X Uncovers Suspected Chinese Network Trying to Shape Narrative on AI Data Centers

On Aug. 28, X’s Global Government Affairs account said its Safety team had identified a suspected Chinese inauthentic “bot farm” of approximately 200,000 accounts.

Within the farm, X said, 200 accounts were posting about the U.S. AI data centers, household electricity prices and strain on the power grid. The 200 accounts represented about 0.1% of the larger network, and X did not describe what the remaining accounts were posting.

The posts included AI-generated cartoons depicting data-center operators profiting while the public absorbed the costs. X said the activity was being conducted “in a manner that could manipulate a legitimate debate” over American AI and energy policy.

X said it suspends accounts that violate its Authenticity policy, but did not publish a tally showing that all 200,000 accounts had been suspended. It also did not identify the operators, disclose how long the network had been active or provide engagement figures for the 200 accounts.

The Data-Center Power Debate Predates the Network

The accounts were entering a U.S. debate that was already underway.

Data-center construction is driving a sharp increase in electricity demand as companies expand infrastructure for artificial intelligence. PJM’s July 2026 capacity auction for 2028-29 fell about 6,831 megawatts short of its reliability requirement, while the grid operator’s independent market monitor, Monitoring Analytics, has said data-center load growth is the primary reason for recent and expected capacity-market conditions.

The resulting capacity-market costs do not translate directly into a uniform increase in household electricity bills. Costs move through wholesale markets and utility rate structures differently across regions, and some of the data-center demand driving forecasts has yet to be built.

The Accounts Used AI-Generated Cartoons

X said some of the 200 accounts used AI-generated cartoons to make the same argument in visual form: data-center operators were benefiting while ordinary consumers were left with higher costs.

The material gave an existing dispute a coordinated set of voices and images. X did not publish the accounts’ individual posts, posting volume or audience data in its disclosure, leaving the scale of their exposure unclear.

The company described the activity as something that could manipulate the policy debate. It did not say the accounts had successfully changed public opinion or established that their content had altered the broader conversation around data centers.

OpenAI Found a Similar Campaign in June

The finding follows a June investigation by OpenAI into a separate China-linked campaign it called “Data Center Bandwagon.”

OpenAI said operators used ChatGPT to generate English-language posts and comic strips portraying data centers as a burden on American households, then posted the material on X through accounts posing as Americans. The operators used Simplified Chinese prompts and VPNs, OpenAI said, and were likely part of a private Chinese technology company providing social-media services to provincial-level government clients.

OpenAI said the campaign generated little authentic engagement. Its investigation did not establish that the operation changed public opinion, and X has not said whether the network it identified in August is connected to it.

The two disclosures describe similar subject matter and tactics: China-linked activity, X accounts presenting themselves as ordinary participants in the debate, and AI-generated material focused on the costs of data-center expansion. They do not establish a common operator or infrastructure.

X’s disclosure provides evidence of a suspected inauthentic network attempting to push claims into an existing debate over AI infrastructure. It does not identify who ran the network, show how far its content traveled or establish that the effort changed public perception.