Storage at data centres is announced, not counted: 52 GW of US batteries operate, but no source says how much sits behind a data-centre meter

What a planner can actually count

Anyone planning around storage at data centres needs one number: how much battery capacity sits behind a data-centre meter and is running. None of the sources opened for this brief publishes it. They give a national total, a count of data-centre self-generation that is almost entirely gas, and a short list of individual sites. The figure in between is missing, and the sites that are named differ in status and in the basis of their numbers.

The measured number is national. The data-centre number is a list of projects.

The national count is measured, and it is not about data centres

The US Energy Information Administration reports that operators had added 8.3 GW of battery storage in the first half of 2026, taking nameplate capacity to nearly 52 GW, up from 43.6 GW at the end of 2025¹. The source is the EIA-860M generator inventory, which is operator-reported, and the figures are nameplate: what the units are rated for, not what they delivered. EIA also lists 54 GW more planned over the next two and a half years, split 14 GW for the second half of 2026, 26 GW for 2027 and 14 GW for 2028. Plans are not operating capacity.

The EIA article does not mention data centres. It gives the baseline against which any data-centre figure should be read, and nothing more.

The behind-the-meter tracker counts gas, not batteries

Cleanview, a data company that sells tracker access and full reports, follows 59 data centres planning their own power, about 90 GW in total². Its method is permits, filings, press releases and satellite imagery (the imagery shown was captured on 13 May 2026). On its count only about 2 GW is operating, 2.2% of the total, with 1.2% under construction, 36% permitted and 60% existing only as announcements or early-stage plans. Colossus 1 and 2 near Memphis account for 1,498 MW of the operating gas turbine capacity.

These are generation figures. The Cleanview summary gives no battery gigawatts or gigawatt-hours, and the full report, including its Colossus section, is for customers only and was not read. The 2 GW is a Cleanview estimate, independent of the operators but not an audited figure.

What the named sites show

The xAI count is the clearest operating example, and it is still a count of units with a manufacturer's rating attached. The 655 MWh is this brief's arithmetic, not a published total. The report does not say whether the batteries are kept for the life of the site.

The Sweden system is the oldest. Saft says it replaces diesel backup and supports grid stability and black-start capability, but the article gives no total capacity and no measured use or revenue. If each of the four groups holds 4 MWh the site holds 16 MWh; that is arithmetic, not a stated figure.

Cheyenne is the largest figure and the least built. Its battery is contracted under a long-term tolling agreement with the local utility, which is a commercial commitment, not energised capacity. At 200 MW and 1,600 MWh it is an eight-hour system on paper, and nothing yet operates.

Aligned's 31 MW is described as an explicit strategy to accelerate grid interconnection. It shows storage being bought to buy time with the grid, which is a different purpose from backup or energy shifting.

What this does and does not establish

It establishes that the opened sources give a measured national total, an estimated 2 GW of operating data-centre self-generation that is gas, and no measured total or measured use for data-centre storage. Every site figure above is a rating, an agreement or an announcement, not a metered result.

It does not establish that little data-centre storage exists. It may well be substantial. The evidence base is thin: one government source (EIA), one commercial tracker, and trade-press and vendor reports for the sites. Nothing independent measures how the batteries are used, and the Saft, xAI and Enbridge figures all come from the parties that benefit from them.

What a planner should take from it

Ask three questions of any data-centre storage number before it enters a plan.

1. Status. Is the capacity installed, energised and dispatching, or contracted or announced? Cheyenne and Aligned are the latter.
2. Basis. Is the figure power (MW), energy (MWh) or both, and is it a nameplate rating or something metered? xAI has a rating per unit but no MW; Cheyenne has both but nothing built.
3. Use. Is the battery for backup, for ride-through, for interconnection timing or for selling grid services, and is there a dispatch record? None of the sources publishes one.

The practical consequence is narrow. Do not scale the national 52 GW, or Cleanview's 2 GW of gas, into a figure for data-centre storage. Where a decision depends on storage at a particular site, such as a connection bargain like Aligned's or a dependency on a neighbouring project, ask the operator for energised capacity in MW and MWh and for dispatch records. Until those exist, treat data-centre storage as announced, not counted.