The growing backlash against AI data centers in the United States may have attracted more than local activists and angry electricity customers.
X says its safety team uncovered a huge network of approximately 200,000 suspected inauthentic accounts linked to China, including roughly 200 accounts specifically posting content about American AI data centers and energy policy.
According to X, those accounts pushed messages claiming data centers were increasing household electricity bills, putting pressure on power grids, and allowing wealthy technology companies to profit while ordinary Americans paid the cost.
Some accounts used American-sounding names and shared legitimate news stories alongside inflammatory commentary and AI-generated cartoons.
The discovery is important because it suggests foreign influence operators may be trying to exploit one of the most politically sensitive issues surrounding the AI boom.
But there is an equally important warning:
Real foreign influence activity does not mean real American concerns about data centers are fake.
Electricity prices, grid congestion, water consumption, noise, land use, and local infrastructure pressures are genuine issues in many communities.
That distinction sits at the center of this story.
X Says It Found About 200,000 Suspicious Accounts
X’s Global Government Affairs team announced the investigation on August 28.
The company said its Safety team had been examining suspected Chinese inauthentic accounts involved in influence operations.
That investigation identified what X described as a bot farm containing approximately 200,000 accounts.
Most of those accounts were not publicly described in detail.
Within the larger network, however, X says it found about 200 accounts posting specifically about AI data centers and American energy policy.
X said the accounts were publishing material in a way that “could manipulate a legitimate debate about American AI and energy policy.”
That wording is worth noticing.
X itself described the underlying debate as legitimate.
The company was not claiming that concerns about data centers were invented by China.
It was saying that suspected foreign operators were trying to insert themselves into an existing American argument and amplify particular narratives.
What Were the Accounts Posting?
The themes were familiar to anyone following the data center debate.
Posts claimed AI data centers were:
Driving up electricity prices.
Straining local power grids.
Creating costs that ordinary households were forced to absorb.
Making large technology companies richer while communities carried the burden.
Some posts linked to genuine reporting about electricity markets and data center demand, including stories about rising capacity costs.
The suspected accounts would then add more emotional commentary around those stories.
Others used AI-generated political cartoons showing wealthy data center operators benefiting while ordinary consumers faced higher bills.
This is a common feature of influence operations.
The underlying fact does not necessarily have to be false.
A more effective strategy can be taking a real issue, exaggerating it, presenting it emotionally, and disguising the identity of the person pushing the message.
That makes propaganda much harder to identify than a completely fabricated story.
Some Accounts Pretended to Look American
One detail that attracted attention was the apparent identity of the accounts.
Examples displayed by X included profiles with names such as:
Thomas Jackson.
Patricia Robinson.
Sandra Thompson.
Kenneth Flores.
Those names made the accounts look like ordinary American users.
That appearance matters.
A post saying:
“People in my town are paying higher electricity bills because of these AI data centers”
feels very different if the reader believes it came from a local American resident rather than someone working overseas as part of a coordinated influence campaign.
The objective is not merely to spread information.
It is to manufacture the impression that a particular viewpoint is more popular, organic, or widespread than it really is.
This technique is often described as astroturfing.
It creates the appearance of grassroots political activity where some of the activity may actually be coordinated.
X Has Not Publicly Proved Who Ordered the Operation
There is an important limitation.
X says the accounts were associated with suspected Chinese influence activity.
But the company has not publicly identified a specific Chinese government agency as the operator of the entire 200,000-account network.
It has also not disclosed a complete list of the accounts or published a full technical report explaining every piece of attribution evidence.
Reporting following the announcement noted that it remains unclear who sponsored or sanctioned the network.
That means claims such as:
“China’s government ordered 200,000 bots to attack U.S. data centers”
would go beyond the evidence X has publicly presented.
The more accurate description is:
X says it identified a large network of suspected inauthentic Chinese accounts, some of which were attempting to influence U.S. AI and energy discussions.
That still matters.
It simply requires careful wording.
OpenAI Found Something Very Similar Earlier This Year
X’s findings are especially interesting because another AI company reported a closely related operation months earlier.
In June, OpenAI disclosed what it called the “Data Center Bandwagon” campaign.
OpenAI said it banned a cluster of ChatGPT accounts that likely originated in China and were being used to generate social-media posts criticizing American AI data center expansion.
The accounts asked ChatGPT to create English-language comments and images claiming data centers were increasing electricity prices for ordinary Americans.
OpenAI said those posts later appeared on X through likely inauthentic accounts pretending to represent Americans from different backgrounds.
That overlap is difficult to ignore.
The themes are nearly identical.
AI data centers.
Electricity prices.
Ordinary households paying the cost.
American-looking social-media profiles.
AI-generated graphics.
OpenAI’s earlier investigation therefore provides independent evidence that at least one China-linked influence effort really was targeting this particular policy debate.
Read OpenAI’s Data Center Bandwagon investigation
OpenAI Found Clues Pointing Toward China
OpenAI provided more detail about how it reached its attribution.
The suspicious ChatGPT accounts interacted with the system in Simplified Chinese.
They were active during working hours consistent with mainland China.
They used VPNs because OpenAI does not provide service directly in mainland China.
And some activity suggested the users may have been associated with a private Chinese technology company working for provincial-level government clients.
OpenAI also found that the accounts asked the model to produce material posing as Americans.
That is much stronger evidence than simply observing that several accounts shared similar political opinions.
Still, even OpenAI used careful wording.
It described the campaign as likely PRC-origin, rather than claiming it had proved direct control by China’s central government.
That cautious distinction should also be applied to X’s latest claims.
Did the Chinese Campaign Actually Change American Opinion?
There is little evidence that it did.
OpenAI explicitly said it found no evidence of meaningful breakout beyond the operation’s own activity.
That means the accounts existed.
The content was generated.
Posts were distributed.
But OpenAI did not find strong evidence that the campaign significantly shifted public opinion.
This is an important point because influence-operation stories can accidentally give relatively ineffective propaganda campaigns far more credit than they deserve.
A few hundred accounts can create noise.
That does not necessarily mean they convinced millions of voters.
Business Insider also questioned how much impact X’s newly identified accounts could realistically have had, even if the network was genuine.
Foreign influence operations may be most useful when they amplify existing divisions rather than creating entirely new ones.
And Americans already have plenty of reasons to argue about data centers without outside assistance.
The Data Center Backlash Is Very Real
Communities across the United States are increasingly resisting new data center projects.
The reasons vary.
Some residents worry about electricity prices.
Others worry about water consumption.
Some dislike noise from cooling equipment and generators.
Others object to farmland or rural land being converted into enormous industrial facilities.
There are also concerns that data centers consume huge quantities of infrastructure while creating relatively few permanent local jobs after construction finishes.
Those arguments are now influencing politics in both Republican and Democratic states.
Texas politicians, for example, have proposed stricter rules around data centers as public opposition grows.
The backlash is therefore not something that needs to be invented by bots.
Foreign operators merely need to amplify it.
Electricity Concerns Are Not Imaginary
One of the most dangerous mistakes would be assuming that because suspicious foreign accounts posted about electricity prices, electricity-price concerns must therefore be misinformation.
The real U.S. grid is under growing pressure.
PJM, the largest electricity market in the United States, has experienced sharply rising transmission congestion costs.
In the first half of 2026, those costs surged 43% to approximately $6 billion, according to Reuters.
Growing electricity demand from data centers is one factor contributing to the broader pressure on the grid, alongside extreme weather, electrification, transmission bottlenecks, and higher gas prices.
PJM’s capacity market is also struggling to keep up with growing demand.
Its 2028/29 capacity auction fell short by about 6.8 gigawatts, while prices reached the market’s maximum cap.
So when someone says:
“Data centers could put pressure on electricity infrastructure,”
that statement is not automatically propaganda.
The manipulation comes from pretending to be someone you are not, coordinating deceptive accounts, and selectively pushing narratives to inflame divisions.
This Is What Makes Modern Influence Operations Effective
The strongest propaganda usually contains some truth.
Suppose an operator creates a completely fake claim:
“AI data centers secretly control the weather.”
Most people will ignore it.
Now take a real issue:
Electricity prices have risen.
Data centers need large amounts of power.
Communities are worried.
Then exaggerate the relationship.
Blame every increase on AI.
Create emotional cartoons.
Pretend dozens of ordinary Americans are independently saying the same thing.
That is much more effective.
The goal is often not to persuade everyone of one specific claim.
It is to increase distrust.
Make people angrier.
Make political compromise harder.
Push existing groups further apart.
OpenAI described the earlier China-linked campaign in almost exactly these terms, saying foreign influence operators often attach themselves to sincere local concerns to amplify divisions and exacerbate public distrust.
Why Would China Care About American Data Centers?
The strategic motivation is straightforward.
Artificial intelligence has become one of the central areas of competition between the United States and China.
Advanced AI requires:
Semiconductors.
Data centers.
Electricity.
Networking.
Capital.
Large-scale computing infrastructure.
If American communities block data centers or make them more difficult and expensive to build, U.S. AI development could theoretically slow.
China is simultaneously investing heavily in its own AI infrastructure.
From a geopolitical perspective, anything that delays a rival’s computing expansion could create an advantage.
That does not prove the Chinese government ordered the X campaign.
But it explains why AI infrastructure would be an attractive topic for a foreign influence operation.
OpenAI itself said the earlier campaign was significant because AI infrastructure is a foundation of U.S. technological leadership and economic competitiveness.
The AI Industry Also Has Reasons to Emphasize Foreign Interference
There is another side to consider.
Technology companies benefit enormously from data center expansion.
Nvidia sells the chips.
Cloud companies sell AI computing.
Data center operators build enormous facilities.
Utilities sell electricity.
AI developers need more processing capacity.
That means companies in the industry have a strong financial interest in keeping new infrastructure projects moving.
So when X or other technology companies highlight foreign interference, the findings should be taken seriously—but also examined critically.
Business Insider noted that X and its owner Elon Musk have significant interests in expanding AI infrastructure and that foreign-influence claims can become a convenient explanation for domestic opposition.
Both things can be true at once.
China-linked operators may be attempting to manipulate the debate.
And real Americans may independently oppose data centers for legitimate reasons.
Treating one fact as proof against the other would be a mistake.
Kevin O’Leary Had Already Blamed China
Foreign interference became part of the public conversation even before X published its investigation.
Investor Kevin O’Leary had claimed that China was spending heavily to turn Americans against data centers and slow U.S. AI development.
Other technology figures have made similar accusations.
Until recently, critics argued there was limited public evidence supporting some of those broad claims.
OpenAI’s June report and X’s latest announcement provide more concrete evidence that at least some China-linked influence activity does exist.
But they still do not establish that the broader anti-data-center movement is primarily foreign-controlled.
Two hundred suspicious X accounts are important.
They are not the same thing as millions of American voters.
AI Makes Bot Farms Much Easier to Operate
The technology at the center of the political dispute also makes the influence operation easier.
Before generative AI, running hundreds of convincing fake accounts required enormous amounts of human labor.
Someone had to write posts.
Create images.
Respond to users.
Adjust tone.
Translate content.
Now AI can generate hundreds of variations almost instantly.
One prompt can produce:
A concerned-parent version.
A retired-worker version.
A local-business-owner version.
A political cartoon.
A short angry post.
A longer explanation.
A translated Chinese version.
A translated English version.
That makes fake grassroots campaigns cheaper and more scalable.
OpenAI’s investigation showed that its own models had been used to generate both written posts and manipulated images for the Data Center Bandwagon operation before the company banned the accounts.
This is an uncomfortable reality of generative AI.
The same tools that help legitimate users communicate can also help influence operators manufacture thousands of synthetic voices.
Platforms Are Now Using AI to Fight AI-Generated Propaganda
The response is becoming equally automated.
Platforms can analyze:
Posting patterns.
Account creation times.
IP addresses.
Language behavior.
Shared links.
Coordinated engagement.
Image similarities.
Automated interaction patterns.
One suspicious post may reveal very little.
Thousands of accounts behaving similarly can reveal a network.
OpenAI’s investigation used behavioral indicators including language, VPN usage, activity timing, and prompt patterns to identify the earlier operation.
X has not released equivalent technical detail publicly for its latest 200,000-account network, but large platforms have access to internal signals ordinary users cannot see.
That gives them a major advantage when identifying coordinated behavior.
The 200,000 Number Needs Context
The headline number is enormous.
200,000 accounts.
But only around 200 were identified by X as participating in the specific AI-data-center narrative.
That is roughly one-tenth of one percent of the larger suspected network.
X has not publicly explained what the remaining accounts were doing.
They may have been involved in different narratives.
Some may have existed mainly to amplify other accounts.
Others may have been dormant.
Without a complete report, we simply do not know.
So the most accurate headline is not:
“200,000 Chinese bots attacked American data centers.”
It is:
X says it found a 200,000-account suspected Chinese bot network, including about 200 accounts posting about U.S. AI data centers.
That difference matters.
The Real Debate Should Not Disappear
The worst outcome would be allowing the bot-farm story to replace the underlying policy discussion.
America genuinely needs to decide:
Who should pay for grid upgrades required by data centers?
Should ordinary customers absorb infrastructure costs?
How much water should enormous facilities be allowed to consume?
Should data centers be required to build their own generation?
How should communities be compensated?
How many jobs should projects create?
Where should facilities be built?
How should emissions be managed?
Those are legitimate questions regardless of who posts about them online.
Even OpenAI, while exposing the earlier campaign, explicitly described the discussion as a legitimate public debate.
The solution to foreign manipulation is not pretending criticism is illegitimate.
It is making sure Americans know who is actually speaking.
The Bigger Lesson
X’s announcement adds another piece of evidence that AI infrastructure has become part of geopolitical information warfare.
According to X, a suspected Chinese bot network contained about 200 accounts pushing anti-data-center narratives to American audiences.
OpenAI independently uncovered a similar China-linked campaign earlier this year.
That makes it increasingly difficult to dismiss the possibility of foreign attempts to influence the debate.
But the second half of the story matters just as much.
Foreign actors did not invent America’s data center problems.
Electricity systems are genuinely under strain.
Communities genuinely worry about costs and infrastructure.
Political opposition is genuinely growing.
The manipulation appears to work by attaching itself to those real concerns rather than creating them from nothing.
That is what makes this kind of campaign difficult to counter.
Deleting fake accounts is relatively easy.
Separating legitimate criticism from deliberately amplified outrage is much harder.
And as generative AI makes fake identities, political cartoons, comments, and entire online personalities cheaper to create, that problem is likely to become much bigger.
For now, the safest conclusion is straightforward:
There is growing evidence that China-linked operators have tried to influence America’s AI data center debate. There is not evidence that the broader American backlash against data centers is simply a foreign propaganda campaign.
Both realities can exist at the same time.
And understanding that distinction may be more important than anything the bots were actually posting.
