into environmental efficiency at the system level. You can use an AI-powered scheduling system to reduce fleet miles for a delivery company, which reduces fuel consumption. You can also use AI to enable next-day delivery of discretionary purchases, which increases fleet miles overall. The net environmental effect depends on which use cases dominate, which depends on business incentives that are not controlled by the technology. This is exactly the kind of sociotechnical entanglement that IS research is positioned to examine, and it points toward a more honest accounting of what digital transformation actually delivers.

There is also a justice dimension that I think IS research tends to underweight. Data centers are not evenly distributed. They require large amounts of water for cooling and space for facilities. They tend to locate near cheap power, which often means near hydroelectric dams in rural areas or near coal power in regions with low energy costs. The communities that host data centers bear the environmental costs of the infrastructure. The productivity gains from that infrastructure accrue to organizations and workers in entirely different locations. That asymmetry is visible in other resource extraction industries. It is not usually discussed in the IS business value literature.

The question I am left with is whether IS researchers studying digital transformation and AI adoption have an obligation to account for the environmental costs of the systems they are studying, or whether it is legitimate to treat those costs as externalities that lie outside the frame. My read is that treating them as externalities has become increasingly difficult to defend, both intellectually and politically, as the scale of digital infrastructure's environmental footprint becomes harder to ignore.