Reducing emissions is important, but net-zero commitments are often unrealistic. If those ambitions are scaled back, what sustainability goals will remain?

Large operators, including Google, Meta, AWS, Iron Mountain and Microsoft, claim to match the energy they use each year with energy attribute certificates (EACs). Many more operators have promised to achieve 100% carbon-free or renewable energy consumption through EAC matching by 2030.

Google and Microsoft have made a more ambitious commitment (in 2020 and 2021, respectively): by 2030, their data centers would consume carbon-free energy (CFE) hour by hour. Microsoft's "100/100/0" pledge promises that 100% of its energy consumption, 100% of the time, would be zero carbon (see Made to measure: sustainability commitment progress and updates). In 2021, Iron Mountain also pledged to achieve 24x7 CFE consumption, but by 2040.

This is a worthy goal, which aligns electricity consumption with production from carbon-free (or zero-carbon) sources. Increasing the share of CFE generation assets in the electricity grid mix is essential to the grid's decarbonization. However, it requires significant investment in new wind and solar generation assets, expanded inter- and intra-region transmission infrastructure, and energy storage, to time-shift intermittent generation from periods of high to low output. New firm zero-carbon sources, such as geothermal and nuclear (including small modular reactors), will also be utilized as those technologies mature.

Microsoft and Google's commitments were designed to shoulder responsibility for their part in this change, making the job easier for the grid providers. Operators can increase grid CFE penetration by purchasing CFE through a green tariff or by committing to long-term offtake contracts to enable the direct installation of new generation capacity.

CFE availability

The average percentage of CFE generation on regional grids remains low, and approaching 100% CFE involves several major challenges:

• Intermittency and fluctuation. Wind and solar availability fluctuate daily and seasonally and, for extended periods of the year, solar (or wind) will not be available, so energy storage is required. Studies have shown that a combination of wind and solar generation assets integrated with batteries to time-shift generation can supply 80-90% of a facility's energy needs economically. Supplying the final 10-20% at economic electricity rates will require the development of long-duration energy storage (LDES) capable of providing electricity for 10-200 hours, firm geothermal and nuclear assets, and extended transmission infrastructure to enable demand/capacity balancing between grid regions.
• Stability. Spinning turbines at fossil, hydro and nuclear plants provide rotational inertia that stabilizes the grid voltage and frequency. Solar and wind plants typically connect through non-synchronous inverters, requiring grid operators to maintain some synchronous generation online to stabilize the grid. Alternatively, operators can deploy equipment such as batteries with fast-response synchronous inverters or synthetic/virtual inertia devices that mimic generator behavior.
• Dispatchable power. Coal and gas plants can be ramped up quickly, but wind and solar sources cannot be manually controlled by a grid operator. Their output depends on available solar radiation and wind speed, which vary continuously over a 24-hour period. Batteries, when properly integrated into the grid infrastructure, can provide the "dispatchable" supply required to address the intermittency of wind and solar generation for a limited period (typically 4 hours). Managing 6-12 hours or multi-day periods of low output and seasonal variations can require much greater LDES capacity.
• Permitting and planning. CFE sources and batteries occupy large sites, and may contribute to noise and viewscape pollution. To the extent possible, these assets should be located near the consuming data center to maximize the use of existing transmission infrastructure and reduce the need for high voltage transmission systems upgrades. These projects are often controversial and can be delayed by the need to procure multiple local, state and national permits and to address local opposition.

These issues vary geographically. In some areas, for instance, renewable sources are limited or difficult to develop, while in other locations regulations may restrict or encourage their development. CFE may command a price premium over the prevailing grid mix, where the percentage of CFE generation in the grid mix is low. In practice, the gigawatt-hours of available CFE and the associated premium will vary hugely by location, the mix of generation and storage assets in the grid region, as well as the date, time of day and season.

Renewable sources have a capacity factor of 20-60% (or more for some large hydropower installations), so they will deliver less power and energy on average than their rated maximum capacity. For example, a wind turbine with a rated capacity of 10 MW and a capacity factor of 35% will, on average, generate 31,000 MWh of electricity — 8,760 hours x 0.35 (capacity factor) x 10 MW — in a year.

Figure 1 shows that only two geographic regions, Latin America and Europe, have achieved greater than 50% CFE generation on their grids.

Figure 1 Average grid CFE generation by region

Some data center operators have been rapidly increasing their CFE consumption, but their total demand has grown even faster (accelerated by the current rapid expansion of AI facilities); as a result, the percentage of CFE in their data center energy consumption is stagnant. Google, for instance, grew its 24x7 CFE consumption rapidly from 6.7 terawatt-hours (TWh) in 2018 to 20.3 TWh in 2024, while its total energy consumption increased from 10 TWh to around 30 TWh over the same period. Its 24x7 CFE energy percentage grew from 61% to 67% in 2020, fell back to 64% in 2022, and has climbed back to 66% in 2024.

Google and Microsoft have both acknowledged the difficulties in meeting the "100% CFE consumption by 2030" target:

• Microsoft is reported to be reconsidering its 24x7 consumption goal. A spokesperson told Bloomberg that its yearly matching goal is also in question.
• Google states that it is still committed to its 24x7 CFE consumption goal, but acknowledges it was "always a moonshot."

Hourly CFE consumption goals

Hourly CFE consumption commitments were very likely made as aspirational marketing statements without accounting for the hard realities of the availability and performance of CFE generation assets. Operators and proponents of 24x7 CFE ignore (or underestimate) the challenges and full cost of attaining this goal based on hourly consumption, which include:

• Procuring intermittent wind and solar generation while ensuring a firm energy supply is delivered to the electricity meter.
• The lack of comprehensive hourly tracking of CFE generation, particularly the CFE percentage in battery storage (quantity of CFE stored and associated emission factors change between charging and discharging due to round-trip efficiency losses).
• A lack of hourly facility consumption data.
• New firm zero-carbon sources such as small modular nuclear reactors and geothermal sources are in the development phase and are not expected to make significant contributions to energy grids (increased 24x7 availability of CFE power) before the mid-2030s.
• The ongoing acceleration in data center and generation asset construction makes tracking and documenting these goals even more challenging.

Despite these issues, 24x7 CFE consumption goals remain a priority in public policy discussions. Hourly matching of CFE generation to facility consumption has been proposed in the draft revisions to the Greenhouse Gas (GHG) Protocol Scope 2 accounting guidance, the most widely used standard for emissions accounting. It is unclear as of June 2026 if the hourly matching requirement will be retained in the final standard, given strong opposition from many large energy consumers. The GHG Protocol is managed by the World Resources Institute (WRI) and is referenced in EU legislation, such as the Corporate Sustainability Reporting Directive, and in voluntary programs, such as the Science Based Targets initiative (SBTi).

Uptime Institute has commented to the GHG Protocol Scope 2 Technical Working Group that hourly matching is not currently viable in market-based carbon accounting (Scope 2 Guidance update: impact on climate disclosure). It would impose a complex accounting process (and expensive tracking software) on inadequate data and negate the value of the many virtual power purchase agreements used to secure EACs.

The SBTi has declined to align with the draft GHG Scope 2 Guidance at this time, and industry majors are expected to develop an alternative Scope 2 accounting standard if the GHG Protocol finalizes its current draft.

Achievable targets

Uptime Institute recommends monthly matching because it is achievable with current data and represents a substantial improvement over yearly matching. Declarations of CFE consumption and net-zero emissions goals should use location-based, not market-based, accounting and operators should progress incrementally over time toward these goals.

Operators can reach 50-80% CFE consumption over a year with a price premium below 5% in markets with high CFE penetration. Beyond 80%, the cost escalates, primarily because there is insufficient LDES to time-shift CFE and provide firm, dispatchable CFE generation.

The cost of 24x7 matching is shown in Figure 2, which uses data from grids with average and optimal levelized cost of energy (LCOE) to model how costs rise with the percentage of consumption matched by clean supply. Achieving 80-90% hourly-matched CFE would represent a significant improvement on most grids, which are currently operating at 27-71% CFE generation (see Table 1), and would help reduce the grid emissions factor (24x7 carbon-free energy (part one): expectations and realities; 24x7 carbon-free energy (part two): getting to 100%).

Figure 2 The cost of electricity as the percentage of 24×7 CFE approaches 100%

Uptime Institute Intelligence reports have highlighted the issues (Unravelling net-zero), predicted net-zero goals would be eased (Five data center predictions for 2024), and noted signs that this shift is already occurring in practice (Net-zero timelines are becoming more realistic).