Google claims that since 2017, it has matched 100% of its global energy consumption with renewable energy purchases. The AES announcement, however, represents the next evolution in the vendor's sustainability efforts as it works toward its 2030 goal of being carbon-free 24/7. The main difference between the 2017 milestone and the 2030 goal is load-matching – essentially, it's a matter of timing. In the US (and elsewhere), Google has invested heavily in both solar and wind energy generation, but neither of those generates electricity every hour of any given day (notably, the company is also investing in geothermal).
To compensate, many organizations will pay for more energy generation during daylight hours (using solar as an example) than they actually consume, such that on an annualized basis, the amount of energy produced is equal to – or greater than – the amount of energy used. Google's ambitions, however, are hour by hour, so more energy produced in one hour has no bearing on the next. During times of low – or no – energy production as compared with the vendor's actual usage, Google inherits the grid's carbon density, which in Virginia is not good (the state's grid is still heavily coal-fired).
To chip away at this problem in Virginia, the company has tapped AES to assemble 500MW of renewable energy generation (we estimate Google has less than 200MW of IT load in Virginia today), which will be complemented by battery storage. During periods of non-generation, the idea is that the stored energy from the renewable sources can be pushed out on the grid, thereby continuing to offset Google's demand on an hourly basis. Despite the $600m expected investment, the deal is anticipated to only cover 90% Google's demand on an hourly basis, underscoring the challenge – and expense – of decarbonizing loads that reside in dirty grids.
The datacenter opportunity
By now, it's no secret that the hyperscale segment has been a major source of demand for the leased datacenter industry. It's also no secret that the hyperscalers have all set lofty goals for decarbonization and aim to do so in short order (many are following the UN's 2030 targets).
The obvious statement here is that to the extent that datacenter providers can be leaders in finding and establishing paths to decarbonization, the better opportunity will be in landing hyperscalers, otherwise the hyperscalers will have to work all the more to offset what they lease. Beyond hyperscalers, however, there are a whole host of enterprises that are looking to better their carbon impact and for those tech-heavy firms, their datacenter portfolio is a major contributor to their carbon emissions.
What if the datacenter industry used its leverage with the power suppliers to drive greener power options? What if, like Google, the datacenter providers insisted on decarbonizing whether the local power suppliers are onboard or not? What if the datacenter industry began to roll out managed-type services to assist companies in identifying their carbon impact, and also offered a way forward to reduce that impact?
The Google-AES collaboration perhaps gives us a framework for what this could look like. It reflects an aspiration to report the company's real-time electricity consumption, exposing the carbon density with the goal of matching it with CO2-free generation in near real-time. This represents a new level of accountability in reducing CO2 footprint than the industry has seen previously. It's logical, then, to assume that in the future datacenters may face pressure to account for their energy usage and CO2 footprint with this same level of transparency.
Datacenters will increasingly want to understand the aggregate load they manage within each interconnected grid, and the good news here is that much work has been done already to do this. Between building management systems and more advanced DCIM systems, providers generally know this already, oftentimes on a real-time basis. The next step will be to understand the generation profile of their current energy usage; investments in renewable generation, including wholesale PPAs; and on-site solar and wind, along with other CO2-free generation, for the purpose of understanding where mismatches between renewable and nonrenewable portfolios exist.
At first it's understandable that this will be for internal use only, but we believe it is possible that the datacenter industry will eventually feel pressure to disclose this to their customers. Where substantial mismatches exist, datacenter operators should look to market providers to offer real-time CO2-free services, including additional green energy investments, grid services, and reservation of storage. Google opted to contract with AES on a turnkey basis for its Virginia datacenters, but this market will likely develop beyond a single-supplier approach.
Finally, there is definitely something to be said for gains in efficiency. At the end of the day, less energy consumed is less carbon emitted. As an industry, it seems we've stalled out of late in continuing to push down PUE numbers for facilities and it is time for some new thought here. Increasingly, though – finally – we're seeing more interest in bringing new products to market that could usher in the next wave of efficiency gains.