X Removes 200,000 China-Linked Accounts It Says Pushed Anti-Data-Center Messaging

X says it has dismantled a network of roughly 200,000 accounts it links to China, and that around 200 of them were quietly seeding American social media with claims that AI data centers are driving up electricity bills and straining local power grids. The takedown, announced by X's Global Government Affairs account on August 28, is the second such disclosure in three months tied to the increasingly contentious politics of US AI infrastructure — but an independent researcher who dug through the archived posts says the actual reach of the campaign may have been close to zero.

In its statement, X said its Safety team "conducted an investigation into suspected Chinese inauthentic accounts involved in influence operations" and "identified a bot farm of approximately 200,000 accounts." Within that larger farm, the company said, it found roughly 200 accounts "posting in a manner that could manipulate a legitimate debate about American AI and energy policy." X described the narratives pushed by those accounts as including claims that data centers are raising household electricity prices, that rapid construction is overloading the grid, and that ordinary ratepayers are subsidizing wealthy AI companies — some of it delivered through AI-generated cartoons depicting data center operators growing rich while families pick up the tab. "We take seriously any attempts to undermine the integrity of the global town square and suspend accounts that violate our Authenticity policy," the company said, adding that it remains committed to keeping X "an open and authentic platform."

Notably, X itself acknowledged that the underlying grievances are not fabricated. The company's own statement conceded that "many of the underlying concerns about electricity demand and consumer bills are indeed supported by genuine data" — meaning the alleged operation, as X frames it, was less about inventing a controversy than amplifying and steering one that already exists among American residents near proposed or operating data center sites. X did not disclose engagement or follower figures for the flagged accounts, nor did it detail what the remaining roughly 199,800 accounts in the farm were doing, or provide technical evidence tying the network to Chinese state or corporate actors beyond calling them "suspected Chinese inauthentic accounts."

That gap between claim and verification is where the story gets complicated. Darren Linvill, a disinformation researcher at Clemson University, dug into the archived posts and told Business Insider that the campaign's real-world footprint looks negligible. Linvill traced activity using the hashtag "capacityauction," which he found attached to exactly 90 posts, all published in January, with individual accounts typically posting only two or three times before X suspended them. Examining at least 59 of the associated accounts, Linvill said many had no followers and generated no engagement — replies, likes, or reposts — at all. "No real humans saw these posts about data centers," Linvill told Business Insider, going on to describe the operation as effectively a "non-story" in terms of its influence on public opinion, even while allowing that it's still possible some users encountered the content before accounts were removed.

The episode follows a similar disclosure by OpenAI in June, when the company published a threat report describing a cluster of accounts it called "Data Center Bandwagon." OpenAI said the operators used ChatGPT, accessed through VPNs from mainland China with prompts written in Simplified Chinese, to generate comic strips and social posts arguing that AI data centers were pushing up electricity prices — content that in some cases drew on regional US newspaper reporting about grid capacity auctions before being repackaged and posted under hashtags including #capacityauction. OpenAI assessed that the operators were likely a social media team working for a private Chinese tech firm serving "provincial-level government clients," and it rated the campaign's measurable impact as showing activity but no evidence of meaningful engagement from real audiences. "Industry, governments, civil society and the public should remain alert for similar attempts to scale these messages and foreign interference activities," OpenAI wrote in that report. The overlapping hashtag and narrative between the OpenAI and X findings suggests researchers may be looking at the same underlying operation, or at least closely related ones, moving across platforms.

Why It Matters

AI data centers have become one of the most visible flashpoints in American politics this year, with local opposition spreading across both red and blue districts over rising electricity costs, water use, noise, and land value concerns — genuine, homegrown grievances that predate any foreign involvement. That backdrop is exactly what makes attribution so fraught. When a platform or an AI company says it has found foreign accounts amplifying a domestic debate, it invites two opposite failure modes: dismissing a real and growing grassroots movement as a foreign psyop, which can delegitimize local residents' actual concerns about their power bills and their grids, or underestimating how cheaply and quickly state-linked or commercially motivated operators can inject synthetic content into a debate that's already primed to spread it. X's own acknowledgment that the underlying concerns are "supported by genuine data" is a rare and useful admission that complicates any narrative of a purely manufactured controversy. Attribution of influence operations is also technically hard: platforms rarely publish the forensic evidence — infrastructure overlap, behavioral signatures, device fingerprints — that would let outside researchers independently verify a "China-linked" label, leaving the public to weigh a company's word against a researcher's spot-check of engagement metrics. As AI infrastructure spending accelerates into the hundreds of billions of dollars, expect information operations, real or alleged, to become a recurring subplot in the broader fight over where data centers get built and who pays for the power they use.

What to Watch

Watch for whether X or independent researchers release more granular data — engagement counts, network graphs, or technical indicators — that would let outsiders verify the China link and gauge actual reach, and whether Congress or federal agencies (Republican lawmakers have already floated FBI scrutiny of data center opposition) treat this as grounds for policy action. Also worth tracking: whether OpenAI, Meta, or other platforms disclose related takedowns tied to the same hashtags or messaging clusters, which would help establish whether this is one small operation or a broader, coordinated effort to shape the AI infrastructure debate.