• Should Google and Microsoft focus on sourcing their own 100% clean power or cleaning up the dirtiest grids?
  • Newsletter
  • Donate
Clean energy journalism for a cooler tomorrow

Should Google and Microsoft focus on sourcing their own 100% clean power or cleaning up the dirtiest grids?

Major companies are spending billions on clean energy. New measurement tools can help them get the most climate impact from their investments.
By Jeff St. John

  • Link copied to clipboard
Google and Microsoft logos
(Halil Sagirkaya/Anadolu Agency)

This is the second of two articles on how corporate clean energy buyers are using new strategies to make the most climate impact with their investments. Read the first article.

Major companies with ambitious clean energy goals face a complicated set of options for how they ought to prioritize their efforts over the coming decade. Should they make their own electricity supply as clean as possible, or should they focus first on cleaning up the dirtiest power grids?

The first approach means striving to get every watt of electricity they use from a carbon-free source. That’s the idea behind Google’s 24/7 clean energy pledge, made a year ago, which sets a 2030 deadline for powering its data centers and corporate campuses with 100 percent carbon-free energy every hour of the year. Microsoft followed up earlier this year with a 100/100/0 pledge to match 100 percent of its corporate power consumption with zero-carbon resources 100 percent of the time by decade’s end.

But a second pathway for maximizing corporate carbon reductions has been gaining traction in recent years: investing in clean energy projects based on their emissionality,” or their ability to directly reduce carbon emissions. Instead of linking clean power with companies’ individual electricity loads, companies would target their dollars at projects with the most capacity to displace dirty energy, regardless of whether those projects are near their facilities or far away. These types of calculations are being made possible through a complex but increasingly data-driven set of methodologies.

So which approach is better? The answer depends on a wide range of variables, said Olivier Corradi, CEO of Tomorrow, a Danish company measuring the carbon-intensity of power grids around the world. But in simple terms, there are two optimization goals,” he said. One is, can I reduce my own footprint? The other is, can I make investments that can avoid emissions somewhere else?”

Corporate procurements of clean power drove about one-fifth of total U.S. clean energy additions in 2020 and are expected to account for 55 to 85 gigawatts of deployment through 2030. That means companies’ decisions about how they purchase clean energy are a big deal. And the rules that guide how they’re able to claim credit for emissions reductions created by their investments are vitally important.

There are pros and cons to both the 24/7 and emissionality approaches. The simplicity and directness of tying clean energy to consumption leave little wiggle room for problematic approaches like buying renewable energy credits for wind or solar that’s already been built. It’s also a way for companies to help curb emissions in their backyards — not just carbon emissions but also other types of air pollution — thereby allowing them to be better neighbors and community members.

But there are practical limits to how far any one company can go on its own to achieve such round-the-clock clean energy targets. Even the wealthiest companies can’t force entire power grids to adopt clean energy. And the costs of achieving 24/7 clean energy within their walls may direct money away from building more clean power on the world’s dirtiest grids where it’s most urgently needed.

That’s a key insight from this summer’s Clean Power by the Hour report from nonprofit research organization RMI, commissioned by Microsoft to help the software giant better understand the implications of its 100/100/0 policy. (Canary Media is an independent affiliate of RMI.) The report’s fundamental finding, said Mark Dyson, a principal with RMI’s Carbon-Free Electricity Practice, is this: Don’t let the perfect be the enemy of the good.”

In this case, the perfect” would be the 100 percent carbon-free power ideal, whereas the good is just building a bunch of wind and solar on the dirtiest grids, and pushing fossil fuels off the system,” he said.

Consider the chart below, which shows the load-matching” costs for a hypothetical data center served by mid-Atlantic grid operator PJM. Building the batteries to store enough carbon-free power to cover the final 40 percent of the data center’s annual energy demand would drive up costs dramatically. 


That money could be better spent building wind power to serve the PJM grid, even if that leaves the data center shy of its round-the-clock clean energy goals, as the next chart shows.


In short, these findings indicate that 24/7 goals like Microsoft’s and Google’s, if taken to extremes, aren’t the most cost-effective way to drive the most aggressive emissions reductions possible. Both companies are acknowledging this fact in their broader corporate decarbonization strategies.

Helping the rest of the world get to 24/7

While Dyson concedes that 24/7 goals are not the most economical way to get clean energy onto the grid today, he adds one important caveat to this observation. We need to build the infrastructure and policy to enable a grid that’s clean 24 hours a day. One of the outcomes of reaching a 24/7 target is that you can’t do it cost-effectively on solar and wind and batteries alone,” he said. This is a huge challenge for mandates around the world requiring 100 percent clean electricity, as multiple studies over the years have pointed out.

So when major companies push to achieve 24/7 goals, they are helping to develop the technologies and markets that will help everyone else get to 24/7 too. 

Today’s battery technologies can cost-effectively manage the variability of wind and solar power from hour to hour. But longer-duration forms of energy storage and more reliable and controllable forms of zero-carbon electricity generation will be needed to manage the mismatches that can arise across 24-hour day-to-night cycles and between winter and summer, as indicated by this chart from a May report by the U.S. National Renewable Energy Laboratory. 

(National Renewable Energy Laboratory)

Corporations that lead the way on 24/7 targets can help solve the long-duration storage problem as well as additional challenges that face utilities, cities, states and other companies with 100% clean energy goals.

The Google approach to the 24/7 challenge

That’s how Michael Terrell, Google’s director of energy, sees the value of the 24/7 clean energy approach. It’s a signaling method to tell the market where we need to go,” he said in a July interview. The ultimate destination is carbon-free power, 24/7, in every location everywhere around the world.”

Google is investing in a range of technologies that could help it get more clean power in the hours when wind and solar can’t easily deliver it, including geothermal power. It’s also trying to shift its data centers’ computing loads to periods when grid power is cleanest.

Given its global footprint, Google has little choice but to work with each grid as it is, Terrell noted. We have five data center sites that are over 90 percent carbon-free,” including one in Oregon where there’s ample hydropower and one in Iowa where Google has invested heavily in wind power. In other markets with dirtier power, such as the U.S. Southeast or Taiwan, the company is pushing for policies such as green tariff programs to expand the opportunities to bring more clean power onto the grid.

Google’s strategy is laid out in more detail in the company’s white paper on carbon-free energy methodologies and metrics, which underpin its assessment of the avoided-emissions” impact of its procurement decisions.

We wanted to align corporate purchasing with where we want to take the grids,” Terrell said, not just sourcing every hour, everywhere, for our own supply, but to do that for the grids we’re working with.”

We’re not doing this on an island,” he said. We’re part of the system, and we recognize that to achieve our goal at every site, we need to drive system change at every site.” 

How Microsoft is balancing near-term and long-term impacts

Microsoft’s carbon plan also tries to balance the pros and cons of optimizing its own 24/7 clean energy goals with driving faster decarbonization on the grid, said Brian Janous, the company’s general manager of energy and renewables, in an August interview.

One of the things to get clear right out of the gate is to explain what it is we’re not trying to do,” he said. Everyone can put solar on their roof and batteries in their garage, and sure, they can get 100 percent clean energy to cover their needs.” But to do that, you’d vastly overbuild the amount of solar and storage you’d need.”

RMI’s Clean Power by the Hour study underscores the diminishing returns of this isolated approach to sourcing 100 percent clean power 100 percent of the time, Janous said. But it also highlights the importance of putting money behind decarbonization approaches that are less cost-effective than wind and solar today but will be needed to reach gridwide 100 percent goals.

It goes far beyond the power that’s coming into our meter,” he said. It’s how we’re driving market behavior, how we’re pushing the boundaries on longer-duration storage — all the things we need by 2030 to some degree, and definitely by 2040.”

At the same time, Microsoft is directing its clean energy investments in ways that target the biggest emissions reductions, Janous said. One example is its work on the locational marginal emissions” tool developed with partner REsurety to site renewables projects where the clean power they produce will displace the most carbon. (Read more about Microsoft’s work with REsurety in our first article in this series.)

An analysis of the Texas grid by REsurety and The Brattle Group, for instance, shows how building a wind or solar farm on the wrong side of a transmission congestion point can nearly halve its effective carbon-reduction impact over the course of a typical year. That’s because much of the power it produces will never reach customers, ultimately going to waste. This chart from the report shows the difference between two hypothetical solar projects in Texas in terms of their locational marginal emissions (LME), a term used to describe this relationship between project location and emissions-reduction impact. 

(REsurety and Brattle Group)

We want to make sure that our solar project is rapidly decarbonizing the grid [and] that I’m not sticking it in West Texas” behind a congestion point, he said.

Not all companies have the same power to influence the energy markets they’re part of, of course. That means that each company’s clean energy strategies will differ based on where they are on their journey and what they’re trying to solve for,” said Priya Barua, director of zero-carbon innovation for the Renewable Energy Buyers Alliance, a trade group representing some of the biggest U.S. companies with ambitious decarbonization commitments.

Is it trying to drive emissions reductions for the long term?” she said. Is it using their buying power to turbo-charge a broader suite of technologies that could drive deeper decarbonization?”

I don’t feel like there’s a wrong or a right strategy,” Barua said. It boils down to how…they want to have an impact as an organization.” 

It’s complicated: Determining cause and effect

There’s a key challenge for companies seeking to add clean power to dirtier grids by prioritizing emissions reduction impact over 24/7 direct clean energy purchases, however: There’s no commonly accepted way for the companies to take credit for the resulting emissions reductions. For companies under scrutiny to prove the effectiveness of their investments, that’s an important issue.

The following chart from the REsurety and Brattle Group Texas grid analysis shows it’s much cheaper for a company to invest in renewables where they can displace the most carbon on an LME basis than for a company to match on-site loads to 24/7 carbon-free energy supply.

But a company could achieve an even more dramatic reduction — and at a lower cost — by moving its data centers and other energy-consuming facilities away from dirtier grids and onto cleaner ones. In fact, this technique could even lead to net negative emissions since it removes demand centers that would otherwise be served by dirtier power.

(REsurety and The Brattle Group)

There is, however, no common set of agreed-upon methods for measuring these options on an apples-to-apples basis. Moving a data center could clean up its owner’s electricity portfolio, for example. But did that act lead to less coal being burned on the grid it departed or more clean energy being built on the grid it chose to connect to? It’s hard to establish a clear cause-and-effect chain in this instance — so it would be hard for the company to prove the claim that its investment should earn carbon-reduction credits.

Trying to ascribe emissions impacts to a corporation’s clean energy siting decisions brings up similar complications. One of the biggest involves the difference between attributional” and consequential” measures of marginal emissions, Tomorrow CEO Corradi said.

These terms are described in detail in a white paper written by Corradi and Gavin McCormick, CEO of WattTime, another company measuring power grid emissions on an hourly basis. (WattTime, like Canary Media, is an independent affiliate of RMI.) While they’re a bit tricky to distinguish from one another, they’re also extremely important to understanding the way that companies can think about the impact of their actions and how society can measure them, Corradi said.

The key difference, he said, is between a company claiming that it’s using clean energy (i.e., attributing” it to its overall energy portfolio) and claiming that the grid at large getting cleaner was the consequence” of a particular decision by the company.

Today, under the Greenhouse Gas Protocol Scope 2 Guidance, corporations report the emissions impact of their energy purchases through what’s called an attributional” accounting method. Attributional means you’re trying to take global emissions and divide them up by all the consumers, and [then] attribute each megawatt-hour produced, and the associated emissions, with someone responsible for causing it,” Corradi said. That’s an important way to differentiate a company’s real-world electricity usage from unbundled renewable energy credits, virtual power-purchase agreements and other methods that frame these fundamental relationships in an abstract manner, he said.

Consequential” emissions changes, by contrast, must be tied directly to the consequence of your actions” as an energy buyer, Corradi explained. In other words, a company might be able to attribute the clean grid power its data center is using to its corporate share of clean energy use, but it couldn’t claim to be responsible for causing that grid to be cleaner by using the clean energy without having a commonly accepted and verifiable way to prove it, he said.

On the other hand, if a company could prove that its actions led directly to less dirty energy being produced on a grid — say, by reducing total grid demand that would otherwise have caused a natural gas peaker plant to be ramped up to serve it — that could be considered a consequential reduction, he said. But measuring the consequences of an action requires comparing it to a counterfactual,” he said. You’re asking what would have happened if I didn’t” take the action.

Any process that involves measuring real-world decisions against counterfactuals — against what might have been, in other words — requires assumptions about what would have happened if the action hadn’t been taken. And if you build models with a lot of assumptions, those assumptions will be challenged,” Corradi said.

The difference between short-term and long-term emissions effects

To add another level of complication, corporate clean energy decisions can have very different effects on short-term carbon emissions versus long-term emissions — and the long-term ones, while they may be more important, are very tricky to calculate.

It’s much easier to make solid, mutually agreed-upon assumptions about what would have happened on the grid as it’s operating today, Corradi said. Take the example of Google shifting its data-center loads on an hour-to-hour basis to periods when more solar and wind power is available and reducing them when coal or gas plants are ramping up.

But it’s a lot harder to say that building a wind farm or a data center in one place versus another will achieve a quantifiable reduction in emissions over longer periods of time, he said. After all, the emissions-intensity of the grid changes with every new unit of clean energy added to it or unit of dirty energy removed.

Google’s white paper on its methodologies makes the same point, noting that its accounting for marginal emissions applies to short-term, small variations in load,” rather than longer-term decisions like where to build a solar or wind farm. “[W]e are looking at clean energy projects that will be operating for decades, and therefore to measure their impact we would need to understand their long-term impact on grid planning and dispatch patterns.”

Companies such as Tomorrow and WattTime that measure marginal emissions reductions differentiate between short-term, or operational,” reductions, like those that come from charging an electric vehicle at a different time of the day, and long-term, or build margin,” reductions, like the decision to construct a wind farm or a data center in one place rather than another.

It is our understanding that there are not yet any standardized, universal sources of reliable build margin data available,” Corradi and his co-authors state in the white paper. That makes it difficult to apply in practice in 2021, though we hope that will change over time.”

Pieter Gagnon, senior energy systems researcher with the National Renewable Energy Laboratory, noted that NREL’s work on long-run marginal emissions rates” could start to shed light on this forward-looking measurement challenge.

We’ve had some good conversations with the Google folks” about this metric, to evaluate the avoided emissions from whatever interventions they’re considering,” he said. But more work needs to be done on these methodologies before they might be used to certify emissions-reduction impact in a transparent and commonly accepted way.

I would love it if there was really good guidance” on the emissions impact of investments in different locations over the long term, Gagnon said. Without them, companies may feel compelled to take actions they can clearly tie to their own electricity consumption, like focusing on 24/7 clean energy, whether or not they’re the best for reducing emissions at large.

In other words, you may choose to do something that doesn’t decrease emissions as much, but it decreases your emissions,” he said. The trade-off there is kind of embedded in those two approaches.”

Jeff St. John is director of news and special projects at Canary Media. He covers innovative grid technologies, rooftop solar and batteries, clean hydrogen, EV charging and more.