Clean energy buying incentives for data centers

Incentives are the key to understanding all behavior. If you understand the incentives, you can understand why something has happened or predict what is likely to happen. That doesn’t mean it’s easy to design good incentives. It’s notoriously hard. Bad incentives alter behavior in ways that betray their original goals because the incentive wasn’t properly understood.

Prices act as an incentive. If we want to encourage the adoption of a product, we can reduce its price or make alternatives more expensive. That works in many cases, but only if price is the sole incentive.

Other metrics can act as incentives. If we assign a software sustainability score based on the consumption of clean energy, we can encourage moving workloads to regions with greater abundance of clean energy. If that score is calculated on a granular enough basis – by the minute or hour – we can also encourage workload scheduling at times of more abundant clean energy.

However, the scoring inputs matter.

A sustainability score based on emissions accounting works on an attributional basis. This considers the impact of individual actions. A data center buying green power can reduce its own emissions through matching (of whatever type), thereby improving the sustainability score. Considering only the sustainability score (ignoring reliability, latency, uptime), there would be an incentive to move workloads to this data center.

A sustainability score based on marginal emissions works on a consequential basis. This considers the impact on the system as a whole. A data center buying green power would have no direct impact on its own emissions and no impact on the sustainability score for workloads running there. The impact of that green power depends on whether it is generating as the marginal power plant at any given time. There would be an incentive to move to that region, but not that data center.

Therefore, using a marginal emissions-based sustainability score incentivizes migrating workloads to a region with a lower marginal emissions factor. The emissions factor for that workload would reduce, but does it incentivize the deployment of new clean energy?

What is the incentive that encourges data center operators to buy clean energy? Is it their desire to do good by helping to decarbonize the grid? Maybe. Is it because their customers and regulators are applying pressure? More likely, and this is growing. I’d bet on the latter, not the former. Competitive advantage leads to customer demand.

Assessing grid emissions on a consequential basis is a good approach for systems-level thinking and overall decarbonization goals. But who is buying all the clean energy? What is the goal of a sustainability score? Is it to reduce the emissions impact of a specific workload? Or is it trying to encourage system level decarbonization? If the latter, can a sustainability score based on marginal emissions that ignores the actions of individual data centers achieve that?